I don't like at all that they're failing to specify base clock speeds. Will this become a trend like it has with their newer graphics cards? For CPUs, that's absolutely unacceptable.
I want AMD to succeed, but their marketers really make it hard to like them.
At least it's not as bad* as the Athlon XP+ days, when they named the processors after the relative performance compared to their older CPUs. "Wait, so the Athlon 3200+ is actually running at 2200MHz?"
*Referring only the the PR campaigns of AMD and Intel, not the actual CPUs themselves.
Show a consumer a P4 2.4ghz, for example and an Athlon 2.2ghz and they'll go for the 'bigger' number every time, even though the Athlon 2.2ghz would walk all over the 2.4ghz Intel chip - so there was at least a fully justified reason for it.
So they had to compare it to something, or they'd be tanked on the showroom floor. Performance rating helped with that.
I'd like to know the low end power usage, idle figures, frankly. I have a CPU monitor on my taskbar (linux, natch) and the CPU doesn't spin up very often - so total TDP isn't that relevant....
But yeah, not ideal by any stretch, and aside from that, AMD need to GTF away from the Piledriver microarch - it's seriously creaking these days, it's sad to say.
Contrary to myth, the performance ratings of the AMD CPUs were not created for comparison to Intel CPUs. There were many people that assumed that to be the case, but the product training that AMD supplied to retail employees (CompUSA, Best Buy, etc.) specifically called out comparisons to AMD's own CPUs when addressing the PR, not Intel's. If I had kept that old literature from my retail days, I would scan it for you to see, but you'll just have to take my word for it. :|
What Flunk said. You were told that lie to keep AMD from being sued. Everyone knew it was a comparison between the performance of the AMD processors and the rated speed of the Intel chips.
I sure miss being able to build an affordable high end rig. Yes, my main rig is an i7 and is a monster. That's pretty much what us enthusiasts have had to do. I just retired my first dual-core (Athlon 64 X2 4400) due to a failing mobo. Makes me sad.
Yep, but now Intel needs a processor rating system since AMD is the one pushing ridiculous clock speeds with diminishing returns. 220w for 5GHz?
An 88w Intel 4790K @ 4GHz stock is faster. And although it costs more, there is more performance headroom in overclocking (4.6GHz+ isn't uncommon) and the fact is runs on nearly 1/3rd the power at full load indicates a lower electric bill and lower cooling costs.
I also question the longterm reliability of motherboard VRM's when pumping out 220w of juice...
Why would Intel need a rating? They have proven that it doesn't matter if they have the better CPU in regards to and. Their still going to sell boatloads more. Since they took back the performance lead we get slot of side ways jumps but nothing like the jump that was seen with the core & athlon before it.
Might be a nice upgrade from my FX-6300. I mainly use that machine for video encoding (Handbrake or Makemkv), so will be keeping an eye here for benchmarks.
The 6-packs remind me of AMD in the late 90's. They used to have 5 packs, and they also came with a sheet of stickers for the front of machines. That was a popular option back when small computer stores were thriving... Buy 5 AMD processors, build 5 machines, post a sale sign, and profit. ;-)
I'm not sure how much of that goes on nowadays. The small clone builder culture seems to have mostly dried up...
Wish they'd repackage these into FM2+ like those other FX CPUs they released, then we can use them on more modern motherboards with more form factor variety (I'm looking at you, mini-itx).
I'd prefer they release A-series processors that totally blow the FX chips out of the water, then AMD can focus on one line of desktop processors instead of two.
They basically are. These are nothing more than slight updates to CPU's that came out almost 2 years ago. They have somewhat lower power usage which means pretty much nothing in the desktop world. It's AMD trying to milk a little more life out of a processor line that is basically dead now. The future is all about the A series for them.
These very minor bumps come from power and frequency binning as well as slightly refined manufacturing tolerances. No new R&D went into making these chips, there's no architecture refinements and no process improvements. As for their remaining line of processors, they're basically laptop chips on steroids. Okay if you need on-board graphics but quoting the Anandtech review: "a dual-core Core i3 Haswell will deliver much better CPU performance than even the fastest Kaveri at a lower price." so if you care enough about graphics to get a discrete card you won't buy it. If you don't care about graphics at all, you won't buy it. I guess it's the cheap all-in-ones but it's the jack of all trades, master of none.
Well I've read some sources that the base clock is also 100Mhz higher. Not that this changes much, and overall I would tend to agree with what you've said. On the other hand if it's a better bin it might be a better overclocker. At a minimum new releases might help to push prices down on the lower models.
Out of curiousity, does anyone know if it's still C0 stepping?
I want AMD to succeed as much as anyone, as competition benefits the technology and the end user, but a TDP of 225 watts for a CPU that can't outperform an Intel processor that generates less than half that much heat? AMD has a long way to go...
Don't really care much about AM socket CPUs anymore, I'm very interested in FM socket CPUs though, I like building a mini ITX systems as a hobby at least once a year.
Come On! All of these FX processors uses Double the Power while being like 50% Slower. That is a 4x difference in pref / watts! Has AMD given up completing in x86 market?
The table in the article has the number of modules for each chip. The integer cores are 2x the number of modules, so all of these processors are 8 cores.
I find some amusement in the complaints regarding AMD's product naming schemes. It's not as intuitive as it should be, but it could be worse. The general rule - as should be understood by now - is that the first number is the core count, the second is the generation (in odd numbers), the third is its position within the hierarchy (i.e. bigger is better) and the fourth is always a zero. It's only really the 9xxx series which has fouled things up totally. The real shame is that we don't get a 200MHz stepping per every 10 in the model number like we did with Phenom II.
"Base frequency" is mostly meaningless with modern CPUs. If the CPU is lightly loaded, you might find it running at 600 MHz or so. Fully load at least one core, and that core will jump to the "turbo" speed. What if you run all cores flat out? Depends on the instruction mix whether the processor will hit the power cap and slow down. How much will it slow down? Depends on the code you are running. I have an AMD 6-core CPU and I dedicate 4 cores to Prime95 if I am not doing other number crunching. Such as trying to get a few percent more performance out of code which fully loads the floating point cores. (How do I get more crunching? Tweaking the code to decrease L2 collisions.)
Now running my fp-code for linear programming, you will top out the fp-units and put a pretty heavy load on the integer CPU portion as well, doing indexing. On an AMD CPU you might as well run the code on half the cores (and make sure to use only even or only odd) cores since that will max out both the fp units and memory access. So what is the right number for max performance? Here you have say three cores running flat out, three cores mostly in halt states, and performance is effectively gated by main memory access. (Yes, I have code that runs much faster when running with all data to and from L2. But for large arrays, you have to use main memory if only to read the data (once) and write the results. And as I said, I keep trying to reduce L2 collisions..)
Oh, and I hope AMD certs these CPUs for DDR3-2166, even though I run my memory there anyway. ;-) Or they could finally start shipping DDR4 SKUs. AFAIK, their CPUs have been DDR4 ready for at least six months, but they need someone to provide mobos.
hey, AMD, you know what would be much more impressive and sell way more than yet ANOTHER binned piledriver chip from 2012? AN EIGHT CORE STEAMROLLER CPU!. or a six core. built on 28nm, of course. You are not intel, you cant say "this is what consumers want" and expect people to buy it. Stop making us wait already and give us a new fx steamroller chip.
It wouldn't work. The process is tuned more towards power savings and as such won't clock as high. If you were to overclock to at least 8350 levels, it'd probably start to use way too much power... precisely the reason why the architecture is flawed.
In any case, there's not a large difference in single threaded performance, and floating point performance has, rather strangely, regressed slightly, so an FX-class chip would only make sense for multithreading, not gaming.
Being a big AMD fan this is kind of discouraging. No new enthusiast dekstop CPU's for years now. Just crappy refreshes and reworks of the power hungry Piledriver architecture. My next system will be X99. Farewell AMD. My wallet will never be the same....
AMD will hardly stand a chance, if developers continue to use Intel's compiler and don't patch their products to stop identifying the CPU on the system and running the "most compatible" (read -much slower-) routine every time they find an AMD chip.
Out of respect for their customers, software developers should offer a correctly and completely optimized product.
Intel's compiler of one of the best ones on the market and it's free. But if Intel is free not to play fair, the software developer should do his duty to his customer and offer a correctly optimized product.
I believe that AnandTech should check to see if the benchmarks used treat the processors objectively and point out to the readers the tests that involve software that runs the "most compatible" routine, each time it finds and AMD chip.
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
40 Comments
Back to Article
JDG1980 - Tuesday, August 26, 2014 - link
I don't like at all that they're failing to specify base clock speeds. Will this become a trend like it has with their newer graphics cards? For CPUs, that's absolutely unacceptable.I want AMD to succeed, but their marketers really make it hard to like them.
nathanddrews - Tuesday, August 26, 2014 - link
At least it's not as bad* as the Athlon XP+ days, when they named the processors after the relative performance compared to their older CPUs. "Wait, so the Athlon 3200+ is actually running at 2200MHz?"*Referring only the the PR campaigns of AMD and Intel, not the actual CPUs themselves.
Beany2013 - Tuesday, August 26, 2014 - link
Actually, it was meant to help consumers understand that a 2200mhz Athlon was trading blows (well, certainly not far off) from the 3.2ghz P4s of the day:http://ixbtlabs.com/articles2/roundupmobo/p4-3.2gh...
Show a consumer a P4 2.4ghz, for example and an Athlon 2.2ghz and they'll go for the 'bigger' number every time, even though the Athlon 2.2ghz would walk all over the 2.4ghz Intel chip - so there was at least a fully justified reason for it.
So they had to compare it to something, or they'd be tanked on the showroom floor. Performance rating helped with that.
I'd like to know the low end power usage, idle figures, frankly. I have a CPU monitor on my taskbar (linux, natch) and the CPU doesn't spin up very often - so total TDP isn't that relevant....
But yeah, not ideal by any stretch, and aside from that, AMD need to GTF away from the Piledriver microarch - it's seriously creaking these days, it's sad to say.
nathanddrews - Tuesday, August 26, 2014 - link
Contrary to myth, the performance ratings of the AMD CPUs were not created for comparison to Intel CPUs. There were many people that assumed that to be the case, but the product training that AMD supplied to retail employees (CompUSA, Best Buy, etc.) specifically called out comparisons to AMD's own CPUs when addressing the PR, not Intel's. If I had kept that old literature from my retail days, I would scan it for you to see, but you'll just have to take my word for it. :|Flunk - Tuesday, August 26, 2014 - link
Yes, that was what AMD claimed to avoid being sued.bigboxes - Tuesday, August 26, 2014 - link
What Flunk said. You were told that lie to keep AMD from being sued. Everyone knew it was a comparison between the performance of the AMD processors and the rated speed of the Intel chips.I sure miss being able to build an affordable high end rig. Yes, my main rig is an i7 and is a monster. That's pretty much what us enthusiasts have had to do. I just retired my first dual-core (Athlon 64 X2 4400) due to a failing mobo. Makes me sad.
Samus - Tuesday, August 26, 2014 - link
Yep, but now Intel needs a processor rating system since AMD is the one pushing ridiculous clock speeds with diminishing returns. 220w for 5GHz?An 88w Intel 4790K @ 4GHz stock is faster. And although it costs more, there is more performance headroom in overclocking (4.6GHz+ isn't uncommon) and the fact is runs on nearly 1/3rd the power at full load indicates a lower electric bill and lower cooling costs.
I also question the longterm reliability of motherboard VRM's when pumping out 220w of juice...
just4U - Wednesday, August 27, 2014 - link
Why would Intel need a rating? They have proven that it doesn't matter if they have the better CPU in regards to and. Their still going to sell boatloads more. Since they took back the performance lead we get slot of side ways jumps but nothing like the jump that was seen with the core & athlon before it.just4U - Wednesday, August 27, 2014 - link
Bah tablet typing.. sorry.piroroadkill - Tuesday, August 26, 2014 - link
So it's still crap, but at least it won't be as hot. This wasn't as exciting as it should have been.barleyguy - Tuesday, August 26, 2014 - link
Might be a nice upgrade from my FX-6300. I mainly use that machine for video encoding (Handbrake or Makemkv), so will be keeping an eye here for benchmarks.The 6-packs remind me of AMD in the late 90's. They used to have 5 packs, and they also came with a sheet of stickers for the front of machines. That was a popular option back when small computer stores were thriving... Buy 5 AMD processors, build 5 machines, post a sale sign, and profit. ;-)
I'm not sure how much of that goes on nowadays. The small clone builder culture seems to have mostly dried up...
ayqazi - Tuesday, August 26, 2014 - link
Wish they'd repackage these into FM2+ like those other FX CPUs they released, then we can use them on more modern motherboards with more form factor variety (I'm looking at you, mini-itx).ayqazi - Tuesday, August 26, 2014 - link
Whoops, I realised my mistake: I thought they had released FX parts on FM2+, they obviously haven't.Flunk - Tuesday, August 26, 2014 - link
I'd prefer they release A-series processors that totally blow the FX chips out of the water, then AMD can focus on one line of desktop processors instead of two.Jumangi - Tuesday, August 26, 2014 - link
They basically are. These are nothing more than slight updates to CPU's that came out almost 2 years ago. They have somewhat lower power usage which means pretty much nothing in the desktop world. It's AMD trying to milk a little more life out of a processor line that is basically dead now. The future is all about the A series for them.Kjella - Tuesday, August 26, 2014 - link
These very minor bumps come from power and frequency binning as well as slightly refined manufacturing tolerances. No new R&D went into making these chips, there's no architecture refinements and no process improvements. As for their remaining line of processors, they're basically laptop chips on steroids. Okay if you need on-board graphics but quoting the Anandtech review: "a dual-core Core i3 Haswell will deliver much better CPU performance than even the fastest Kaveri at a lower price." so if you care enough about graphics to get a discrete card you won't buy it. If you don't care about graphics at all, you won't buy it. I guess it's the cheap all-in-ones but it's the jack of all trades, master of none.just4U - Wednesday, August 27, 2014 - link
I'd take. A A8-7600 over any i3 on the market today.. the CPU may be slower but for most that's not going to be noticed....flyingpants1 - Wednesday, August 27, 2014 - link
I'm sorry, what?Masospaghetti - Tuesday, August 26, 2014 - link
AMD's marketing is awful. I've rooted for them forever, but they must have a bunch of 10 year olds coming up with new names and brands.Flunk - Tuesday, August 26, 2014 - link
New chips that slot into the middle of their product line aren't particularly exciting. 8350 + 100Mhz of turbo = 8370 I guess.Alexvrb - Tuesday, August 26, 2014 - link
Well I've read some sources that the base clock is also 100Mhz higher. Not that this changes much, and overall I would tend to agree with what you've said. On the other hand if it's a better bin it might be a better overclocker. At a minimum new releases might help to push prices down on the lower models.Out of curiousity, does anyone know if it's still C0 stepping?
mgilbert - Tuesday, August 26, 2014 - link
I want AMD to succeed as much as anyone, as competition benefits the technology and the end user, but a TDP of 225 watts for a CPU that can't outperform an Intel processor that generates less than half that much heat? AMD has a long way to go...hojnikb - Tuesday, August 26, 2014 - link
They should rather focus on the chipset front, insteand of releasing new bins of old arch.9 series is getting really old now.
SlowSpyder - Tuesday, August 26, 2014 - link
Awesome seeing AM3+ still getting some love from AMD!Malih - Tuesday, August 26, 2014 - link
Don't really care much about AM socket CPUs anymore, I'm very interested in FM socket CPUs though, I like building a mini ITX systems as a hobby at least once a year.willis936 - Tuesday, August 26, 2014 - link
>new FX SKUsWoohoo!
>Vishera
:(
iwod - Tuesday, August 26, 2014 - link
Come On! All of these FX processors uses Double the Power while being like 50% Slower.That is a 4x difference in pref / watts!
Has AMD given up completing in x86 market?
LarsBars - Tuesday, August 26, 2014 - link
If these new SKUs were steamroller or excavator, I think there would be a lot more excitement around them...LarsBars - Tuesday, August 26, 2014 - link
Ha! I like how Brett's AMD FX CPU comparison doesn't even have a core count, that's great! :)barleyguy - Tuesday, August 26, 2014 - link
The table in the article has the number of modules for each chip. The integer cores are 2x the number of modules, so all of these processors are 8 cores.silverblue - Tuesday, August 26, 2014 - link
I find some amusement in the complaints regarding AMD's product naming schemes. It's not as intuitive as it should be, but it could be worse. The general rule - as should be understood by now - is that the first number is the core count, the second is the generation (in odd numbers), the third is its position within the hierarchy (i.e. bigger is better) and the fourth is always a zero. It's only really the 9xxx series which has fouled things up totally. The real shame is that we don't get a 200MHz stepping per every 10 in the model number like we did with Phenom II.silverblue - Tuesday, August 26, 2014 - link
(with a general understanding that the bottom of the rung models would be 2.4GHz as they had two trailing zeroes)eachus - Tuesday, August 26, 2014 - link
"Base frequency" is mostly meaningless with modern CPUs. If the CPU is lightly loaded, you might find it running at 600 MHz or so. Fully load at least one core, and that core will jump to the "turbo" speed. What if you run all cores flat out? Depends on the instruction mix whether the processor will hit the power cap and slow down. How much will it slow down? Depends on the code you are running. I have an AMD 6-core CPU and I dedicate 4 cores to Prime95 if I am not doing other number crunching. Such as trying to get a few percent more performance out of code which fully loads the floating point cores. (How do I get more crunching? Tweaking the code to decrease L2 collisions.)Now running my fp-code for linear programming, you will top out the fp-units and put a pretty heavy load on the integer CPU portion as well, doing indexing. On an AMD CPU you might as well run the code on half the cores (and make sure to use only even or only odd) cores since that will max out both the fp units and memory access. So what is the right number for max performance? Here you have say three cores running flat out, three cores mostly in halt states, and performance is effectively gated by main memory access. (Yes, I have code that runs much faster when running with all data to and from L2. But for large arrays, you have to use main memory if only to read the data (once) and write the results. And as I said, I keep trying to reduce L2 collisions..)
Oh, and I hope AMD certs these CPUs for DDR3-2166, even though I run my memory there anyway. ;-) Or they could finally start shipping DDR4 SKUs. AFAIK, their CPUs have been DDR4 ready for at least six months, but they need someone to provide mobos.
chiechien - Tuesday, August 26, 2014 - link
New Piledriver cores? Not helping.TheinsanegamerN - Tuesday, August 26, 2014 - link
hey, AMD, you know what would be much more impressive and sell way more than yet ANOTHER binned piledriver chip from 2012? AN EIGHT CORE STEAMROLLER CPU!. or a six core. built on 28nm, of course. You are not intel, you cant say "this is what consumers want" and expect people to buy it. Stop making us wait already and give us a new fx steamroller chip.silverblue - Tuesday, August 26, 2014 - link
It wouldn't work. The process is tuned more towards power savings and as such won't clock as high. If you were to overclock to at least 8350 levels, it'd probably start to use way too much power... precisely the reason why the architecture is flawed.In any case, there's not a large difference in single threaded performance, and floating point performance has, rather strangely, regressed slightly, so an FX-class chip would only make sense for multithreading, not gaming.
roadapathy - Wednesday, August 27, 2014 - link
Oh, so it's -still- 32nm wafer. Yeah. Talk about putting lipstick on a pig. Does the PR department run AMD or do they sometimes listen to the techies?Leonard Emery - Wednesday, August 27, 2014 - link
Being a big AMD fan this is kind of discouraging. No new enthusiast dekstop CPU's for years now. Just crappy refreshes and reworks of the power hungry Piledriver architecture. My next system will be X99. Farewell AMD. My wallet will never be the same....flyingpants1 - Wednesday, August 27, 2014 - link
But you knew this would happen. This is how capitalism works. What exactly do you expect them to do?East17 - Wednesday, September 3, 2014 - link
AMD will hardly stand a chance, if developers continue to use Intel's compiler and don't patch their products to stop identifying the CPU on the system and running the "most compatible" (read -much slower-) routine every time they find an AMD chip.Out of respect for their customers, software developers should offer a correctly and completely optimized product.
Intel's compiler of one of the best ones on the market and it's free. But if Intel is free not to play fair, the software developer should do his duty to his customer and offer a correctly optimized product.
I believe that AnandTech should check to see if the benchmarks used treat the processors objectively and point out to the readers the tests that involve software that runs the "most compatible" routine, each time it finds and AMD chip.