By the time it's fixed, we're half way to Z68 anyhow.. might as well wait for both CPU and mobo chipsets to mature.
I, for one, could not just use the 2 6Gbps ports. I had enough trouble finding Mobo's with 5 sockets and the # of HDs I'm getting just keeps increasing. There's no way I could only survive on 2 HDs, especially if 2 of them are going to be SSDs.
Again, I say, don't ask for the motherboards back, just send out a new one to people that either (1) have a problem down the line, or (2) can prove they have the B-stepping. Though, the idea of extremely affordable refurbished P67 motherboards makes me warm, to the point that I might say - who cares about Z68, I'll bite on a $70 P67.
If I can trade up to a z68 based motherboard.....that would be friggen awesome. I don't even care about this issue as I'm already using the 6GBs ports so I wouldn't be affected. However, I'd love to trade up if possible :D
Or if they didn't want to go with a 3rd party like Highpoint (although I've had great luck with them, even in servers) they could just subsidize their own RAID controllers:
Not quite. The P67 chipset has insufficient PCIe lanes. Having a bandwidth hungry PCIe cards means you can't run dual GPU without affecting peripheral performance such as USB3.0 or Gigabit LAN
The diagram in the article makes it look like the P67 chipset has 8 lanes of PCIe 2.0 by itself, and the 16 lanes that GPUs would use connect directly to the CPU.
So why then don't everyone connect their harddrives to the 6GBPS connection?
Is there any clarity if when connecting the harddrive to one 3GBPS which say would fail in 3 years, if the other ports are affected?
If not it'll be a matter of just switching SATA port every 3 years, good for another 15 years, by which time most likely the motherboard and processor can be dumped in the trash. PC hardware lasts no longer than 8 years before it underperforms too much.
what do you suppose the chances of many of these boards being turned into refurbs, with the only change being the 3Gbps connectors removed i.e. a defeatured board.
It will never happen, they will repair the board by isolating the voltage to the corrupt transistor. Refurbishing is a waste of silicon and consumer ends up with 2 6G ports for the cost of 2 6G + 4 3G ports.
"isolating the voltage to the corrupt transistor" means manufacturing a new chip where no voltage is sent to the transistor. Current chips can't be changed. They need to be replaced with new ones.
The affected transistor is just one of the million of transistors inside the small, many-legged chip used for I/O. And the problem is a lack of something, which can't be solved by nuking something with a pinpoint X-Ray or something.
I disagree. If the motherboard manufacturer could disable the 3Gbps ports at the BIOS level. Then they will never be used. The physical connectors would remain, though.
From a manufacturing point of view, they'll refurb the ones that haven't shipped. If you can do it on a large enough scale, reheating the board, lifting the old chip, and surface mounting the replacement will be cost effective. Especially with Intel fully covering or heavily subsidizing the costs.
I suspect that they'll just provide an extended replacement to those who already have a board.
Considering the 700 million it would likely be cheaper to just send out new chips any way. Changing the chipset shouldn't cost much. They can't sell boards and then disable the ports, and it would make no sense manufacturing with your current stock and still mount the connectors to the board, just to disable the ports. With lots of unhappy customers. Would make more sense waiting for new chips or redesigning the PCB and use some other controller for SATA-support.
But sounds like they would like to wait just for the chip to fail and replace it then. Just like when nvidiachips fail :)
I definitely wasn't suggesting Intel isn't going to release a new stepping of the chipset that fixes the flaw insite the chip. Of course they will. I mostly know this because they already said it. I was only responding to the point that it is not impossible to take an existing manufactured motherboard and permanently disable the SATA 3Gbps ports in the BIOS. If you don't need or use those ports, the flaw has zero affect on you. The main point is that it's an offsetting calculation by the actuaries who calculate the percentage of returns / exchanges that will be required.
Most of the boards are from OEM's, so Intel will just ship them a fixed version of the chipset and the OEM's will be in charge of either swapping the chipsets out or just putting the new chipsets on new boards. For the boards people bought directly from Intel, Intel will probably ship them a brand new board and then swap out the bad chipset and sell that board as a refurb later.
Most motherboards that are sold retail for around $100. That means the manufacturing cost, in toto, is somewhere around $35-50. The rest is packaging, shipping, advertising, and profit.
Shipping piles of boards (properly protected from ESD and physical damage, mind you) to China so that reflow-solder workers can remove the chipset, carefully clean the board, align and install the new one, reflow oven, test the newly made motherboard extensively, and ship them back... it's gonna be more expensive than simply dumping them in the landfill. Also, replacing the chipset is a major operation, and you would have to do it by hand rather than robotic placement, and I'm not convinced that the quality control wouldn't have to dump half the boards into the incinerator anyway.
Removing the broken ports at the distributors and then disabling them in the bios, though, that might be worthwhile -- except for the fact that 2 ports is just too few to be interesting. You could only really do it with OEM manufacturers -- the pallets worth of boards that are stacked up waiting to go into HPaqs and Dells. And you need 1 esata port, 1 ssd port, 1 storage port, and 1 optical port even in basic machines -- possibly omitting the ssd, but even then, 3 is more than two and no expansion at all tends to make consumers sad. Especially when it's the one upgrade people actually do sometimes, slinging an extra HD or Optical in there.
"... the best gesture of goodwill on Intel’s part would be to enable motherboard manufacturers to replace P67/H67 motherboards with Z68 boards..." That might be the only way for Intel to win hearts and minds.
Even so, I'm sure when the LGA 2011-based mobos launch later this year, it will make early adopters pause to think twice.
And RU482's got a point - mobo manufacturers might choose to refurb, given the very specific nature of the problem.
Ok, $700M over 8M products, I get the $87.50 there. But there's the round trip shipping cost, labor, and the bad stock which will be quite expensive, much more than $87.50, especially for laptops. I just ordered a Sager laptop with a Sandy Bridge CPU i7-2720QM. I can't imagine it'll be cheap to replace the motherboard on that thing. I just hope I can get better then UPS ground shipping because that would be a ten day round trip for me... :(
1) Many of those chips have probably not made it into motherboards yet, so it's a simple shipping and/or destruction cost. (I doubt they will destroy them, but who knows..) Quite simple to ask all the OEMs just to ship back or destroy the pallets/trays/however they are bulk packaged, I'm sure it won't cost 87.50 each to ship.
2) I'm sure there are many people that don't care/don't know that there is a potential problem, although that depends I suppose on how much (if any) mainstream media attention there is.
3) As Anand mentioned, there may be quite a few laptop chips that are unaffected that are probably included in that 8m number, many OEMs may have chosen to only implement the sataIII ports.
4) They may be lowballing the dollar amount to make things look better than they actually are. I remember reading somewhere that Intel set aside a certain amount for the floating point bug, but ended up spending a good bit more.
Intel would really really not like to have a million boards (15% of 8 million chips) losing SATA ports (even if three years from now). People have long memories for this kind of bad accidents, and might decide that Intel is no longer a dependable vendor. The reason Intel mainboards (Pentium II and !!! days) of old were bought was not feature set or performance or overclockability or something else - they were bought because they were rock solid, during the time when Anandtech testing of mainboard crashed them usually more than thrice in 48 hours. And I remember when the mainboards came that crashed in that burn test only when they used interleaved memory access, but not in non-interleaved memory access.
Any chips that aren't already in boards will go to landfill, and they probably won't even be shipped back to intel for it. The only reason for Intel to require them back is to verify that they are *really* destroyed, and not just labeled as such and then inserted into no-brand cheap motherboards.
$87.50 per chip on average is far from a lowball. That means they can easily be planning on spending $150 for ATX and even more per notebook that's affected.
As the article points out, laptops are highly unlikely to get covered / recalled / replaced at all. There's a very good chance that you would never be affected because most laptops don't use anything beyond the 0+1 ports - which are completely unaffected by this flaw.
And that is standard in the notebooks and notebook chipset? No, but the eSATA port of HM67/65 pretty much is. Then you also have lightpeak over copper, followed by fiber. Just saying. USB3 support don't rule out dead eSATAs. Professional laptos will always use SATA for the docking station too.
You're not going to be shipping your Saber laptop to Intel. You'll be returning it to the OEM. All Intel is doing is shipping out new chips to the OEMs.
But I run 6 sata devices, two hard drives in a raid 1, two independent hard drives, and two optical drives. My Asus P67 EVO has exactly 6 sata connections that are uneffected, if I string two of my devices out the case into the external ports. Not exactly the solution I paid for.
If reports are correct, you could just leave your optical drives on the "defective" ports, since the ports will simply stop working without any actual data/hardware damage. At least supposedly.
There might be data damage (certainly the odds go up if there are corrupted frames regularly, if the right bits get flipped error correction will miss it and you get actual data corruption instead of a dropped frame), but even if there is corruption on the bus to your dvd drive.. really, who cares.
Can we verify that there will not be actual media damage due to this error? If so, the urgency of this issue drops dramatically (keep using the board until the SATA II ports fail, then switch your drives over to SATA III or RMA the board).
Indeed, that's a good idea, other than it might stop working at a critical moment. But at least if you use the faulty ports, you'll assure yourself of being getting a non-defective motherboard eventually, when it fails and you qualify for warranty replacement. First it would be nice to see the motherboard manufacturers support their early adopters with a better-than-they-are-required-to-do replacement policy.
And thanks to Anand for being all over this story with some hard facts.
Intel says that the incidence of corrupted frames will get so high that there is significant impact to performance, because corrupted frames are detected and discarded. In the simple case of a CRC check, you get 8 bits and then another bit to checksum it, and if *two* bits flip in that one byte, you have an undetected error. In a similar way, it is always possible for both data and checksum to get corrupted in just such a way that you have an undetectable error (it's just usually a bit harder than in the simple example).
So with 100s of megabytes of corrupt packets per second going across your SATA bus, the odds of getting eventually an undetected error are probably not zero.
From what I heard, this recall only affects Sandy-Bridge-based desktop PCs. If you need a new performance PC during the next several weeks, you have three options:
1. Purchase a PC with an AMD high-end CPU (i.e. AMD Phenom II X6 1090T) 2. Purchase a PC with a Nehalem Core i5 or Core i7 CPU (i.e. Intel Core i7 930) 3. Purchase a Sandy-Bridge-based laptop PC (laptops are not affected by the recall since they use a mobile chipset type)
Nope, the notebooks use the same chipsets. The thinking is that most laptops won't support more then two SATA devices (a harddrive/SSD and optical drive) and as long as they only use the two SATA 6.0Gbps ports, they won't have any problems.
What I need to know is: Is this a 5% chance that the sata port will fail and you'll be out of luck for the next month. Or a 5% chance of silent data corruption. I have been editing customer photos on this machine and even a small chance of data corruption could cost me business.
About the data corruption; I assume that you're making backups regularly, to multiple destinations, and verifying that they're valid (the most important point?) If so, what do you have to worry, besides the downtime? If not, why aren't you properly backing up your company's data?
Because a read error will copy to the backup. How do I verify that they are valid? That would be extremely beneficial as I have thousands of pictures and over 500gb of files. Is there a program that scans each file and verifies that it is not corrupt?
My work flow goes like this (RAW files) Camera (memory card) -> Computer hard drive (convert files to adobe DNG) -> lightroom editing -> convert finished photos to JPEG -> Backup original DNG files, edited DNG files, finished jpegs -> format memory card
There are several points in this process that the main files or backup files could become corrupt from a bad read.
--- but now I am pulling the plug. I will not build a system with a known faulty chipset!
I need a new computer now, not in a few a months. This was going to be my first Intel build - thanks to this, I'm sticking with the tried and true AMD.
You could have just bought a P55 or X58 system and been perfectly happy. These are still much quicker than AMD options. Unless of course performance isn't your key concern.
Much as I love my X58, it doesn't make any kind of sense to dump ~$350 (minimum - $200 MC Nehalem special + mobo) into it when you can get a 2500K combo for $280 + tax at MC.
I'm in the same boat as you, but I'm purchasing an OEM system. You can get the AMD Phenom II X6 1100T or the Intel Core i7 980X for fast performance. Unless the SB systems make a comeback by the end of this month, I'm getting a PC with an AMD Phenom II X6 processor or perhaps an Intel Core i7 930 or 950.
Why exchange the chip when motherboard manufacterers can simply take the chip, put an additional sata controller to the board and use the 2 6G ports from the chip and the other 4 ports from the additional sata controller.
The mobo manufacturers can take the layout for boards with additional sata chip from their premium versions. - There is a much shorter delay, - The chips are not wasted and can be reused - It costs only a few $ for a sata chip and slightly different mobo instead of 87,50$ - Intel could even continue production of the chips, just selling them a few $ cheaper
Of course it only works for chips not already used on mobos yet.
The chips aren't worth anything. Any chips that have not made it into boards are already landfill, or will be by the end of the week. A chipset has a manufacturing cost to Intel of somewhere like 10 bucks at most.
87.50 is an average, and that includes the cost to replace motherboards in desktops and laptops, which can be a lot more than a measly hundred bucks.
Intel could always repurpose the old chips for notebooks and SFF PCs that only offer 2xSATA ports. Many countries have laws prohibiting the sale of such used/remanufactured items in new products, but they could always be used for something like charity notebooks for third-world countries... like a Sandy Bridge version of the OLPCs. ;)
"Intel claims that only 5-percent of Sandy Bridge owners will experience the flaw. But speaking with manufacturers, we were told that numbers appear to be far greater than that. Some companies told us that this is definitely a "panic-mode" situation for them."
It will be interesting to see how folks like Asus and Gigabyte will handle this. Will the end user be forced into the traditional method of replacing the motherboard, ie the end user sends off their board and waits for up to 8 weeks for the replacement?? If so, that would be very disappointing.
Boards should be swappable at the point of sale. Maybe Anand can ask Gary how Asus plan to handle this.
Within two days of my P67-based board running a SATA II port died. Plugged my DVD drive into another port and so-far so-good. Yes, it sucks that we're stuck in situation, but for me personally I wouldn't have a problem just getting an expansion card supporting four SATA II devices. I hate to see my board being scrapped for this, but I do also need a permanent resolution whether it be a new board or expansion card.
Sending you a SATA-card won't really be much cheaper then them repairing your board by RMA with a new chip from Intel which they don't pay for. Some work and shipping though. Which Intel might pay for any way. (This means sending you a new or refurbished board). But it's not like the business with motherboards sold as components retail are hugh. Would mean more problem with all the laptops with dead esata. It's usually a hell to get those repaired.
If a single SATA II port died, this is not related to the problem. You just got a dud board. The issue as described involves the PLL Clock circuit for all four ports simultaneously, so either all or none of them work properly.
I hope teaches intel next time not to abandon a tried and true socket design like the 1156. Now you can buy a sb, but have no motherboard to put it in!
Intel will pay roughly 1 billion before this is all done but this is not a big deal for consumers. Most chipsets are sold in laptops. Most laptops use SATA port 0 and 1 which are not affected. See my full story here, comments are welcome! www.myezweb.net
Makes you wonder how much power/heat/battery life could be improved by going through a ground up redesign or at least a comprehensive analysis of re-used circuitry.
I got an RMA for my mb last night. Originally they were deducting a restocking fee...lol. I chatted with a rep and they removed the restocking fee and emailed me a UPS shipping label gratis. Very nice service.
I hope you don't plan on sending it in right away (unless this is not a primary system for you). I just upgraded to SNB and already donated my old system to my parents. Intel should pay Newegg to send me a new board before I send mine to them... IMO
I'll be submitting my request for a new motherboard. This is quite bad. The transistor will short circuit and probably can cause damage to other components. I won't be surprised if they do a recall within the next month or two. Intel should have just announced a recall rather than lie to their customers.
learn to read...Why they use larger voltage than necessary? Energy efficiency is important.
They will not use a voltage level if not necessary, so it is not a fix without consequences. They are hiding something to save face, and don't should to be trusted.
I'm not sure what you think they are hiding, they said the only way to fix the problem, is to replace the chipset. They have stopped shipping the bad chips and for now have put aside $700M to deal with people who have the bad chips. While there are alot of guesses posted here as to how it will be handled, people are going to have to wait to see how each OEM and MB maker decided to deal with this. Samsung has already said they will offer refunds for PC's sold with the bad chipset.
I highly doubt they analyze and control the voltage applied to each transistor individually. If this explanation is true I would guess it is part of a bank of transistors that all receive the same voltage. That said it does sound weird that they would insert something into the design that was at least partly irrelevant.
Chips aren;t designed by humans, especially not at the transistor level. Humans assemble libraries of existing components.
Much like software, it's entirely possible for there to be pieces of code or transistor that have been unchanged since 1995 and sometimes aren;t even used any more.
regularly having to switch around sata ports when ssd drive not found in bios could these be related also my optical drive also not working for some time now mb msi P55 gd80' processor i5 655k' ssd kingston snvp 325s264gb, dvdcd gh22ls40 ata.My so called remedy is flash cmos when necessary including switching out sata cables from devices and ports anyway my dvd drive has not worked again ssd still ticking are these one in the same because of type of chip just wondering
Are we going to see SB motherboards on sale at newegg for $39? I can live with 2 fully functional ports for $39. All I need is one for an SSD and one for a DVD. Can use NAS for the rest.
I've said Intel shot itself in the foot with the sandy bridge release with the motherboard and I was right.
This is all caused by greed to make a new chipset only for customers to pay more money. They could have probably used 1156 pins for sandy if they really wanted to, but they decided to go for more profits and now it comes back to haunt them.
It would have been good if they also supported usb 3.0 and have a z68 boatherboard from the start.
Now they loose 1 billion and I hope it serves them as a lesson!
Come on AMD bring us your new processors sata 6gbps, usb 3.0 and SSD cahcing as well as better choices!
Yeah, bad Intel! Really, they should be even more consumer conscious and stop trying to take all our money away! Some people don't even have jobs! How are they going to decode blu-rays on a quad-core eight-threaded core i7 2000? I think Intel should just sell all their stuff at manufacturing costs so we can all have equal computer access. Anything else is just pure greed.
The easier fix for people that have a problem in the field (aside from notebook users who would likely never see the problem) would be to just ship those customers a PCIe SATA card. Less down time, less cost, faster solution to a happy customer.
"These are the same tests that all Intel chipsets must go through, testing things like functionality, reliability and behavior at various conditions (high temps, load temps, high voltage, low voltage, etc...). The chipset made it through all of these tests just fine."
A design will go through "4 corner tests" - hi-temp/hi-voltage, hi-temp/lo-voltage, lo-temp//lo-voltage, lo-temp/lo-volts. A number of samples are tested at high throughput for a few hours at each corner (with maybe a few more "corners" thrown in). You repeat this test with samples from 'fast' 'slow' and nominal process corners. This doesn't take too long but it verifies that the design is robust over a set of extreme conditions but for a short period of time
Then there's High Temperature Operating Life (HTOL) Testing ( google WHAT IS HTOL ), or accelerated life testing. We (where I work, NOT Intel) usually take a larger number of samples and operate them at high temperature for a couple of weeks in an oven, perhaps at high voltage. This takes a lot longer. We usually do it after the 4-corner tests are complete.
I guess that if you release a chip before HTOL tests are complete, you risk losing a billion dollars.
Earlier samples didn't have the problem. Revisions don't need to be proven same as an new design so it should be quite easy sneaking in new problems there.
Here's a question. Just got the i7-2600 last week to build new unit. Still in box. Anyone know of a way to send it for replacement BEFORE installation?
I would suggest you just go ahead and build your system. Use the SATA6G ports first. If you need more than two, add a $20 expansion card for more SATA ports. Enjoy the system and when the replacement info comes out, worry about it then.
I just picked up Sandy Bridge components last Friday. Assembled the system on Sunday. Intel can pry it from my cold, dead hands. It folds proteins like a demon.
Why sit on your very fast parts for a month (realistically it will take that long for a new motherboard)? It is really a minor issue, now that we know what the problem is.
If it's a transistor that fails and causes a problem, but it isn't important and not used anyway so the fix is to disable it, what makes it a problem if it "disables itself" by failing?
About a week ago, I ordered a Vostro 460 desktop with a new Sandy Bridge processor. When the news came out about the recall yesterday, I called Dell to see whether my order would be delayed. The tech told me that my expected shipping date had been postponed a week, but did not have any details about the recall. Then, my order unexpectedly shipped this morning.
I guess Dell was not effected by the recall; maybe they were only using the SATA 6GB ports or maybe they were using the A stepping or maybe they just didn't care and shipped the broken chipset. Either way, I'm not going to take any chances: I am going to ensure that the two hard drives (one SSD) are installed on the 6GB sata ports. To make room, I will need to move the CD drive to one of the 3GB ports because it is used less so it should be less affected by the bug.
HP doesn't really suck. Their Pavilion PCs use quality parts and are built to last, not to mention energy-efficient (probably why they use 300W PSUs). If you order an SB unit from HP.com, the estimated build date will just be delayed until March 1 to allow HP time to receive the updated hardware.
That's weird. Dell has removed the XPS 8300 from its website and is not available for retail at electronics stores either. For those of you who have not ordered already, you can purchase the AMD-based XPS 7100 or Nehalem Core-based XPS 9100 though.
I am a little confused about how an upgrade offer to Z68 would be a show of goodwill to affected users. Having just built my new P67 rig with an 1155 socket processor, I don't see how that would help me. Is Intel going to offer me a 2011 socket processor and a triple channel memory kit to go along with this new Z68 board that they may (or may not) offer me? Anand, why do you suggest that will be a nice offer to current P67 users.
Another reason to skip first gen, I learned my lesson with the first batch of Pentium 3's.
I still want a Sandy Bridge, perhaps they'll release another version chipset for the current i5/i7 Sandy Bridges, right now the AMD X6 are a waste of money since even the latest X4's still has better overall performance, I'm happy with my Phenom II X4 955BE but the i5-2500K costs less than the X6 1090T and practically STOMPS all over the X6 1100T.
Can't beat the cost of the X3 740 BE combos I got from newEgg during Black Friday. IIRC, ~$130 for the CPU + mobo. All 6 of them unlocked to 4 cores and OCed to 3.7GHz.
My main is a X58 + i7 920 @ 4.1GHz, but damn! if those AMDs aren't cost effective o.O
Just think if all motherboards had a 15 cent socket installed for the chipset . This new found flaw wouldn't be that big of a deal, and would cost a fraction of the one billion Intel is going to shell out (probably would only cost a few million).
Intel could send you a small box with a new chip and small plastic tool to pop the old chip out and then insert the new one. Problem fixed. No RMA, very little waste, and down time would be extremely short for most for most people.
A socket for a high density large pin count BGA costs a whole lot more than 15 cents. You could very well expect to see your MOBO cost 12-16$ more at the retail level if it accomodated a chip set socket.
Don't forget the extra real estate for the chip (board area), the heatsink problems, the installation cost, and the fact that you won't be able to use large GPUs any more.
I think it will be easier and cheaper to replace the 4 chipset SATA300 ports with a PCIe card with 4 SATA ports for existing systems. MB makers can now re-design to add 3rd party PCIe to SATA chip to MB and remove the Intel 4 ports. Can we reduce the voltage to the chipset to reduce the risk of failure ? No overclocking allowed ?
They have to bullshit us with some kind of explanation. Intel and Anandtech is trying to put the cherry on a turd and sugarcoat this, but the truth is that the real situation is far worse! How do I know? Well, I know the same way you know: they had to publicly disclose this to avoid big expensive - public & humiliating class action law suits... it would have been a PR nightmare. Kind of the same way NVidia shipped those bad notebook GPUs and then got sued by a a bunch of people. Lucky me, I had an ASUS notebook with one of those turds. Anyway, Intel sucks for being this stupid... I wish there was a way to went from this, but I'm fuming with frustration over this.
This is only the outbound loss. I want to know the internal cost as well, how much to spin out 8million new chipsets and then just hand them out for free?
Ok, just to be clear, if I do buy a p67 or h67 and ONLY USE THE BLUE SATA-6 ports, will I ever get to see the problem? I don't intend to run more then 2 hard drives anyway, for now I'm just sticking to a sata II 1.5tb 64mb cash western digital on my rig. Am I fine?
Will you be fine? Well, that depends on what you eat, if you work out and look around as you cross the street... but actually, no you wont be fine. You are going to die.
Woohoo!! Bought my new Sandy Bridge Build on the 29th just before Intel announced the recall. Supplier send me a refund on the motherboard but I still have all the other parts sent to me which are now completely useless for two months as I cant get hold of a p67 Board anywhere. So I am now sat with £750 worth of components (which will depreciate in value) without the most important part unavailable and April seems to be the earliest that new boards will be released. So a nice two month wait till I can even start building my comp.
To make it worse I agreed to sell my old computer for a lot more than it was worth which I was using to help offset the cost of the build and now that wont happen.
Gnomey, I'm sure you can get the Z6* boards soon, or maybe your supplier can do a swap deal with you. Hang onto your parts as it shouldn't be too long (1 month maybe). I'm waiting myself to get a Sandy Bridge setup.
i have this motherboard the computer frezzes up every 5 mins i try to transfer files i keep getting errors cant read from hdd this bug happens all the time not in 3 years
I think between Intel and the board manufacturers, they should just send us all new boards without even having to rip out our old ones to send back. They did this with my Logitech speaker system that had the snap,crackle & pop in the audio and paid well over $100 dollars for. (Now have 2 speaker systems) A proof of purchasing slip should be all that is required and send us all new and possibly updated versions. And in the end there would still be frosting on the cake for Intel, as Im sure alot of people would buy new chips to go with their new boards!
Instead of repairing these motherboards, they should be re-purposed for schools or non profit organizations (maybe even the THE INTEL COMPUTER CLUBHOUSE NETWORK). Let the 2 6gb sata ports be used with some notice (sticker?) Its not worth the $$$ for Intel to deal with. Thats my $.02
OK, 4-5 years on from this, am I now starting to see an issue? I have a Tyan S5510 server board that uses the Couger C204 chipset. A while back I ended up with some data corruption on drives that were connected to the SATA connector, but it was all very sporadic. I thought the drives might be failing, and although I replaced them (as a mater of course), I have since tested them with various diagnostic utilities on other systems and couldn't find anything actually wrong with them.
The drives I used as replacements were OK for a while, but again this week, one started to get odd errors, and it was indeed connected to one of the 3Gbps ports.
Hi guys, very in-depth article. I own a Toshba Qosmio X770 laptop (UK model). It comes with HM65 chipset. CPU-z calls it: Intel Sandy Bridge Rev 09 - - Intel HM65 Rev B2
And Device Manager displays the dreaded: Intel(R) 6 Series/C200 Series Chipset family PCI Express Root Port 1 (1C10), Port 2 (1C12), Port 4 (1C16) and Port 6 (1C1A).
This is my 4th year running this. I am glad this intel-defect does not affect my USB3 port (which I have only one). There are no eSata ports on my laptop. I have an SSD (primary), a 7200rpm HDD (secondary) and a DVD-RW drive.
From what I understand, Toshiba engineers should only use Sata3 ports 0 and 1 because the Sata2 ports (2-5) will cause problems. (1) If my primary hard-drive (Samsung Evo830 SSD) is connected to port 0, then why am I only getting Sata2 speeds (250-300 MB/s read and write)? All benchmark utilities I have used report similar speeds (Windows power settings set to Highest). They also report that SSD is on Sata3 (iaStor - Ok) and the secondary hard-drive is on a Sata-2 port. I conducted the SSD test on another laptop and the speeds reported were 520MB/s R/W seq. So the SSD is not faulty. (2) If SATA3 ports 0 were soldered to primary hard-drives and SATA3 port 1 to the DVD drive, does this mean that my secondary hard-drive is on a faulty SATA2 port? (3) I wish there was a software which could gather this information and tell me who is connected to what on the HM65 sata ports. Is there one? Maybe a hacking utility, which could enable Sata-3 speeds for my SSD (if in deed it is connected to SATA3 PORT 0). (4) Can I swap HM65 with a non-faulty B3 chipset or HM70? Thanks for any answers. Best regards.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
127 Comments
Back to Article
MonkeyPaw - Monday, January 31, 2011 - link
"I always screw up some mundane detail!"InterClaw - Monday, January 31, 2011 - link
"Oh, what is this fairly mundane detail, Michael!?"vol7ron - Monday, January 31, 2011 - link
By the time it's fixed, we're half way to Z68 anyhow.. might as well wait for both CPU and mobo chipsets to mature.I, for one, could not just use the 2 6Gbps ports. I had enough trouble finding Mobo's with 5 sockets and the # of HDs I'm getting just keeps increasing. There's no way I could only survive on 2 HDs, especially if 2 of them are going to be SSDs.
Again, I say, don't ask for the motherboards back, just send out a new one to people that either (1) have a problem down the line, or (2) can prove they have the B-stepping. Though, the idea of extremely affordable refurbished P67 motherboards makes me warm, to the point that I might say - who cares about Z68, I'll bite on a $70 P67.
samspqr - Tuesday, February 1, 2011 - link
if I was in charge of this at intel, I'd handle this as follows:offer every SNB motherboard owner to either:
* accept a $40 discount on their recently-bought motherboard, use only two SATA ports, and call it a day
* send the board in (for a chip swap, or for a full motherboard swap, whatever comes out cheaper)
aaron88_7 - Tuesday, February 1, 2011 - link
If I can trade up to a z68 based motherboard.....that would be friggen awesome. I don't even care about this issue as I'm already using the 6GBs ports so I wouldn't be affected. However, I'd love to trade up if possible :DSamus - Tuesday, February 1, 2011 - link
For $87.50 couldn't they can just buy everyone a decent PCIe RAID controller?http://www.newegg.com/Product/Product.aspx?Item=N8...
Or if they didn't want to go with a 3rd party like Highpoint (although I've had great luck with them, even in servers) they could just subsidize their own RAID controllers:
http://www.newegg.com/Product/Product.aspx?Item=N8...
I'm sure *most* people wouldn't complain as the performance and features are probably greater than the onboard 4 ports anyway.
AnnihilatorX - Tuesday, February 1, 2011 - link
Not quite. The P67 chipset has insufficient PCIe lanes. Having a bandwidth hungry PCIe cards means you can't run dual GPU without affecting peripheral performance such as USB3.0 or Gigabit LANMercAB3 - Wednesday, February 2, 2011 - link
The diagram in the article makes it look like the P67 chipset has 8 lanes of PCIe 2.0 by itself, and the 16 lanes that GPUs would use connect directly to the CPU.oritpro - Thursday, February 3, 2011 - link
Unfortunately If you populate both x16 slots they run at x8.ProDigit - Saturday, February 5, 2011 - link
So why then don't everyone connect their harddrives to the 6GBPS connection?Is there any clarity if when connecting the harddrive to one 3GBPS which say would fail in 3 years, if the other ports are affected?
If not it'll be a matter of just switching SATA port every 3 years, good for another 15 years, by which time most likely the motherboard and processor can be dumped in the trash.
PC hardware lasts no longer than 8 years before it underperforms too much.
RU482 - Monday, January 31, 2011 - link
what do you suppose the chances of many of these boards being turned into refurbs, with the only change being the 3Gbps connectors removed i.e. a defeatured board.andrewsolid - Monday, January 31, 2011 - link
It will never happen, they will repair the board by isolating the voltage to the corrupt transistor. Refurbishing is a waste of silicon and consumer ends up with 2 6G ports for the cost of 2 6G + 4 3G ports.yuhong - Monday, January 31, 2011 - link
They can't do this from the outside the chip.andrewsolid - Monday, January 31, 2011 - link
Can u pls explain that?MamiyaOtaru - Monday, January 31, 2011 - link
"isolating the voltage to the corrupt transistor" means manufacturing a new chip where no voltage is sent to the transistor. Current chips can't be changed. They need to be replaced with new ones.andrewsolid - Tuesday, February 1, 2011 - link
Thank you,"The source of the problem is actually not even a key part of the 6-series chipset design, it’s remnant of an earlier design that’s no longer needed."
An useless piece costs them 1 Billion dollars......This sucks..
Calin - Tuesday, February 1, 2011 - link
The affected transistor is just one of the million of transistors inside the small, many-legged chip used for I/O. And the problem is a lack of something, which can't be solved by nuking something with a pinpoint X-Ray or something.Focher - Monday, January 31, 2011 - link
I disagree. If the motherboard manufacturer could disable the 3Gbps ports at the BIOS level. Then they will never be used. The physical connectors would remain, though.From a manufacturing point of view, they'll refurb the ones that haven't shipped. If you can do it on a large enough scale, reheating the board, lifting the old chip, and surface mounting the replacement will be cost effective. Especially with Intel fully covering or heavily subsidizing the costs.
I suspect that they'll just provide an extended replacement to those who already have a board.
Penti - Monday, January 31, 2011 - link
Considering the 700 million it would likely be cheaper to just send out new chips any way. Changing the chipset shouldn't cost much. They can't sell boards and then disable the ports, and it would make no sense manufacturing with your current stock and still mount the connectors to the board, just to disable the ports. With lots of unhappy customers. Would make more sense waiting for new chips or redesigning the PCB and use some other controller for SATA-support.But sounds like they would like to wait just for the chip to fail and replace it then. Just like when nvidiachips fail :)
Focher - Tuesday, February 1, 2011 - link
I definitely wasn't suggesting Intel isn't going to release a new stepping of the chipset that fixes the flaw insite the chip. Of course they will. I mostly know this because they already said it. I was only responding to the point that it is not impossible to take an existing manufactured motherboard and permanently disable the SATA 3Gbps ports in the BIOS. If you don't need or use those ports, the flaw has zero affect on you. The main point is that it's an offsetting calculation by the actuaries who calculate the percentage of returns / exchanges that will be required.cjb110 - Tuesday, February 1, 2011 - link
That's a good idea...better than the boards ending up in landfill.Remove the broken ports, sell it as a 2 port board (quite a few pc's only have 2 drives).
But in actuality what will happen? Is it viable to 'remake' the boards with components replaced?
cbass64 - Tuesday, February 1, 2011 - link
Most of the boards are from OEM's, so Intel will just ship them a fixed version of the chipset and the OEM's will be in charge of either swapping the chipsets out or just putting the new chipsets on new boards. For the boards people bought directly from Intel, Intel will probably ship them a brand new board and then swap out the bad chipset and sell that board as a refurb later.JasperJanssen - Tuesday, February 1, 2011 - link
Most motherboards that are sold retail for around $100. That means the manufacturing cost, in toto, is somewhere around $35-50. The rest is packaging, shipping, advertising, and profit.Shipping piles of boards (properly protected from ESD and physical damage, mind you) to China so that reflow-solder workers can remove the chipset, carefully clean the board, align and install the new one, reflow oven, test the newly made motherboard extensively, and ship them back... it's gonna be more expensive than simply dumping them in the landfill. Also, replacing the chipset is a major operation, and you would have to do it by hand rather than robotic placement, and I'm not convinced that the quality control wouldn't have to dump half the boards into the incinerator anyway.
Removing the broken ports at the distributors and then disabling them in the bios, though, that might be worthwhile -- except for the fact that 2 ports is just too few to be interesting. You could only really do it with OEM manufacturers -- the pallets worth of boards that are stacked up waiting to go into HPaqs and Dells. And you need 1 esata port, 1 ssd port, 1 storage port, and 1 optical port even in basic machines -- possibly omitting the ssd, but even then, 3 is more than two and no expansion at all tends to make consumers sad. Especially when it's the one upgrade people actually do sometimes, slinging an extra HD or Optical in there.
Zoomer - Wednesday, February 2, 2011 - link
There's the ide port which can be used for the optical drive. So 2 ports, no esata, sounds good.Phylyp - Monday, January 31, 2011 - link
"... the best gesture of goodwill on Intel’s part would be to enable motherboard manufacturers to replace P67/H67 motherboards with Z68 boards..."That might be the only way for Intel to win hearts and minds.
Even so, I'm sure when the LGA 2011-based mobos launch later this year, it will make early adopters pause to think twice.
And RU482's got a point - mobo manufacturers might choose to refurb, given the very specific nature of the problem.
htwingnut - Monday, January 31, 2011 - link
Ok, $700M over 8M products, I get the $87.50 there. But there's the round trip shipping cost, labor, and the bad stock which will be quite expensive, much more than $87.50, especially for laptops. I just ordered a Sager laptop with a Sandy Bridge CPU i7-2720QM. I can't imagine it'll be cheap to replace the motherboard on that thing. I just hope I can get better then UPS ground shipping because that would be a ten day round trip for me... :(nikclev - Monday, January 31, 2011 - link
I'm sure they are factoring in several things:1) Many of those chips have probably not made it into motherboards yet, so it's a simple shipping and/or destruction cost. (I doubt they will destroy them, but who knows..) Quite simple to ask all the OEMs just to ship back or destroy the pallets/trays/however they are bulk packaged, I'm sure it won't cost 87.50 each to ship.
2) I'm sure there are many people that don't care/don't know that there is a potential problem, although that depends I suppose on how much (if any) mainstream media attention there is.
3) As Anand mentioned, there may be quite a few laptop chips that are unaffected that are probably included in that 8m number, many OEMs may have chosen to only implement the sataIII ports.
4) They may be lowballing the dollar amount to make things look better than they actually are. I remember reading somewhere that Intel set aside a certain amount for the floating point bug, but ended up spending a good bit more.
nikclev - Monday, January 31, 2011 - link
Forgot to add: Maybe the plan to offset the cost somehow... Maybe sell them as keychains in the Intel gift shop!Calin - Tuesday, February 1, 2011 - link
Intel would really really not like to have a million boards (15% of 8 million chips) losing SATA ports (even if three years from now). People have long memories for this kind of bad accidents, and might decide that Intel is no longer a dependable vendor.The reason Intel mainboards (Pentium II and !!! days) of old were bought was not feature set or performance or overclockability or something else - they were bought because they were rock solid, during the time when Anandtech testing of mainboard crashed them usually more than thrice in 48 hours. And I remember when the mainboards came that crashed in that burn test only when they used interleaved memory access, but not in non-interleaved memory access.
JasperJanssen - Tuesday, February 1, 2011 - link
Any chips that aren't already in boards will go to landfill, and they probably won't even be shipped back to intel for it. The only reason for Intel to require them back is to verify that they are *really* destroyed, and not just labeled as such and then inserted into no-brand cheap motherboards.$87.50 per chip on average is far from a lowball. That means they can easily be planning on spending $150 for ATX and even more per notebook that's affected.
Focher - Monday, January 31, 2011 - link
As the article points out, laptops are highly unlikely to get covered / recalled / replaced at all. There's a very good chance that you would never be affected because most laptops don't use anything beyond the 0+1 ports - which are completely unaffected by this flaw.Penti - Monday, January 31, 2011 - link
eSATA just saying.bigboxes - Tuesday, February 1, 2011 - link
When USB3 is fully implemented by mobo mfgs you can say goodbye to eSATA. Just saying.Penti - Tuesday, February 1, 2011 - link
And that is standard in the notebooks and notebook chipset? No, but the eSATA port of HM67/65 pretty much is. Then you also have lightpeak over copper, followed by fiber. Just saying. USB3 support don't rule out dead eSATAs. Professional laptos will always use SATA for the docking station too.Squuiid - Tuesday, February 1, 2011 - link
Say Goodbye to any eSATA ports on a laptop though...cbass64 - Tuesday, February 1, 2011 - link
You're not going to be shipping your Saber laptop to Intel. You'll be returning it to the OEM. All Intel is doing is shipping out new chips to the OEMs.EdgeOfDetroit - Monday, January 31, 2011 - link
But I run 6 sata devices, two hard drives in a raid 1, two independent hard drives, and two optical drives. My Asus P67 EVO has exactly 6 sata connections that are uneffected, if I string two of my devices out the case into the external ports. Not exactly the solution I paid for.jimhsu - Monday, January 31, 2011 - link
If reports are correct, you could just leave your optical drives on the "defective" ports, since the ports will simply stop working without any actual data/hardware damage. At least supposedly.JasperJanssen - Tuesday, February 1, 2011 - link
There might be data damage (certainly the odds go up if there are corrupted frames regularly, if the right bits get flipped error correction will miss it and you get actual data corruption instead of a dropped frame), but even if there is corruption on the bus to your dvd drive.. really, who cares.jimhsu - Monday, January 31, 2011 - link
Can we verify that there will not be actual media damage due to this error? If so, the urgency of this issue drops dramatically (keep using the board until the SATA II ports fail, then switch your drives over to SATA III or RMA the board).EdgeOfDetroit - Monday, January 31, 2011 - link
Indeed, that's a good idea, other than it might stop working at a critical moment. But at least if you use the faulty ports, you'll assure yourself of being getting a non-defective motherboard eventually, when it fails and you qualify for warranty replacement. First it would be nice to see the motherboard manufacturers support their early adopters with a better-than-they-are-required-to-do replacement policy.And thanks to Anand for being all over this story with some hard facts.
JasperJanssen - Tuesday, February 1, 2011 - link
Intel says that the incidence of corrupted frames will get so high that there is significant impact to performance, because corrupted frames are detected and discarded. In the simple case of a CRC check, you get 8 bits and then another bit to checksum it, and if *two* bits flip in that one byte, you have an undetected error. In a similar way, it is always possible for both data and checksum to get corrupted in just such a way that you have an undetectable error (it's just usually a bit harder than in the simple example).So with 100s of megabytes of corrupt packets per second going across your SATA bus, the odds of getting eventually an undetected error are probably not zero.
banzaigtv - Monday, January 31, 2011 - link
From what I heard, this recall only affects Sandy-Bridge-based desktop PCs. If you need a new performance PC during the next several weeks, you have three options:1. Purchase a PC with an AMD high-end CPU (i.e. AMD Phenom II X6 1090T)
2. Purchase a PC with a Nehalem Core i5 or Core i7 CPU (i.e. Intel Core i7 930)
3. Purchase a Sandy-Bridge-based laptop PC (laptops are not affected by the recall since they use a mobile chipset type)
danjw - Monday, January 31, 2011 - link
Nope, the notebooks use the same chipsets. The thinking is that most laptops won't support more then two SATA devices (a harddrive/SSD and optical drive) and as long as they only use the two SATA 6.0Gbps ports, they won't have any problems.LeftSide - Monday, January 31, 2011 - link
What I need to know is: Is this a 5% chance that the sata port will fail and you'll be out of luck for the next month. Or a 5% chance of silent data corruption. I have been editing customer photos on this machine and even a small chance of data corruption could cost me business.Jovec - Monday, January 31, 2011 - link
I don't Intel is going to say anything concrete to limit any potential liability.jimhsu - Monday, January 31, 2011 - link
About the data corruption; I assume that you're making backups regularly, to multiple destinations, and verifying that they're valid (the most important point?) If so, what do you have to worry, besides the downtime? If not, why aren't you properly backing up your company's data?LeftSide - Tuesday, February 1, 2011 - link
Because a read error will copy to the backup. How do I verify that they are valid? That would be extremely beneficial as I have thousands of pictures and over 500gb of files. Is there a program that scans each file and verifies that it is not corrupt?My work flow goes like this (RAW files)
Camera (memory card) -> Computer hard drive (convert files to adobe DNG) -> lightroom editing -> convert finished photos to JPEG -> Backup original DNG files, edited DNG files, finished jpegs -> format memory card
There are several points in this process that the main files or backup files could become corrupt from a bad read.
Andreos - Monday, January 31, 2011 - link
--- but now I am pulling the plug. I will not build a system with a known faulty chipset!I need a new computer now, not in a few a months. This was going to be my first Intel build - thanks to this, I'm sticking with the tried and true AMD.
mapesdhs - Tuesday, February 1, 2011 - link
You could have just bought a P55 or X58 system and been perfectly happy. These
are still much quicker than AMD options. Unless of course performance isn't your
key concern.
Ian.
JasperJanssen - Tuesday, February 1, 2011 - link
Yes, spend too much for an X58 system that is already outdated, great idea.WhoBeDaPlaya - Tuesday, February 1, 2011 - link
Much as I love my X58, it doesn't make any kind of sense to dump ~$350 (minimum - $200 MC Nehalem special + mobo) into it when you can get a 2500K combo for $280 + tax at MC.cactusdog - Saturday, February 5, 2011 - link
It also doesnt make sense to buy a P67 board when Z68 are coming out soon. The true overclocking boards for the socket 1155 platform.banzaigtv - Wednesday, February 2, 2011 - link
I'm in the same boat as you, but I'm purchasing an OEM system. You can get the AMD Phenom II X6 1100T or the Intel Core i7 980X for fast performance. Unless the SB systems make a comeback by the end of this month, I'm getting a PC with an AMD Phenom II X6 processor or perhaps an Intel Core i7 930 or 950.bavaria2 - Monday, January 31, 2011 - link
Why exchange the chip when motherboard manufacterers can simply take the chip, put an additional sata controller to the board and use the 2 6G ports from the chip and the other 4 ports from the additional sata controller.The mobo manufacturers can take the layout for boards with additional sata chip from their premium versions.
- There is a much shorter delay,
- The chips are not wasted and can be reused
- It costs only a few $ for a sata chip and slightly different mobo instead of 87,50$
- Intel could even continue production of the chips, just selling them a few $ cheaper
Of course it only works for chips not already used on mobos yet.
JasperJanssen - Tuesday, February 1, 2011 - link
The chips aren't worth anything. Any chips that have not made it into boards are already landfill, or will be by the end of the week. A chipset has a manufacturing cost to Intel of somewhere like 10 bucks at most.87.50 is an average, and that includes the cost to replace motherboards in desktops and laptops, which can be a lot more than a measly hundred bucks.
CZroe - Monday, January 31, 2011 - link
Intel could always repurpose the old chips for notebooks and SFF PCs that only offer 2xSATA ports. Many countries have laws prohibiting the sale of such used/remanufactured items in new products, but they could always be used for something like charity notebooks for third-world countries... like a Sandy Bridge version of the OLPCs. ;)Touche - Monday, January 31, 2011 - link
"Intel claims that only 5-percent of Sandy Bridgeowners will experience the flaw. But speaking with
manufacturers, we were told that numbers appear
to be far greater than that. Some companies told us
that this is definitely a "panic-mode" situation for
them."
www.tomshardware.com/news/sandy-bridge-sata-error-sata-3,12112.html
cactusdog - Monday, January 31, 2011 - link
It will be interesting to see how folks like Asus and Gigabyte will handle this. Will the end user be forced into the traditional method of replacing the motherboard, ie the end user sends off their board and waits for up to 8 weeks for the replacement?? If so, that would be very disappointing.Boards should be swappable at the point of sale. Maybe Anand can ask Gary how Asus plan to handle this.
travman733 - Monday, January 31, 2011 - link
Within two days of my P67-based board running a SATA II port died. Plugged my DVD drive into another port and so-far so-good. Yes, it sucks that we're stuck in situation, but for me personally I wouldn't have a problem just getting an expansion card supporting four SATA II devices. I hate to see my board being scrapped for this, but I do also need a permanent resolution whether it be a new board or expansion card.Penti - Monday, January 31, 2011 - link
Sending you a SATA-card won't really be much cheaper then them repairing your board by RMA with a new chip from Intel which they don't pay for. Some work and shipping though. Which Intel might pay for any way. (This means sending you a new or refurbished board). But it's not like the business with motherboards sold as components retail are hugh. Would mean more problem with all the laptops with dead esata. It's usually a hell to get those repaired.JasperJanssen - Tuesday, February 1, 2011 - link
If a single SATA II port died, this is not related to the problem. You just got a dud board. The issue as described involves the PLL Clock circuit for all four ports simultaneously, so either all or none of them work properly.Pneumothorax - Monday, January 31, 2011 - link
I hope teaches intel next time not to abandon a tried and true socket design like the 1156. Now you can buy a sb, but have no motherboard to put it in!myEZweb_net - Monday, January 31, 2011 - link
Intel will pay roughly 1 billion before this is all done but this is not a big deal for consumers.Most chipsets are sold in laptops. Most laptops use SATA port 0 and 1 which are not affected. See my full story here, comments are welcome! www.myezweb.net
redisnidma - Monday, January 31, 2011 - link
I just hope Anand defends AMD the way he does so passionately for intel. ;)Starcub - Monday, January 31, 2011 - link
Makes you wonder how much power/heat/battery life could be improved by going through a ground up redesign or at least a comprehensive analysis of re-used circuitry.Gothmoth - Tuesday, February 1, 2011 - link
what do you think would a complete redesign cost?people want everthing as cheap as possible.
if you want to safe engery then tell the americans not to use aircondition in every room and with open windows......
chicagotechjunkie - Monday, January 31, 2011 - link
Facts, from the source, with details the tech community appreciate.This site, and specifically you, make the internet a better place.
Thank you.
Tigerwraith - Monday, January 31, 2011 - link
Newegg has pulled all 1155 mb and cpus.ratbert1 - Tuesday, February 1, 2011 - link
I got an RMA for my mb last night. Originally they were deducting a restocking fee...lol. I chatted with a rep and they removed the restocking fee and emailed me a UPS shipping label gratis. Very nice service.mike2100 - Tuesday, February 1, 2011 - link
I hope you don't plan on sending it in right away (unless this is not a primary system for you).I just upgraded to SNB and already donated my old system to my parents. Intel should pay Newegg to send me a new board before I send mine to them... IMO
ghitz - Tuesday, February 1, 2011 - link
Agreed. I'm not going to be without a system even for a 1 day because of this. My new machine is my daily production machine.glad2meetu - Tuesday, February 1, 2011 - link
I'll be submitting my request for a new motherboard. This is quite bad. The transistor will short circuit and probably can cause damage to other components. I won't be surprised if they do a recall within the next month or two. Intel should have just announced a recall rather than lie to their customers.marraco - Tuesday, February 1, 2011 - link
If the mother is fixed by running on a lower voltage: ¿Why they used a higher voltage? ¿To waste energy?This looks like a lame excuse. I don't trust.
I will no buy P67/H67.
Gothmoth - Tuesday, February 1, 2011 - link
learn to read.... it was an oversight by the engineers.you people can not even read a simple article but complain about engineers who have to oversee a billion transistor CPU..
marraco - Tuesday, February 1, 2011 - link
learn to read...Why they use larger voltage than necessary? Energy efficiency is important.They will not use a voltage level if not necessary, so it is not a fix without consequences. They are hiding something to save face, and don't should to be trusted.
bruce24 - Tuesday, February 1, 2011 - link
I'm not sure what you think they are hiding, they said the only way to fix the problem, is to replace the chipset. They have stopped shipping the bad chips and for now have put aside $700M to deal with people who have the bad chips. While there are alot of guesses posted here as to how it will be handled, people are going to have to wait to see how each OEM and MB maker decided to deal with this. Samsung has already said they will offer refunds for PC's sold with the bad chipset.strikeback03 - Tuesday, February 1, 2011 - link
I highly doubt they analyze and control the voltage applied to each transistor individually. If this explanation is true I would guess it is part of a bank of transistors that all receive the same voltage. That said it does sound weird that they would insert something into the design that was at least partly irrelevant.JasperJanssen - Tuesday, February 1, 2011 - link
Chips aren;t designed by humans, especially not at the transistor level. Humans assemble libraries of existing components.Much like software, it's entirely possible for there to be pieces of code or transistor that have been unchanged since 1995 and sometimes aren;t even used any more.
catommy - Tuesday, February 1, 2011 - link
regularly having to switch around sata ports when ssd drive not found in bios could these be related also my optical drive also not working for some time now mb msi P55 gd80' processor i5 655k' ssd kingston snvp 325s264gb, dvdcd gh22ls40 ata.My so called remedy is flash cmos when necessary including switching out sata cables from devices and ports anyway my dvd drive has not worked again ssd still ticking are these one in the same because of type of chip just wonderingJasperJanssen - Tuesday, February 1, 2011 - link
No, this cannot be related. The bug was introduced in the B sepping of the latest chipset and certainly hasn't been around since P55 days.Your problem sound more like mine, which I tracked down to faulty cabling either in the power to the HDD rack or the SATA cables.
FearTec - Tuesday, February 1, 2011 - link
I have an Engineering Sample GA-PH67A-UD3 with a ID1C4A revision 4 SB, would it be busted?http://valid.canardpc.com/show_oc.php?id=1631459
Taft12 - Tuesday, February 1, 2011 - link
Yes.Shadowmaster625 - Tuesday, February 1, 2011 - link
Are we going to see SB motherboards on sale at newegg for $39? I can live with 2 fully functional ports for $39. All I need is one for an SSD and one for a DVD. Can use NAS for the rest.slickr - Tuesday, February 1, 2011 - link
I've said Intel shot itself in the foot with the sandy bridge release with the motherboard and I was right.This is all caused by greed to make a new chipset only for customers to pay more money. They could have probably used 1156 pins for sandy if they really wanted to, but they decided to go for more profits and now it comes back to haunt them.
It would have been good if they also supported usb 3.0 and have a z68 boatherboard from the start.
Now they loose 1 billion and I hope it serves them as a lesson!
Come on AMD bring us your new processors sata 6gbps, usb 3.0 and SSD cahcing as well as better choices!
seapeople - Tuesday, February 1, 2011 - link
Yeah, bad Intel! Really, they should be even more consumer conscious and stop trying to take all our money away! Some people don't even have jobs! How are they going to decode blu-rays on a quad-core eight-threaded core i7 2000? I think Intel should just sell all their stuff at manufacturing costs so we can all have equal computer access. Anything else is just pure greed.Demon-Xanth - Tuesday, February 1, 2011 - link
The easier fix for people that have a problem in the field (aside from notebook users who would likely never see the problem) would be to just ship those customers a PCIe SATA card. Less down time, less cost, faster solution to a happy customer.333DNow - Tuesday, February 1, 2011 - link
slickrI absolutely agree with you. And those who had already bought SB can't stick upcoming Ivy Bridge to motherboards with Intel's Cougar Point chipset .
http://fudzilla.com/processors/item/21621-ivy-brid...
FullHiSpeed - Tuesday, February 1, 2011 - link
"These are the same tests that all Intel chipsets must go through, testing things like functionality, reliability and behavior at various conditions (high temps, load temps, high voltage, low voltage, etc...). The chipset made it through all of these tests just fine."A design will go through "4 corner tests" - hi-temp/hi-voltage, hi-temp/lo-voltage, lo-temp//lo-voltage, lo-temp/lo-volts. A number of samples are tested at high throughput for a few hours at each corner (with maybe a few more "corners" thrown in). You repeat this test with samples from 'fast' 'slow' and nominal process corners. This doesn't take too long but it verifies that the design is robust over a set of extreme conditions but for a short period of time
Then there's High Temperature Operating Life (HTOL) Testing ( google WHAT IS HTOL ), or accelerated life testing. We (where I work, NOT Intel) usually take a larger number of samples and operate them at high temperature for a couple of weeks in an oven, perhaps at high voltage. This takes a lot longer. We usually do it after the 4-corner tests are complete.
I guess that if you release a chip before HTOL tests are complete, you risk losing a billion dollars.
Penti - Tuesday, February 1, 2011 - link
Earlier samples didn't have the problem. Revisions don't need to be proven same as an new design so it should be quite easy sneaking in new problems there.makercab - Tuesday, February 1, 2011 - link
Here's a question. Just got the i7-2600 last week to build new unit. Still in box. Anyone know of a way to send it for replacement BEFORE installation?mapesdhs - Tuesday, February 1, 2011 - link
Live in the UK and have an absolute 7-day right of return. ;D
Ian.
valenti - Tuesday, February 1, 2011 - link
I would suggest you just go ahead and build your system. Use the SATA6G ports first. If you need more than two, add a $20 expansion card for more SATA ports. Enjoy the system and when the replacement info comes out, worry about it then.I just picked up Sandy Bridge components last Friday. Assembled the system on Sunday. Intel can pry it from my cold, dead hands. It folds proteins like a demon.
Why sit on your very fast parts for a month (realistically it will take that long for a new motherboard)? It is really a minor issue, now that we know what the problem is.
cbass64 - Tuesday, February 1, 2011 - link
Why would you replace your 2600?Touche - Tuesday, February 1, 2011 - link
If it's a transistor that fails and causes a problem, but it isn't important and not used anyway so the fix is to disable it, what makes it a problem if it "disables itself" by failing?dcollins - Tuesday, February 1, 2011 - link
About a week ago, I ordered a Vostro 460 desktop with a new Sandy Bridge processor. When the news came out about the recall yesterday, I called Dell to see whether my order would be delayed. The tech told me that my expected shipping date had been postponed a week, but did not have any details about the recall. Then, my order unexpectedly shipped this morning.I guess Dell was not effected by the recall; maybe they were only using the SATA 6GB ports or maybe they were using the A stepping or maybe they just didn't care and shipped the broken chipset. Either way, I'm not going to take any chances: I am going to ensure that the two hard drives (one SSD) are installed on the 6GB sata ports. To make room, I will need to move the CD drive to one of the 3GB ports because it is used less so it should be less affected by the bug.
Squuiid - Tuesday, February 1, 2011 - link
Having worked with Dell for years and years, I'm going with the latter, "they just didn't care and shipped the broken chipset."Dell suck:
http://www.nytimes.com/2010/06/29/technology/29del...
JasperJanssen - Tuesday, February 1, 2011 - link
Dell may suck, but so does every other vendor.banzaigtv - Wednesday, February 2, 2011 - link
HP doesn't really suck. Their Pavilion PCs use quality parts and are built to last, not to mention energy-efficient (probably why they use 300W PSUs). If you order an SB unit from HP.com, the estimated build date will just be delayed until March 1 to allow HP time to receive the updated hardware.banzaigtv - Wednesday, February 2, 2011 - link
That's weird. Dell has removed the XPS 8300 from its website and is not available for retail at electronics stores either. For those of you who have not ordered already, you can purchase the AMD-based XPS 7100 or Nehalem Core-based XPS 9100 though.Bourbo - Tuesday, February 1, 2011 - link
I am a little confused about how an upgrade offer to Z68 would be a show of goodwill to affected users. Having just built my new P67 rig with an 1155 socket processor, I don't see how that would help me. Is Intel going to offer me a 2011 socket processor and a triple channel memory kit to go along with this new Z68 board that they may (or may not) offer me? Anand, why do you suggest that will be a nice offer to current P67 users.ViRGE - Tuesday, February 1, 2011 - link
Z68 is the IGP-enabled version of P67. It's not a Socket 2011 chipset.Teknobug - Tuesday, February 1, 2011 - link
Another reason to skip first gen, I learned my lesson with the first batch of Pentium 3's.I still want a Sandy Bridge, perhaps they'll release another version chipset for the current i5/i7 Sandy Bridges, right now the AMD X6 are a waste of money since even the latest X4's still has better overall performance, I'm happy with my Phenom II X4 955BE but the i5-2500K costs less than the X6 1090T and practically STOMPS all over the X6 1100T.
WhoBeDaPlaya - Tuesday, February 1, 2011 - link
Can't beat the cost of the X3 740 BE combos I got from newEgg during Black Friday.IIRC, ~$130 for the CPU + mobo. All 6 of them unlocked to 4 cores and OCed to 3.7GHz.
My main is a X58 + i7 920 @ 4.1GHz, but damn! if those AMDs aren't cost effective o.O
dualsmp - Tuesday, February 1, 2011 - link
Just think if all motherboards had a 15 cent socket installed for the chipset . This new found flaw wouldn't be that big of a deal, and would cost a fraction of the one billion Intel is going to shell out (probably would only cost a few million).Intel could send you a small box with a new chip and small plastic tool to pop the old chip out and then insert the new one. Problem fixed. No RMA, very little waste, and down time would be extremely short for most for most people.
mkaras - Tuesday, February 1, 2011 - link
A socket for a high density large pin count BGA costs a whole lot more than 15 cents. You could very well expect to see your MOBO cost 12-16$ more at the retail level if it accomodated a chip set socket.JasperJanssen - Tuesday, February 1, 2011 - link
Don't forget the extra real estate for the chip (board area), the heatsink problems, the installation cost, and the fact that you won't be able to use large GPUs any more.$30 retail would be optimistic, says my gut.
tygrus - Tuesday, February 1, 2011 - link
I think it will be easier and cheaper to replace the 4 chipset SATA300 ports with a PCIe card with 4 SATA ports for existing systems. MB makers can now re-design to add 3rd party PCIe to SATA chip to MB and remove the Intel 4 ports.Can we reduce the voltage to the chipset to reduce the risk of failure ? No overclocking allowed ?
toyotabedzrock - Tuesday, February 1, 2011 - link
How is disabling the PLL transistor going to help, doesn't it need it to run correctly? If it doesn't then why would a failure cause any problem?
I hope you pointed out how allowing more end user voltage adjustments and not using internal PLL chips might have prevented this.
wolfman3k5 - Tuesday, February 1, 2011 - link
They have to bullshit us with some kind of explanation. Intel and Anandtech is trying to put the cherry on a turd and sugarcoat this, but the truth is that the real situation is far worse! How do I know? Well, I know the same way you know: they had to publicly disclose this to avoid big expensive - public & humiliating class action law suits... it would have been a PR nightmare. Kind of the same way NVidia shipped those bad notebook GPUs and then got sued by a a bunch of people. Lucky me, I had an ASUS notebook with one of those turds. Anyway, Intel sucks for being this stupid... I wish there was a way to went from this, but I'm fuming with frustration over this.JohnZoidberg - Wednesday, February 2, 2011 - link
You don't know nothing.And if you don't like AT don't read it.
choyak - Thursday, December 29, 2011 - link
'You don't know nothing' = you know it all!!!!!!!Blessedman - Wednesday, February 2, 2011 - link
This is only the outbound loss. I want to know the internal cost as well, how much to spin out 8million new chipsets and then just hand them out for free?kikoo182 - Wednesday, February 2, 2011 - link
Ok, just to be clear, if I do buy a p67 or h67 and ONLY USE THE BLUE SATA-6 ports, will I ever get to see the problem? I don't intend to run more then 2 hard drives anyway, for now I'm just sticking to a sata II 1.5tb 64mb cash western digital on my rig. Am I fine?misium - Wednesday, February 2, 2011 - link
Will you be fine? Well, that depends on what you eat, if you work out and look around as you cross the street... but actually, no you wont be fine. You are going to die.CrapONez - Thursday, February 3, 2011 - link
"It sounds to me like an engineer did something without thinking and this was the result.""Altogether we’re talking about a billion dollar penalty."
So do you think they'll payroll deduct?
Gnomey - Thursday, February 3, 2011 - link
Woohoo!! Bought my new Sandy Bridge Build on the 29th just before Intel announced the recall. Supplier send me a refund on the motherboard but I still have all the other parts sent to me which are now completely useless for two months as I cant get hold of a p67 Board anywhere.So I am now sat with £750 worth of components (which will depreciate in value) without the most important part unavailable and April seems to be the earliest that new boards will be released. So a nice two month wait till I can even start building my comp.
To make it worse I agreed to sell my old computer for a lot more than it was worth which I was using to help offset the cost of the build and now that wont happen.
Thanks Intel. That will sure teach me.
End of QQ.
Teknobug - Thursday, February 3, 2011 - link
Gnomey, I'm sure you can get the Z6* boards soon, or maybe your supplier can do a swap deal with you. Hang onto your parts as it shouldn't be too long (1 month maybe). I'm waiting myself to get a Sandy Bridge setup.SPOOOK - Thursday, February 3, 2011 - link
i have this motherboard the computer frezzes up every 5 mins i try to transfer files i keep getting errors cant read from hdd this bug happens all the time not in 3 yearsanan - Saturday, February 5, 2011 - link
There is no way that intel would lose on this. The replacement motherboard will be higher in cost to offset the recall.Conspiricy theory is that Sandy Bridge is running close or better than their Extreeme Edition Cpu's at a fraction of the cost.
TivoLi - Saturday, February 5, 2011 - link
Changing the MB will have an impact on the OS OEM versions (or so I believe), who will pay for that?Will MS be so nice just to give another try?
Assgier - Sunday, February 6, 2011 - link
You're implying everyone uses Windows.I have news for you: not everyone uses Windows.
METALMORPHASIS - Saturday, February 5, 2011 - link
I think between Intel and the board manufacturers, they should just send us all new boards without even having to rip out our old ones to send back. They did this with my Logitech speaker system that had the snap,crackle & pop in the audio and paid well over $100 dollars for. (Now have 2 speaker systems) A proof of purchasing slip should be all that is required and send us all new and possibly updated versions. And in the end there would still be frosting on the cake for Intel, as Im sure alot of people would buy new chips to go with their new boards!Sactodd - Friday, February 11, 2011 - link
Instead of repairing these motherboards, they should be re-purposed for schools or non profit organizations (maybe even the THE INTEL COMPUTER CLUBHOUSE NETWORK). Let the 2 6gb sata ports be used with some notice (sticker?) Its not worth the $$$ for Intel to deal with. Thats my $.02Groovester - Thursday, February 24, 2011 - link
Why not fix the GPU problem while they're at it?From: http://www.anandtech.com/show/4083/the-sandy-bridg...
"despite having a 23Hz setting in the driver, Intel’s GPU would never output anything other than 24Hz to a display"
swinster - Monday, May 25, 2015 - link
OK, 4-5 years on from this, am I now starting to see an issue? I have a Tyan S5510 server board that uses the Couger C204 chipset. A while back I ended up with some data corruption on drives that were connected to the SATA connector, but it was all very sporadic. I thought the drives might be failing, and although I replaced them (as a mater of course), I have since tested them with various diagnostic utilities on other systems and couldn't find anything actually wrong with them.The drives I used as replacements were OK for a while, but again this week, one started to get odd errors, and it was indeed connected to one of the 3Gbps ports.
ManInBrown - Sunday, May 1, 2016 - link
Hi guys, very in-depth article. I own a Toshba Qosmio X770 laptop (UK model). It comes with HM65 chipset. CPU-z calls it: Intel Sandy Bridge Rev 09 - - Intel HM65 Rev B2And Device Manager displays the dreaded: Intel(R) 6 Series/C200 Series Chipset family PCI Express Root Port 1 (1C10), Port 2 (1C12), Port 4 (1C16) and Port 6 (1C1A).
This is my 4th year running this. I am glad this intel-defect does not affect my USB3 port (which I have only one). There are no eSata ports on my laptop. I have an SSD (primary), a 7200rpm HDD (secondary) and a DVD-RW drive.
From what I understand, Toshiba engineers should only use Sata3 ports 0 and 1 because the Sata2 ports (2-5) will cause problems.
(1) If my primary hard-drive (Samsung Evo830 SSD) is connected to port 0, then why am I only getting Sata2 speeds (250-300 MB/s read and write)? All benchmark utilities I have used report similar speeds (Windows power settings set to Highest). They also report that SSD is on Sata3 (iaStor - Ok) and the secondary hard-drive is on a Sata-2 port. I conducted the SSD test on another laptop and the speeds reported were 520MB/s R/W seq. So the SSD is not faulty.
(2) If SATA3 ports 0 were soldered to primary hard-drives and SATA3 port 1 to the DVD drive, does this mean that my secondary hard-drive is on a faulty SATA2 port?
(3) I wish there was a software which could gather this information and tell me who is connected to what on the HM65 sata ports. Is there one? Maybe a hacking utility, which could enable Sata-3 speeds for my SSD (if in deed it is connected to SATA3 PORT 0).
(4) Can I swap HM65 with a non-faulty B3 chipset or HM70?
Thanks for any answers. Best regards.