For everyone, the TechSpot review neglects to account for AMD's A10-6700 (65w TDP) APU while Hardware.fr did and found it energy efficient while outperforming in graphics and games similar priced Intel offerings.
The reviewer at TechSpot also neglects to utilize any OpenCL and BasemarkCL multiprocessing benchmarks even though apps like Adobe, Aviary and numbers of other apps utilize this acceleration.
So, the TechSpot is pretty much a "hit and run" type of review and only waste time on it if you have time to waste.
The A10-6700 is just an underclocked part, not sure how "interesting" is the metric of choice here. Also, they have tested Photoshop CS6, care to read the review before commenting?
I and others have suspected that on mobile Richland the extra clock speeds and such comes from less conservative voltage binnings, but on these desktop ones i am not sure.
I can see two main scenarios: 1. AMD binned desktop chips better (overclocking should be better than trinity) 2. AMD just kept bins the same for desktop and just raised the clocks (because they had the voltage to do so)
A Thai review managed to OC the top Richland APU to 5.0Ghz on air so it looks like your first scenario is right. If that's true, then Richland should not be underestimated as it'll cause serious concerns for the venerable HTPC i3-3225 (until haswell i3 hit stores) and at the same time gain some OC credit. Incidentally, Richland is pretty much same price as Trinity in the UK.
It is completely possible the extra clocks are courtesy of a more mature process at GF. Perhaps even a silicon revision or two. It is not uncommon for engineers to make minor tweaks to a design's layout. Apple specifically purchased a company who made a living out of tweaking chip designs to fit certain power consumption and performance profiles. They of course leveraged that expertise, added on to it and started designing their own SoCs. I would assume if AMD was tweaking voltages to achieve higher clocks without process enhancements, far better yields and good binning practices; there would be a noticeable difference in power consumption and TDP. These new chips are a case of too little too late as usual. This time Intel has the graphics prowess to compete as well. Let's see if Intel can get their drivers straight...
that probably helps, but the voltages on Llano and Trinity have both been very far over.... i managed to get my A10-5800K from 3.8 base to 4.4 base without raising the voltage.
i managed to lower the voltage of an older Llano A6 quad core by .15 and raise the base clock from 1.4 to 1.8.
I hope it is both :)
And Intel is still far behind Trinity/Richland if you want to push for 30-40fps at highest possible resolution/settings.... i ran a lot of numbers (44 games)
It is ahead of AMD only with GT3 (probably at low settings, even/slightly ahead at highest playable) and GT33 parts cost more than my whole A10-5800K build...
Here is a youtube link showing 3dmark11 and windows index rating for the 4770k 3.5ghz Haswell. Not overclocked. This is apparently around 10-20fps slower than the 6800k in most games. And almost twice the price!! Youtube link: http://www.youtube.com/watch?v=k7Yo2A__1Xw
I also expect that these prices will decrease slightly as more vendors have the product in the USA area at least (only newegg according to www.pcpartpicker.com) AND/OR once trinity get EOLed.
I've been an AMD loyalist since I built my first PC in 2000 around the Athlon 700MHz chip which was better than the Pentium. Six or seven years ago the Athlon II 64 was the best chip in the world and I felt like the Jedi defeated the evil Empire, but then Intel roared back and AMD laid a stinker with the original Phenom and its infamous TLB bug. And it's been ugly since. I felt a little hopeful when Anand recently wrote "The King Returns." I'm glad that AMD appreciates the importance of its human capital, but for me it's too little and too late because yesterday I bit the bait and bought the new Intel Core i7 4770K when Microcenter put it on sale for $280. This is bitter sweet for me; I know I'm going to rock 'n roll, but I which it could've been with AMD.
AMD Richland is a refreah ala the Nvidia/AMD rebranding. The real thing to wait for is Kaveri and seeing how GCN plays with Steamroller. I myself will begetting an lappy near the holidays and hopefully will get to see GCN vs HD5200. AMD's been doing catch up no doubt about it.
What about that new motherboard you need ($150-$250)? What about that new and expensive cooler to cool down that hot Haswell that idles at over 100 degrees fahrenheit? ($50-$100+) and what about that new power supply with that Haswell ($100-$150)?
So, it's simply more than a single item purchase that Intel has forced on those choosing to upgrade.
I tried to buy an A4-4000 a couple of weeks ago, and was bitten by AMD being incompetent at sales chain management in comparison with Intel; since they didn't change the socket for Richland, there are motherboards on the shelves into which Richland fits but doesn't work, and where the advice to buyers is 'just fit a Trinity CPU and upgrade the BIOS'.
I left the Trinity CPU in, and have a surplus A4-4000 which the dealer won't take back; anyone want it?
So Newegg got these yesterday, the day before the NDA is released, but you managed to get your hands on one a couple of weeks ago? Did you think that maybe you should wait for a BIOS update from your board manufacturer before jumping in head first. And your crying over $46? Go away Intel troll.
I'm in England, and I assure you that on 20 May I ordered an "AMD A4 4000 sFM2 65W 1MB" from scan.co.uk which arrived on 22 May. Maybe they broke an embargo; so be it.
I expect that, if I buy a motherboard saying 'FM2 socket' on the box, and a processor saying 'FM2 socket' on the box, they should work together. Either AMD should contrive that their processor doesn't require a BIOS upgrade, or they should say that it requires an FM2+ motherboard.
Yes, there's a BIOS upgrade available from the board manufacturer, but an end-user can't install it without BUYING A NEW PROCESSOR; I can't see how anyone could consider that acceptable.
I am going to complain about wasting £32.88 in the hope that it stops other people from wasting £32.88.
LOL, not everybody here is in the States and as such has access to Newegg. Even so, is it that unreasonable to expect chipset and driver support before launch? If I am correct, the latter certainly happened.
Personally, I'd have a word with Scan and ask them to change the listing at the very least. I've had the same with Dabs before - a Thermaltake PSU I bought about a decade back didn't even come with a UK cable.
Haswell thermal-throttles perfectly reasonably; with the Intel heatsink that came in the box it quietly runs at average 2.9GHz when doing eight threads of number-crunching 24/7, and I start to wonder if it's worth 30W (ie $150 over three years) for 500MHz more.
You don't need a new PSU unless you want to enable the super-idle functionality; that saves you five watts at idle, which doesn't add up to being worth a new PSU.
Considering that you don't know which CPU he came from, you have to assume a new MB with a Richland purchase as well. Richland fits in FM2 but if you come from a Llano or a Phenom you have to upgrade the MB. And if you are buying a high end CPU, chances are you already have a high end cooler or don't care about noise much. And saying you need a new PSU for Haswell is pretty stupid as well. Many current PSUs support the low power states and even if not I haven't seen anyone say using a PSU without that support is dangerous.
As others have said.. this comment totally missed reality. Current PSUs work jsut fine without super-idle (not any worse than Ivy). And power consumption is lower than Sandy, so you don't need a beefier cooler (unless you go for extreme OC - but then you need a beefy cooler with any chip).
Correct on all counts. That is why I am hoping the Ivy Bridge pricing will drop now that Haswell has launched. That said I am disappointed in both Ivy Bridge and Haswell from a thermal perspective. We have a mixture of Sandy and Ivy Bridge chips in out organization. Being the geek I am I often benchmark the shit out of everything I lay my hands on. I have noticed though that the Ivy CPU's idle is much higher than Sandy Bridge and the difference in temperature between cores is more pronounced and in my opinion hampers overclocking by lowering the thermal ceiling for the entire CPU.
Quite a few people who bought boards with the new BIOS that patched up the TLB issue. It dropped performance all-round by about 10% or more, depending on the workload in question. If your system never ran into the kinds of situations that brought the bug on, you could undo the patch in AMD Overdrive, because AMD persuaded motherboard vendors to ship new boards with the new BIOS with the TLB patch set to "On" by default.
AMD desktop and server part are simply no longer competitive in the areas of raw performance and power consumption. Most likely AMD will continue to focus on the low end desktop market but put most of its efforts into mobile.
They're competitive in terms of cheapness if you don't need much compute; I bought mine because I wanted a reasonably energy-efficient fileserver, but insist on having Linux software RAID and a command-line to manage it so am not happy with commercial NAS offerings. Cheaper than anything Intel had when you add a mini-ITX motherboard and want USB3.
The thing with Bulldozer and PIledriver is that their decoders can only work on one thread per clock, but as power consumption is managed on a modular level, the module is being underworked but still at full load. The very fact that Steamroller looks to be able to work on both threads at the same time would suggest that it'd do a lot more work for little penalty. This would obviously help performance, though I still doubt that a 2M/4C Steamroller would quite overhaul even a SB 4C/4T design. Regardless, it'd most likely crush previous FX designs. Larger L1 caches plus queues means fewer misses and fewer trips back to L3, let alone main memory, which should reduce power anyway.
With an 8-core CPU that can more efficiently use its resources, 125W wouldn't actually appear all too bad. Of course, we have to wait to see this, but AMD could do with throwing out some more information on the subject so we're not reduced to speculation until the last moment.
not everybody wants to spend 1500$ and more on a cpu... for the highest compute. milj cpu are all sold in mid and low range.
we buy every year a few hundred Servers for virtualization and are all based on AMD and recently moved to the 16core and more then happy with it. Its old school IT market and managers that is killing this, just like your post, remember no competition is only negative for consumers, many still don't get that. Even in reviews today you see tech sites comparing the A10 against the 4770 and then mention that the CPU is slower.... duh
At what cost though? performance-cost ratio is waaay off the mark IMO. The frugal i3 with HD4000 served many HTPC owners quite happily; I certainly would not recommend Haswell for a HTPC build, it's prohibitively expensive (atm). On the other hand, A10 6800K sell for around £10 more than an i3-3225 (and once you throw in a mITX, they are actually the same price). This is very much in line with the article statement: " Intel competes against the A10-6800K with their Core i3 CPUs, which on the desktop remain Ivy Bridge for now." Thus, if A10 6800K/A10 6700 takes the performance lead in that price bracket (and by that I mean both CPU + iGPU); I'd say AMD would have taken a big slice of the HTPC pie (however small it is).
AMD didn't lose the iGPU performance crown. According to this article and shown in Anands's Iris pro article, Trinity beats Intel's: HD4000,4200,4400 and 4600, Richland should do even better. Besides the, i7 4770R, which Haswell i7, i5 or i3 will have an iGPU faster than Trinty or Richland?
loosing what? have a look at some review sites that actually take the time to review the new release i.s.o posting a list of asus motherboards..... you'll notice in current released cpu that they are still way better.
the iris pro hasn't show its scaling into desktop yet.
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
41 Comments
Back to Article
AFQ - Wednesday, June 5, 2013 - link
When we'll get to see a review?JarredWalton - Wednesday, June 5, 2013 - link
I'm not sure if Anand has all the numbers or not -- if not, he's at Computex so most likely we'd have to wait until he gets back.medi02 - Thursday, June 6, 2013 - link
Judging from your "trending topics":CPUs Haswell Intel GPUs Iris Mobile NVIDIA Kepler GeForce
Where Intel has 2 spots, Nvidia has 3 spots and AMD has 0 spots, we sure have to wait...
JulioFranco - Wednesday, June 5, 2013 - link
We got ours posted on TechSpot if you want to check out some benchmarks in the meantime: http://www.techspot.com/review/681-amd-a10-6800k-a...Other reviews and discussion here:
http://www.overclock.net/t/1397733/various-amd-ric...
axien86 - Wednesday, June 5, 2013 - link
For everyone, the TechSpot review neglects to account for AMD's A10-6700 (65w TDP) APU while Hardware.fr did and found it energy efficient while outperforming in graphics and games similar priced Intel offerings.The reviewer at TechSpot also neglects to utilize any OpenCL and BasemarkCL multiprocessing benchmarks even though apps like Adobe, Aviary and numbers of other apps utilize this acceleration.
So, the TechSpot is pretty much a "hit and run" type of review and only waste time on it if you have time to waste.
BMNify - Wednesday, June 5, 2013 - link
The A10-6700 is just an underclocked part, not sure how "interesting" is the metric of choice here. Also, they have tested Photoshop CS6, care to read the review before commenting?MySchizoBuddy - Wednesday, June 5, 2013 - link
he is taking about opencl in photoshop cs6.testbug00 - Wednesday, June 5, 2013 - link
I and others have suspected that on mobile Richland the extra clock speeds and such comes from less conservative voltage binnings, but on these desktop ones i am not sure.I can see two main scenarios:
1. AMD binned desktop chips better (overclocking should be better than trinity)
2. AMD just kept bins the same for desktop and just raised the clocks (because they had the voltage to do so)
K_Space - Wednesday, June 5, 2013 - link
A Thai review managed to OC the top Richland APU to 5.0Ghz on air so it looks like your first scenario is right. If that's true, then Richland should not be underestimated as it'll cause serious concerns for the venerable HTPC i3-3225 (until haswell i3 hit stores) and at the same time gain some OC credit. Incidentally, Richland is pretty much same price as Trinity in the UK.testbug00 - Wednesday, June 5, 2013 - link
That is good, but i managed to get my A10-5800k to 4.4 without raising the voltages... what site did that get to 5Ghz on air on... thanks!Einy0 - Wednesday, June 5, 2013 - link
It is completely possible the extra clocks are courtesy of a more mature process at GF. Perhaps even a silicon revision or two. It is not uncommon for engineers to make minor tweaks to a design's layout. Apple specifically purchased a company who made a living out of tweaking chip designs to fit certain power consumption and performance profiles. They of course leveraged that expertise, added on to it and started designing their own SoCs. I would assume if AMD was tweaking voltages to achieve higher clocks without process enhancements, far better yields and good binning practices; there would be a noticeable difference in power consumption and TDP. These new chips are a case of too little too late as usual. This time Intel has the graphics prowess to compete as well. Let's see if Intel can get their drivers straight...testbug00 - Wednesday, June 5, 2013 - link
that probably helps, but the voltages on Llano and Trinity have both been very far over.... i managed to get my A10-5800K from 3.8 base to 4.4 base without raising the voltage.i managed to lower the voltage of an older Llano A6 quad core by .15 and raise the base clock from 1.4 to 1.8.
I hope it is both :)
And Intel is still far behind Trinity/Richland if you want to push for 30-40fps at highest possible resolution/settings.... i ran a lot of numbers (44 games)
It is ahead of AMD only with GT3 (probably at low settings, even/slightly ahead at highest playable) and GT33 parts cost more than my whole A10-5800K build...
let us see :)
majorleague - Wednesday, June 5, 2013 - link
Here is a youtube link showing 3dmark11 and windows index rating for the 4770k 3.5ghz Haswell. Not overclocked.This is apparently around 10-20fps slower than the 6800k in most games. And almost twice the price!!
Youtube link:
http://www.youtube.com/watch?v=k7Yo2A__1Xw
testbug00 - Wednesday, June 5, 2013 - link
I also expect that these prices will decrease slightly as more vendors have the product in the USA area at least (only newegg according to www.pcpartpicker.com) AND/OR once trinity get EOLed.anonymous_user - Wednesday, June 5, 2013 - link
Is it just me or does there seem to be a significant price gap between the A6-6400K and A4-4000?JarredWalton - Wednesday, June 5, 2013 - link
Trinity APUs like the A4-5300 continue to plug that hole I think is the plan.7beauties - Wednesday, June 5, 2013 - link
I've been an AMD loyalist since I built my first PC in 2000 around the Athlon 700MHz chip which was better than the Pentium. Six or seven years ago the Athlon II 64 was the best chip in the world and I felt like the Jedi defeated the evil Empire, but then Intel roared back and AMD laid a stinker with the original Phenom and its infamous TLB bug. And it's been ugly since. I felt a little hopeful when Anand recently wrote "The King Returns." I'm glad that AMD appreciates the importance of its human capital, but for me it's too little and too late because yesterday I bit the bait and bought the new Intel Core i7 4770K when Microcenter put it on sale for $280. This is bitter sweet for me; I know I'm going to rock 'n roll, but I which it could've been with AMD.stickmansam - Wednesday, June 5, 2013 - link
AMD Richland is a refreah ala the Nvidia/AMD rebranding. The real thing to wait for is Kaveri and seeing how GCN plays with Steamroller. I myself will begetting an lappy near the holidays and hopefully will get to see GCN vs HD5200. AMD's been doing catch up no doubt about it.testbug00 - Wednesday, June 5, 2013 - link
Richland actually has some significant power saving achievements in mobile... on desktop agree is respin with better OC/baseaxien86 - Wednesday, June 5, 2013 - link
Congrats, but simply buying a Haswell for $280?
What about that new motherboard you need ($150-$250)? What about that new and expensive cooler to cool down that hot Haswell that idles at over 100 degrees fahrenheit? ($50-$100+) and what about that new power supply with that Haswell ($100-$150)?
So, it's simply more than a single item purchase that Intel has forced on those choosing to upgrade.
TomWomack - Wednesday, June 5, 2013 - link
I tried to buy an A4-4000 a couple of weeks ago, and was bitten by AMD being incompetent at sales chain management in comparison with Intel; since they didn't change the socket for Richland, there are motherboards on the shelves into which Richland fits but doesn't work, and where the advice to buyers is 'just fit a Trinity CPU and upgrade the BIOS'.I left the Trinity CPU in, and have a surplus A4-4000 which the dealer won't take back; anyone want it?
superflex - Wednesday, June 5, 2013 - link
So Newegg got these yesterday, the day before the NDA is released, but you managed to get your hands on one a couple of weeks ago? Did you think that maybe you should wait for a BIOS update from your board manufacturer before jumping in head first. And your crying over $46?Go away Intel troll.
TomWomack - Wednesday, June 5, 2013 - link
I'm in England, and I assure you that on 20 May I ordered an "AMD A4 4000 sFM2 65W 1MB" from scan.co.uk which arrived on 22 May. Maybe they broke an embargo; so be it.I expect that, if I buy a motherboard saying 'FM2 socket' on the box, and a processor saying 'FM2 socket' on the box, they should work together. Either AMD should contrive that their processor doesn't require a BIOS upgrade, or they should say that it requires an FM2+ motherboard.
Yes, there's a BIOS upgrade available from the board manufacturer, but an end-user can't install it without BUYING A NEW PROCESSOR; I can't see how anyone could consider that acceptable.
I am going to complain about wasting £32.88 in the hope that it stops other people from wasting £32.88.
silverblue - Wednesday, June 5, 2013 - link
LOL, not everybody here is in the States and as such has access to Newegg. Even so, is it that unreasonable to expect chipset and driver support before launch? If I am correct, the latter certainly happened.Personally, I'd have a word with Scan and ask them to change the listing at the very least. I've had the same with Dabs before - a Thermaltake PSU I bought about a decade back didn't even come with a UK cable.
TomWomack - Wednesday, June 5, 2013 - link
Haswell thermal-throttles perfectly reasonably; with the Intel heatsink that came in the box it quietly runs at average 2.9GHz when doing eight threads of number-crunching 24/7, and I start to wonder if it's worth 30W (ie $150 over three years) for 500MHz more.You don't need a new PSU unless you want to enable the super-idle functionality; that saves you five watts at idle, which doesn't add up to being worth a new PSU.
I'll admit you need a new motherboard.
Death666Angel - Wednesday, June 5, 2013 - link
Considering that you don't know which CPU he came from, you have to assume a new MB with a Richland purchase as well. Richland fits in FM2 but if you come from a Llano or a Phenom you have to upgrade the MB.And if you are buying a high end CPU, chances are you already have a high end cooler or don't care about noise much. And saying you need a new PSU for Haswell is pretty stupid as well. Many current PSUs support the low power states and even if not I haven't seen anyone say using a PSU without that support is dangerous.
MrSpadge - Wednesday, June 5, 2013 - link
As others have said.. this comment totally missed reality. Current PSUs work jsut fine without super-idle (not any worse than Ivy). And power consumption is lower than Sandy, so you don't need a beefier cooler (unless you go for extreme OC - but then you need a beefy cooler with any chip).Veroxious - Wednesday, June 5, 2013 - link
Correct on all counts. That is why I am hoping the Ivy Bridge pricing will drop now that Haswell has launched. That said I am disappointed in both Ivy Bridge and Haswell from a thermal perspective. We have a mixture of Sandy and Ivy Bridge chips in out organization. Being the geek I am I often benchmark the shit out of everything I lay my hands on. I have noticed though that the Ivy CPU's idle is much higher than Sandy Bridge and the difference in temperature between cores is more pronounced and in my opinion hampers overclocking by lowering the thermal ceiling for the entire CPU.testbug00 - Wednesday, June 5, 2013 - link
Haswell is not bad, it is just not a leap in any one area unless your getting GT3(e)and over 100 fahrenheit? um... why does it matter in idle?? power consumption is down and idle heat isn't that big of a deal (at that temp).
silverblue - Wednesday, June 5, 2013 - link
Who did the TLB bug actually affect, in all honesty?Cataclysm_ZA - Wednesday, June 5, 2013 - link
Quite a few people who bought boards with the new BIOS that patched up the TLB issue. It dropped performance all-round by about 10% or more, depending on the workload in question. If your system never ran into the kinds of situations that brought the bug on, you could undo the patch in AMD Overdrive, because AMD persuaded motherboard vendors to ship new boards with the new BIOS with the TLB patch set to "On" by default.tech6 - Wednesday, June 5, 2013 - link
AMD desktop and server part are simply no longer competitive in the areas of raw performance and power consumption. Most likely AMD will continue to focus on the low end desktop market but put most of its efforts into mobile.TomWomack - Wednesday, June 5, 2013 - link
They're competitive in terms of cheapness if you don't need much compute; I bought mine because I wanted a reasonably energy-efficient fileserver, but insist on having Linux software RAID and a command-line to manage it so am not happy with commercial NAS offerings. Cheaper than anything Intel had when you add a mini-ITX motherboard and want USB3.blckgrffn - Wednesday, June 5, 2013 - link
idle power usage is pretty close as well - and that is where most CPU's spend their time anyway.silverblue - Wednesday, June 5, 2013 - link
The thing with Bulldozer and PIledriver is that their decoders can only work on one thread per clock, but as power consumption is managed on a modular level, the module is being underworked but still at full load. The very fact that Steamroller looks to be able to work on both threads at the same time would suggest that it'd do a lot more work for little penalty. This would obviously help performance, though I still doubt that a 2M/4C Steamroller would quite overhaul even a SB 4C/4T design. Regardless, it'd most likely crush previous FX designs. Larger L1 caches plus queues means fewer misses and fewer trips back to L3, let alone main memory, which should reduce power anyway.With an 8-core CPU that can more efficiently use its resources, 125W wouldn't actually appear all too bad. Of course, we have to wait to see this, but AMD could do with throwing out some more information on the subject so we're not reduced to speculation until the last moment.
duploxxx - Wednesday, June 5, 2013 - link
not everybody wants to spend 1500$ and more on a cpu... for the highest compute. milj cpu are all sold in mid and low range.we buy every year a few hundred Servers for virtualization and are all based on AMD and recently moved to the 16core and more then happy with it. Its old school IT market and managers that is killing this, just like your post, remember no competition is only negative for consumers, many still don't get that. Even in reviews today you see tech sites comparing the A10 against the 4770 and then mention that the CPU is slower.... duh
price/performance/power they are competative.
Mountainjoy - Wednesday, June 5, 2013 - link
Wow that did not take long for AMD to lose the iGPU performance crown.silverblue - Wednesday, June 5, 2013 - link
There's also the fact that its competition is priced three times higher.K_Space - Wednesday, June 5, 2013 - link
At what cost though? performance-cost ratio is waaay off the mark IMO. The frugal i3 with HD4000 served many HTPC owners quite happily; I certainly would not recommend Haswell for a HTPC build, it's prohibitively expensive (atm). On the other hand, A10 6800K sell for around £10 more than an i3-3225 (and once you throw in a mITX, they are actually the same price). This is very much in line with the article statement: " Intel competes against the A10-6800K with their Core i3 CPUs, which on the desktop remain Ivy Bridge for now."Thus, if A10 6800K/A10 6700 takes the performance lead in that price bracket (and by that I mean both CPU + iGPU); I'd say AMD would have taken a big slice of the HTPC pie (however small it is).
superunknown98 - Wednesday, June 5, 2013 - link
AMD didn't lose the iGPU performance crown. According to this article and shown in Anands's Iris pro article, Trinity beats Intel's: HD4000,4200,4400 and 4600, Richland should do even better. Besides the, i7 4770R, which Haswell i7, i5 or i3 will have an iGPU faster than Trinty or Richland?duploxxx - Wednesday, June 5, 2013 - link
loosing what? have a look at some review sites that actually take the time to review the new release i.s.o posting a list of asus motherboards..... you'll notice in current released cpu that they are still way better.the iris pro hasn't show its scaling into desktop yet.