So that makes 4 different things that were causing instability on Intel HW. 1st the MB settings, 2nd the microcode requesting incorrect voltages, 3rd the oxidation, and 4th SW/HW problems in laptops.
I am imagining just how hard this was for Intel's engineering team to troubleshoot. It must have been a nightmare.
Assuming the problem is, through the forthcoming microcode update, resolved, I'd like to congratulate them on a job well done.
I don't pretend to know everything, but having followed Intel and AMD CPUs since before Zen, and the only HW longevity problems I've heard of are those mentioned above.
Intel can afford to admit to microcode bugs because they have had those before. What they cannot afford to admit to is "we pushed 5 nodes in 4 years too fast"...
Look at what they say: " Our analysis of returned processors confirms that the elevated operating voltage is stemming from a microcode algorithm resulting in incorrect voltage requests to the processor. " That would be legally defensible as "true" even if the basic problem were "Intel Foundry gave us the wrong numbers to plug into our Turbo microcode"...
Obviously I have no insider knowledge of Intel. Maybe the problem is 100% incompetence in designing the "algorithm" for Turbo control of these chips. But the way I'd bet is that this was a manufacturing issue (in the sense that manufacturing was rushed out too fast, without enough time to fully characterize how the devices behave under long-term stress conditions), but admitting it as a Foundry issue rather than as a microcode issue has a much larger business impact, so...
A similar thought I've had is that the fault might be process-related, but the only thing they can feasibly do about it is tweaking microcode to better cater to the chip's true limitations.
I should add that the timing seems highly suspicious. It's as if they want to be sure the fix lands *after* all the reviews of Ryzen 9000 get published, so that reviewers will still compare against the unpatched Raptor Lakes, not hampered by any performance-robbing mitigations resulting from the fix.
I'm actually curious to see how various review websites are going to react, now that the new Ryzens were slightly delayed. I too found the timing quite suspicious.
I was the only one to call our intel's dumpster fire of cpu's in anandtech's review on the i9-14900KS, if you check the comments, this is before this voltage/microcode disaster was published. Intel should not be commended at the time the article was written. 8P+16E core cpu's should not be consuming 350-400W at stock, that insanity. 128 core AMD EPYC cpus consume 280W, and even that i think is too much. Intel is a complete garbage cpu compared to the playing field, instead of increasing IPC 5-10% each generation, they have only been increasing clocks, which people can do on their own if they want. Intel needs to take a time out and retool their fabs to current gen and fix these issues asap.
After the root-cause resolution itself, probably the next most important factor is addressing permanent damage incurred by said voltage excursions. That'll be the "contact us" part. I'm guessing Intel will doing an all-expenses-paid CPU exchange program for anyone with on-going problems after the patch is applied.
I’m curious what those elevated operating voltages actually were in terms of raw value. Were those voltages in ranges that most people would consider safe only under LN2 like 1.7 V? Or even higher?
The one piece of good news is that lower voltages should also reduce power consumption. Curious what that difference will be as well:
Yep, Intel screwed up. But it shouldn't be any performance hit as these voltages are clearly not normal for even overclocking. It's wrecking the CPUs - which will still need replaced after the patch. That said, the prior recommendations for limiting the damage have already had a performance hit. So unless that gets reversed then I guess everyone has taken the hit already.
The 300+ watts won't be removed. That is expected behaviour for the high end settings. The incorrect voltages were not part of the allowed power limits.
If you think that 300W 'extreme' default manufacturer profiles were 'normal' and won't be un-defaulted then I have an Intel ARC 'high end graphics card' to sell you...
No. Typical for Intel is: Premium Performance and Reliability, for a Premium Price, but at very high power. Though their reliability is currently recent badly tarnished, they have provided several decades of performant and reliable processors. Were it not for that, they would not be so broadly deployed for both consumers and commercial servers.
76% market share is "collapsed"? they ship 50 million units to AMDs 8 million... that's "plummeted"? Steam survey shows 66.8% Intel procs...
I'm sure you are going with the passmark numbers, which represent benchmark tests run... not actual units out there. And AMD will be badly over-represented there because they are predominantly only used by perpetually online hobbyists desperately looking for validation in the form of high benchmark scores... Intel computers are built for, and used by, people who just want to use their computers
This is all good and well, I guess, but you still haven't answered a pending issue since May:
Given that major motherboard manufacturers have now adapted their firmware to use Intel Default Settings by default, and given that's what Intel has been recommending to both manufacturers and end users, will you finally re-test the involved CPUs under those new, more conservative settings, so users can make an educated comparaison and decide with all informations at hand ?
Or should we conclude that the "ridiculous power settings" you are referring to in this article are only selectively ridiculous ?
"will you finally re-test the involved CPUs under those new, more conservative settings, so users can make an educated comparaison and decide with all informations at hand ?"
It's going to take some time to get it all done, but yes.
A fairly simple test would be great - vanilla 14900K vs current intel recommended. One may call it preliminary, but given your credibility and quality Anandtech has presented (at least for me) for years, that would be enought. Sure, I probably could (if I were an unlucky owner of 14900K) conduct such test, but that would be anecdotical and unrepresentative. While yours is backed by years of experience. A test here, on AT is not some kind of twitter crap, it a solid piece of review, even if short.
Good. Then the question is: why wasn't it done yet, and how is that taking so much time?
As a reminder, one of Anandtech's authors claimed he'd do just that "over the weekend" in the article about the i9-14900KS at the beginning of May. So I guess that if it was not going to take longer than a weekend *then* for a given CPU shouldn't take much longer *now* (I'd expect a little more time would be needed, to rebuild the test configuration, etc), right?
Hopefully at least in the meantime, I guess you'll add clear warnings on the charts in the recent July's buyer's guide article, so that readers don't get the wrong information about those CPUs and can make an educated choice based on the content of that article ?
Intel have already said "contact us" when the patch fails to solve it. That's already them saying they expect many permanently broken CPUs. Which they will replace for free obviously.
On INTC financial on channel average weighed price I calculate 125,966,924 to 150 million Raptor desktop components (maybe a smidge more on bundle deal) that are 62,598,714 13th and 38,173,935 14th generation components.
On replacement Intel's looking at around $36 at variable cost and $60 marginal cost per component plus handling and administration.
14900/13900K_ represent 1/3rd of Raptors production and KS alone = 6%. 12900K_ volume is 20.1% with KS alone 2.4%. 9900K_ mature 14+ shows a similar volume to Raptors on the maturity of SF10 at 34% however 9th KS represent a meager 0.0017%.
10th i9's represent 10% and adding i7's = 22.7%. Obviously noteworthy reference concerning the stability and evolutionary improvement of design process to hit frequency targets.
All i9/7/5 K_ by generation.
14th = 73.9% and Intel needed them to drive up full line gross margin 13th = 62.5% similar for margin and AMD did same w/R7K emphasizing 16/12C volume. 12th = 46.8% 11th = 41.9% and 18.7% are i9 10th = 27.1% and 10% are i9 9th = 49.35% and 34% are i9
Here's Raptor percent by SKU on channel supply weight.
Maybe I misunderstood something, but when you wrote: "On replacement Intel's looking at around $36 at variable cost and $60 marginal cost per component plus handling and administration."
You meant it will only cost Intel $60 plus handling to replace each CPU? That's a bit low compared to the prices of even the cheaper SKUs sold today.
@bellsysemlord, the average marginal cost of Raptors is approximately $60 and subject depreciated process no more than $67 currently.
At $60 Intel price in volume to OEMs and large retail is $247.50 in a full line procurement mirroring sort out of the fab; 46.9% i9, 23.9% i7, 8.9%, 22.7% i5, 6.2% i3.
Intel wholesale trade price at x4 over marginal cost of production equals a traditional Intel competitive profit point.
In a full line procurement since i5 1x500 at $1K is underwater at $247 they present a negotiating point between the buyer and Intel. Savvy procurement negotiates for n/c 'bundle deal' that is sales close incentive and Intel might agree as a 'reward' for contract performance. Buyers are willing to pay $247 for a 1x500 or less to get the high margin i7/i9 allocation. High volume can be interpreted as full line production volume divided among seven to eleven top customers.
Otherwise, the basic rule for determining what some mid volume customer, say an SI pays through distribution. is $1K / 2. For some local operator for a tray of 10 CPUs < $1K x 10 < 10%.
I will add to this, see the percent by SKU split, for 14th and 13th desktop up comment string, that is what you get in a full line procurement for $247 each but you're also buying millions.
For everyone else and this is an end buyer necessity to know, at volume retail, they can get the OEM deal from Intel but they may also be getting something not 'so sweetened'.
What the end buyer will benefit from is awareness of retail cost subject price margin to know when you're getting a deal which is toward run end. Or what the margin is on the CPU specifically when buying a lot of PC kit there's some price negotiation room here especially if some of that kit is open box because sometimes open box has already been paid for in a warranty claim.
For Intel rely on $1K / 2 for approximately what that reseller paid for the Core processor available for sale to the end buyer and for AMD Ryzen or TR its MSRP / 3 * 1.5.
Also know for Intel everything down bin, under a 13600K, is possibly secured, highly likely by a volume retail operator for nothing. Which means they're making a 100% margin basically at end sale. Knowing this the end buyer has negotiation potential in a PC kit buy too secure a discount. For AMD it's the hexa core and quads that are thrown into sales package for free and in a large retail purchase the APUs.
Brings memories of the Pentium FDIV bug, where in 1 in 9 billion FP resulted in the 4th digit and the rest to be wrong. Intel didn't want to issue a recall, but said they would replace the CPU's for those that proved they needed a replacement. As, the FP error didn't affect average user. They eventually had to replace all CPU's from customers requesting it, as the story blew up in the media and users felt betrayed. Intel handled the PR badly from the start, and honestly they seem to be repeating that mistake!
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
49 Comments
Back to Article
ballsystemlord - Monday, July 22, 2024 - link
So that makes 4 different things that were causing instability on Intel HW. 1st the MB settings, 2nd the microcode requesting incorrect voltages, 3rd the oxidation, and 4th SW/HW problems in laptops.I am imagining just how hard this was for Intel's engineering team to troubleshoot. It must have been a nightmare.
Assuming the problem is, through the forthcoming microcode update, resolved, I'd like to congratulate them on a job well done.
fallaha56 - Tuesday, July 23, 2024 - link
Errr…the voltage problem has likely caused degradation issues ie chips already malfunctioning will not be fixed by thisAnd of course you’ll notice Intels new RL stepping planned for release
Time overdue for Intel to come clean about this
Oxford Guy - Tuesday, July 23, 2024 - link
The assertion is that situation went from many years without problems to having 4 separate simultaneous problems?ballsystemlord - Tuesday, July 23, 2024 - link
I don't pretend to know everything, but having followed Intel and AMD CPUs since before Zen, and the only HW longevity problems I've heard of are those mentioned above.Oxford Guy - Friday, July 26, 2024 - link
I was just highlighting the situation.name99 - Tuesday, July 23, 2024 - link
Except how much of the above is the FULL truth?Intel can afford to admit to microcode bugs because they have had those before.
What they cannot afford to admit to is "we pushed 5 nodes in 4 years too fast"...
Look at what they say:
"
Our analysis of returned processors confirms that the elevated operating voltage is stemming from a microcode algorithm resulting in incorrect voltage requests to the processor.
"
That would be legally defensible as "true" even if the basic problem were "Intel Foundry gave us the wrong numbers to plug into our Turbo microcode"...
Obviously I have no insider knowledge of Intel. Maybe the problem is 100% incompetence in designing the "algorithm" for Turbo control of these chips. But the way I'd bet is that this was a manufacturing issue (in the sense that manufacturing was rushed out too fast, without enough time to fully characterize how the devices behave under long-term stress conditions), but admitting it as a Foundry issue rather than as a microcode issue has a much larger business impact, so...
mode_13h - Tuesday, July 23, 2024 - link
Well said.A similar thought I've had is that the fault might be process-related, but the only thing they can feasibly do about it is tweaking microcode to better cater to the chip's true limitations.
mode_13h - Tuesday, July 23, 2024 - link
I should add that the timing seems highly suspicious. It's as if they want to be sure the fix lands *after* all the reviews of Ryzen 9000 get published, so that reviewers will still compare against the unpatched Raptor Lakes, not hampered by any performance-robbing mitigations resulting from the fix.kkilobyte - Thursday, July 25, 2024 - link
I'm actually curious to see how various review websites are going to react, now that the new Ryzens were slightly delayed. I too found the timing quite suspicious.Oxford Guy - Tuesday, July 30, 2024 - link
Andy Edser: 'No recall, no halting of sales, and no comment on warranty extensions. … the damage to affected chips may be permanent'edlee - Thursday, August 1, 2024 - link
I was the only one to call our intel's dumpster fire of cpu's in anandtech's review on the i9-14900KS, if you check the comments, this is before this voltage/microcode disaster was published. Intel should not be commended at the time the article was written. 8P+16E core cpu's should not be consuming 350-400W at stock, that insanity. 128 core AMD EPYC cpus consume 280W, and even that i think is too much. Intel is a complete garbage cpu compared to the playing field, instead of increasing IPC 5-10% each generation, they have only been increasing clocks, which people can do on their own if they want. Intel needs to take a time out and retool their fabs to current gen and fix these issues asap.Oxford Guy - Friday, August 2, 2024 - link
The size of a core is a simplistic measure but something clearly is amiss.evanh - Monday, July 22, 2024 - link
After the root-cause resolution itself, probably the next most important factor is addressing permanent damage incurred by said voltage excursions. That'll be the "contact us" part. I'm guessing Intel will doing an all-expenses-paid CPU exchange program for anyone with on-going problems after the patch is applied.Kevin G - Monday, July 22, 2024 - link
I’m curious what those elevated operating voltages actually were in terms of raw value. Were those voltages in ranges that most people would consider safe only under LN2 like 1.7 V? Or even higher?The one piece of good news is that lower voltages should also reduce power consumption. Curious what that difference will be as well:
trparky - Tuesday, July 23, 2024 - link
What I'm interested in seeing is how much of a hit to performance this will be.No matter what, this is a black mark on Intel's reputation.
evanh - Tuesday, July 23, 2024 - link
Yep, Intel screwed up. But it shouldn't be any performance hit as these voltages are clearly not normal for even overclocking. It's wrecking the CPUs - which will still need replaced after the patch.That said, the prior recommendations for limiting the damage have already had a performance hit. So unless that gets reversed then I guess everyone has taken the hit already.
fallaha56 - Tuesday, July 23, 2024 - link
Er of course there will be perf drops‘Voltage fix’ is code for ‘we ran the voltages too high to try and compete with AMD but we damaged the chips’
300W+ was always silly for desktop PCs
evanh - Tuesday, July 23, 2024 - link
The 300+ watts won't be removed. That is expected behaviour for the high end settings. The incorrect voltages were not part of the allowed power limits.fallaha56 - Tuesday, July 23, 2024 - link
If you think that 300W 'extreme' default manufacturer profiles were 'normal' and won't be un-defaulted then I have an Intel ARC 'high end graphics card' to sell you...fallaha56 - Tuesday, July 23, 2024 - link
2/2 to be clear I'm a (very) annoyed shareholder looking forward to recompenseIntel clearly knew about degradation months ago but never admitted it, never mind the added 'turbo default' farce
eloyard - Tuesday, July 23, 2024 - link
Performance drop after all the benchmarks, large part of sales, most promotions and main marketing push is done.Typical Intel.
dwbogardus - Tuesday, July 23, 2024 - link
No. Typical for Intel is: Premium Performance and Reliability, for a Premium Price, but at very high power. Though their reliability is currently recent badly tarnished, they have provided several decades of performant and reliable processors. Were it not for that, they would not be so broadly deployed for both consumers and commercial servers.fallaha56 - Tuesday, July 23, 2024 - link
Er no, Intels market share has plummeted and their market share has collapsed -and I’m a shareholder (for now anyway)101: no one serious in the self-build community goes or recommends Intel any more
Jorgp2 - Tuesday, July 23, 2024 - link
Lol, wutfallaha56 - Tuesday, July 23, 2024 - link
Yes Jorg, or did you just build a 14900K system?That would be the 'lol' surely
temps - Wednesday, July 24, 2024 - link
76% market share is "collapsed"? they ship 50 million units to AMDs 8 million... that's "plummeted"? Steam survey shows 66.8% Intel procs...I'm sure you are going with the passmark numbers, which represent benchmark tests run... not actual units out there. And AMD will be badly over-represented there because they are predominantly only used by perpetually online hobbyists desperately looking for validation in the form of high benchmark scores... Intel computers are built for, and used by, people who just want to use their computers
kkilobyte - Tuesday, July 23, 2024 - link
This is all good and well, I guess, but you still haven't answered a pending issue since May:Given that major motherboard manufacturers have now adapted their firmware to use Intel Default Settings by default, and given that's what Intel has been recommending to both manufacturers and end users, will you finally re-test the involved CPUs under those new, more conservative settings, so users can make an educated comparaison and decide with all informations at hand ?
Or should we conclude that the "ridiculous power settings" you are referring to in this article are only selectively ridiculous ?
TheinsanegamerN - Tuesday, July 23, 2024 - link
They'll retest, as soon as they build a new GPU bench. It's only been 3 years, you have to give them TIME, MAN!Ryan Smith - Wednesday, July 24, 2024 - link
"will you finally re-test the involved CPUs under those new, more conservative settings, so users can make an educated comparaison and decide with all informations at hand ?"It's going to take some time to get it all done, but yes.
m3city - Wednesday, July 24, 2024 - link
A fairly simple test would be great - vanilla 14900K vs current intel recommended. One may call it preliminary, but given your credibility and quality Anandtech has presented (at least for me) for years, that would be enought. Sure, I probably could (if I were an unlucky owner of 14900K) conduct such test, but that would be anecdotical and unrepresentative. While yours is backed by years of experience. A test here, on AT is not some kind of twitter crap, it a solid piece of review, even if short.kkilobyte - Wednesday, July 24, 2024 - link
Good. Then the question is: why wasn't it done yet, and how is that taking so much time?As a reminder, one of Anandtech's authors claimed he'd do just that "over the weekend" in the article about the i9-14900KS at the beginning of May. So I guess that if it was not going to take longer than a weekend *then* for a given CPU shouldn't take much longer *now* (I'd expect a little more time would be needed, to rebuild the test configuration, etc), right?
Hopefully at least in the meantime, I guess you'll add clear warnings on the charts in the recent July's buyer's guide article, so that readers don't get the wrong information about those CPUs and can make an educated choice based on the content of that article ?
Oxford Guy - Friday, July 26, 2024 - link
It's not May anymore? I guess I've taken a long weekend, indeed.fallaha56 - Tuesday, July 23, 2024 - link
Errr come on Intel…the voltage problem has likely caused degradation issues ie chips already malfunctioning will not be fixed by thisAnd of course you’ll notice Intels new RL stepping planned for release
Time overdue for Intel to come clean about this (Intel shareholder here)
evanh - Tuesday, July 23, 2024 - link
Intel have already said "contact us" when the patch fails to solve it. That's already them saying they expect many permanently broken CPUs. Which they will replace for free obviously.fallaha56 - Tuesday, July 23, 2024 - link
That's fine but as one of many shareholders sitting on losses due to material dishonesty we'll be looking for compensationcatavalon21 - Sunday, July 28, 2024 - link
So go the risks of investment, dude.Bruzzone - Tuesday, July 23, 2024 - link
On INTC financial on channel average weighed price I calculate 125,966,924 to 150 million Raptor desktop components (maybe a smidge more on bundle deal) that are 62,598,714 13th and 38,173,935 14th generation components.On replacement Intel's looking at around $36 at variable cost and $60 marginal cost per component plus handling and administration.
14900/13900K_ represent 1/3rd of Raptors production and KS alone = 6%. 12900K_ volume is 20.1% with KS alone 2.4%. 9900K_ mature 14+ shows a similar volume to Raptors on the maturity of SF10 at 34% however 9th KS represent a meager 0.0017%.
10th i9's represent 10% and adding i7's = 22.7%. Obviously noteworthy reference concerning the stability and evolutionary improvement of design process to hit frequency targets.
All i9/7/5 K_ by generation.
14th = 73.9% and Intel needed them to drive up full line gross margin
13th = 62.5% similar for margin and AMD did same w/R7K emphasizing 16/12C volume.
12th = 46.8%
11th = 41.9% and 18.7% are i9
10th = 27.1% and 10% are i9
9th = 49.35% and 34% are i9
Here's Raptor percent by SKU on channel supply weight.
14900KS 1.38%
14900K 30.29%
14900KF 5.09%
14900 0.02%
14900F 1.21%
14900T 0.77%
14790F 0.35%
14700K 14.07%
14700KF 5.22%
14700 0.04%
14700F 3.09%
14700T 0.66%
14600K 14.71%
14600KF 3.24%
14600 0.01%
14600T 0.05%
14500 2.79%
14500T 0.37%
14490F 0.02%
14400 6.11%
14400T 0.08%
14400F 2.35%
14100 7.19%
14100F 0.76%
14100T 0.12%
100.00%
13900KS 7.06%
13900K 28.57%
13900 3.84%
13900KF 2.39%
13900F 1.77%
13900T 0.40%
13790F 0.18%
13700K 14.04%
13700 4.45%
13700KF 2.40%
13700F 2.83%
13700T 1.47%
13600K 5.68%
13600 0.13%
13600KF 2.40%
13600T 0.59%
13500 5.51%
13500T 2.75%
13490F 0.21%
13400 2.98%
13400F 2.15%
13400T 1.51%
13100 4.79%
13100F 1.13%
13100T 0.76%
100.00%
mb
Bruzzone - Tuesday, July 23, 2024 - link
Clarification, 14th generation run to date is 63,368,210 and 13th 62,598,714 components for range low 125,966,924 units produced. mbballsystemlord - Tuesday, July 23, 2024 - link
Maybe I misunderstood something, but when you wrote: "On replacement Intel's looking at around $36 at variable cost and $60 marginal cost per component plus handling and administration."You meant it will only cost Intel $60 plus handling to replace each CPU? That's a bit low compared to the prices of even the cheaper SKUs sold today.
Bruzzone - Thursday, July 25, 2024 - link
@bellsysemlord, the average marginal cost of Raptors is approximately $60 and subject depreciated process no more than $67 currently.At $60 Intel price in volume to OEMs and large retail is $247.50 in a full line procurement mirroring sort out of the fab; 46.9% i9, 23.9% i7, 8.9%, 22.7% i5, 6.2% i3.
Intel wholesale trade price at x4 over marginal cost of production equals a traditional Intel competitive profit point.
In a full line procurement since i5 1x500 at $1K is underwater at $247 they present a negotiating point between the buyer and Intel. Savvy procurement negotiates for n/c 'bundle deal' that is sales close incentive and Intel might agree as a 'reward' for contract performance. Buyers are willing to pay $247 for a 1x500 or less to get the high margin i7/i9 allocation. High volume can be interpreted as full line production volume divided among seven to eleven top customers.
Otherwise, the basic rule for determining what some mid volume customer, say an SI pays through distribution. is $1K / 2. For some local operator for a tray of 10 CPUs < $1K x 10 < 10%.
mb
Bruzzone - Thursday, July 25, 2024 - link
I will add to this, see the percent by SKU split, for 14th and 13th desktop up comment string, that is what you get in a full line procurement for $247 each but you're also buying millions.For everyone else and this is an end buyer necessity to know, at volume retail, they can get the OEM deal from Intel but they may also be getting something not 'so sweetened'.
What the end buyer will benefit from is awareness of retail cost subject price margin to know when you're getting a deal which is toward run end. Or what the margin is on the CPU specifically when buying a lot of PC kit there's some price negotiation room here especially if some of that kit is open box because sometimes open box has already been paid for in a warranty claim.
For Intel rely on $1K / 2 for approximately what that reseller paid for the Core processor available for sale to the end buyer and for AMD Ryzen or TR its MSRP / 3 * 1.5.
Also know for Intel everything down bin, under a 13600K, is possibly secured, highly likely by a volume retail operator for nothing. Which means they're making a 100% margin basically at end sale. Knowing this the end buyer has negotiation potential in a PC kit buy too secure a discount. For AMD it's the hexa core and quads that are thrown into sales package for free and in a large retail purchase the APUs.
mb
ballsystemlord - Friday, July 26, 2024 - link
ThanksBruzzone - Wednesday, July 31, 2024 - link
u'r welcome, anytime, if the data comes to use report back the result. thks mbGeoffreyA - Tuesday, July 23, 2024 - link
"Something is rotten in the state of Denmark."Ryan Smith - Wednesday, July 24, 2024 - link
More like the state of Oregon, if we're being technical.GeoffreyA - Wednesday, July 24, 2024 - link
Indeed, good thinking!catavalon21 - Sunday, July 28, 2024 - link
+1Bluetooth - Saturday, July 27, 2024 - link
Brings memories of the Pentium FDIV bug, where in 1 in 9 billion FP resulted in the 4th digit and the rest to be wrong. Intel didn't want to issue a recall, but said they would replace the CPU's for those that proved they needed a replacement. As, the FP error didn't affect average user. They eventually had to replace all CPU's from customers requesting it, as the story blew up in the media and users felt betrayed. Intel handled the PR badly from the start, and honestly they seem to be repeating that mistake!Oxford Guy - Tuesday, July 30, 2024 - link
Here's what Andy Edser wrote: 'No recall, no halting of sales, and no comment on warranty extensions. … the damage to affected chips may be permanent'Intel certainly isn't handling this well.