AMD Piledriver rumours ... and expert conjecture

[Reynod]

We have had several requests for a sticky on AMD's yet to be released Piledriver architecture ... so here it is.

I want to make a few things clear though.

Post a question relevant to the topic, or information about the topic, or it will be deleted.

Post any negative personal comments about another user ... and they will be deleted.

Post flame baiting comments about the blue, red and green team and they will be deleted.

Enjoy ...

 

[sarinaide]

Oh yeah, a 5Ghz Zambezi should be faster overall compared to a 4.2Ghz PhII; I don't argue that point. I'm just saying, when in the OC territory, PhIIs tend to have a better showing since Zambezi doesn't OC that far away from them and at 4.5Ghz, it's either equal or worse than a PhII even below 4.2Ghz according to most benchies out there. That's one of the things that Vishera should have fixed... I hope... Higher OC room and a tad better IPC, making it an upgrade from heavily OCed PhIIs.

Anyway, if SR's desktop incarnation is indeed headed towards AM3+, I might skip Vishera as well. Not that I like the sound of that anyway... I'd rather have a new platform at this point =/

Cheers!

I will say you are correct to wait for Kaviri, with a new chipset and motherboard revisions spending now on a 990FX only to find a new chipset offering refreshed features is a waste of money, that said all SR chips will operate on any current 900 series motherboard after bios updates, but you will not get the features limited to Steam Roller parts.

 

[sarinaide]

Been reading the debate on AMD's HSA approach on OCN, its all fine and well for AMD to offer this path but if the rest of the players, that is writers and developers don't jump on board it will be a massive opportunity for all, it really changes the computing landscape but sadly apart from a small handfull of developers it seems like they are not interested or too lazy.

 

[gamerk316]

Id say closer to 50/50
http://images.hardwarecanucks.com/image/mac/reviews/AMD/Bulldozer/AMD_FX-8150-201.jpg

Even on toms review, BD was within 1 fps of the phenom II 980 in all the game benches (wich is actually clocked 100 mhz faster)

Proving the existence of the GPU bottleneck in the configurations tested. Hence why I again argue FPS in an insufficient benchmarking tool for games. Average Frame Latency gives you a FAR better picture of things.

 

[truegenius]

Id say closer to 50/50
http://images.hardwarecanucks.com/image/mac/reviews/AMD/Bulldozer/AMD_FX-8150-201.jpg

Even on toms review, BD was within 1 fps of the phenom II 980 in all the game benches (wich is actually clocked 100 mhz faster)

As I said, when people use selective "cherry picking" on what you want to believe, sure, only look at the red numbers and delete all the black ones.

Is Phenom faster than BD? id say you have a 50/50 shot at the same clock speed of having a miniscule measureable difference of 0.1-5% either way.

Does PD improve on BD? without a doubt, but we won't actually KNOW how much till NDA is lifted.

if its a stock chart
then phenom lacks turbo

 

[sarinaide]

At first it was FPS, when that failed we moved on to frame latency. Well thats not going to affect FPS much anyways because the results will still be the same irrespective, as for transition times neither is noticeable so the latency phenomena is still rather moot at this juncture.

It is small irrelevancies that have drawn people needlessly away from AMD, despite the mootness of AMD taking on a architecture that is probably two years before its time, despite all the single threaded trials and tribulations the bottom line is a AMD processor still remains excessive, I don't really ever recall so much negitive sentiment about AMD trying something when Intel basically screwed around with P4 and P-D. AMD did it clean and on resources less than intel pays its high ranking officers what did you expect?

I am starting to get irritated at how everything is stark minimal in this thread and when you jump to the other thread the figures are 15% improvement and 2x improvement to iGPU, both spin off fanfares we will let that disappointment sink in next year.

Now I had my rant for the day and feel better thanks for asking 😀

 

[Dangi]

Now that we have de new FM2 APU's is there any info regarding the other APU's made by AMD, the profesional one with the FirePro integrated instead the Radeons ???

That should be interesting, having your iGPU for minor rendering, a CPU that is a bit weaker than a i3 put where you can put a 6670 and end up with performance close to an i3

 

[truegenius]

so 8350 is only 15% improvement over 8150 ?

base clock of 8350 with resect to 8150 is ~11% higher so is that mean only 4% gain via ipc gains

my point is 8350 clearly sounds ~20-25% gain over 8150

ivy was a die shrink with minimal improvement and got a ~5-10% performance boost crown

and bd is much improveable and is only getting ~10%


which apu is going to/using pcie instead of ht?

 

[de5_Roy]

AMD Richland uses FM2 socket
http://www.fudzilla.com/home/item/28986-amd-richland-uses-fm2-socket

little offtopic
AMD to launch something “mutated” on October 9
http://semiaccurate.com/2012/10/05/amd-to-launch-something-mutated-on-october-9/

 

[Cazalan]

hexus tests trinity's 'dual graphics'
http://hexus.net/tech/reviews/cpu/46157-amd-a10-5800k-dual-graphics-evaluation/
indeed that is something intel cannot do.

Some interesting numbers on dual graphics.
Great for the budget gamer as those cards are around $60.

 

[Cazalan]

well 10-15% is there, in some instances its higher some up to 25ish% others as low as a percent, some actually less but irrelevent to performance. Overall though the 10-15 is about as narrow as I want to go, if I broadened it I would say more along the 7-11% range.

The highest percentages only occur if the newer Trinity instructions are used. Programs using FMA3 for instance which Bulldozer didn't support, but Ivy/Sandy Bridge does.

 

[$hawn]

Hey, does anyone seem to know the story behind the FX 8350's pathetic 200MHz turbo speed? I mean for a 4GHz chip, 200MHz is peanuts!! That's only 5%!!
If the 8320 can ramp up by a whole 500MHz (+14%), then why is the flagship processor so limited in its turbo?

My guess is that the 8350 vishera chips are already pushed to the maximum clock frequency possible within a 125W TDP. Probably RCM's tuned frequencies have something to do here. Even the previous generation falgship FX-8150 had a generous 600 MHz turbo.

 

[-Fran-]

Hey, does anyone seem to know the story behind the FX 8350's pathetic 200MHz turbo speed? I mean for a 4GHz chip, 200MHz is peanuts!! That's only 5%!!
If the 8320 can ramp up by a whole 500MHz (+14%), then why is the flagship processor so limited in its turbo?

My guess is that the 8350 vishera chips are already pushed to the maximum clock frequency possible within a 125W TDP. Probably RCM's tuned frequencies have something to do here. Even the previous generation falgship FX-8150 had a generous 600 MHz turbo.

You're looking at it the other way around.

The CPU will now stay more at 4Ghz than at 4.2Ghz. Remember Turbo is a very situational speed, when you're not taxing every module and the CPU has room to speed up.

So, it's not the "lack of turbo" you should look at, but the increase in "base speed". Overall, since PD won't improve that much in the 32nm process, you'll probably will be within striking distance of Zambezi's overclocking potential at 140W or 150W. I really want to know what speed you'll get when reaching that power consumption, since PhIIs are comfortable in those ceilings.

Cheers!

 

[fazers_on_stun]

Been reading the debate on AMD's HSA approach on OCN, its all fine and well for AMD to offer this path but if the rest of the players, that is writers and developers don't jump on board it will be a massive opportunity for all, it really changes the computing landscape but sadly apart from a small handfull of developers it seems like they are not interested or too lazy.

Can you link this discussion? I'd like to see whassup with HSA as well. If the devs don't wanna support it, then I guess it would be up to AMD to do the heavy lifting, if they can afford it, or else cajole their partners to share the burden. If Intel jumped on that bandwagon, then it would almost assuredly take off since the devs would be guaranteed the entire x86 market would support it, as opposed to just AMD's share..

 

[sarinaide]

The only HSA players seem to be AMD, SAMSUNG, ARM, Oracle and a few other independent developers. The issue is that HSA is still early, possibly 3 years from any semblance of full HSA, this is why I am wondering as to why BD/PD were so heavily laden towards this. We may only start seeing vestiges of HSA based computing during 2013 but still relatively low scale.

 

[fazers_on_stun]

At first it was FPS, when that failed we moved on to frame latency. Well thats not going to affect FPS much anyways because the results will still be the same irrespective, as for transition times neither is noticeable so the latency phenomena is still rather moot at this juncture.

It is small irrelevancies that have drawn people needlessly away from AMD, despite the mootness of AMD taking on a architecture that is probably two years before its time, despite all the single threaded trials and tribulations the bottom line is a AMD processor still remains excessive, I don't really ever recall so much negitive sentiment about AMD trying something when Intel basically screwed around with P4 and P-D. AMD did it clean and on resources less than intel pays its high ranking officers what did you expect?

I am starting to get irritated at how everything is stark minimal in this thread and when you jump to the other thread the figures are 15% improvement and 2x improvement to iGPU, both spin off fanfares we will let that disappointment sink in next year.

Now I had my rant for the day and feel better thanks for asking 😀

Well gameplay smoothness was always an AMD benefit previously, as jerky movement/microstuttering can be pretty distracting in intense combat. But just like video encoding quality 😛, beauty is in the eye of the beholder (aka EOB I - III 😀).

As for the performance deflation for PD in this thread, probably due to PD's being just days away from release and a fairly decent gauge can be made from desktop Trinity. OTOH, Haswell is still some 8+ months away, not much detail given out at IDF, so we are free to speculate up the wazoo, unburdened by actual facts and other pesky stuff 😀.

 

[fazers_on_stun]

You're looking at it the other way around.

The CPU will now stay more at 4Ghz than at 4.2Ghz. Remember Turbo is a very situational speed, when you're not taxing every module and the CPU has room to speed up.

So, it's not the "lack of turbo" you should look at, but the increase in "base speed". Overall, since PD won't improve that much in the 32nm process, you'll probably will be within striking distance of Zambezi's overclocking potential at 140W or 150W. I really want to know what speed you'll get when reaching that power consumption, since PhIIs are comfortable in those ceilings.

Cheers!

Hmm, I'd hafta agree with $hawn on this one - a 200MHz turbo ain't much of a turbo at 4GHz base. And I agree with his thinking it has something to do with RCM, as too far outside the resonant frequency means there's no power savings..

Now if it was a 4 Hz base, turboing up to 2,000,004 Hz, then that would have been impressive! About 20 years ago, that is 😀..

 

[sarinaide]

Hmm, I'd hafta agree with $hawn on this one - a 200MHz turbo ain't much of a turbo at 4GHz base. And I agree with his thinking it has something to do with RCM, as too far outside the resonant frequency means there's no power savings..

Now if it was a 4 Hz base, turboing up to 2,000,004 Hz, then that would have been impressive! About 20 years ago, that is 😀..

If i am being hypercritical (which I am #superseriousface#) then I will say BD and PD failed to achieve the clockspeed frontier AMD were banking on. BD was intended to be 4Ghz for FX 4XXX parts up to 4.5ghz turbo, the FX 6XXX was intended to be a 4.2ghz with a turbo around 4.6ghz while the high end parts were expected to have 4.3ghz base clocks and up to 4.7ghz on turbo. PD did not achieve this either. It does disappoint me a bit as its free performance on clock speed which is not really a problem, the higher frequencies mitigate the loss of IPC, hence a 4.3ghz BD part would have given out impressive results.

4-4.2ghz is modest if you ask me.

 

[jdwii]

At first it was FPS, when that failed we moved on to frame latency. Well thats not going to affect FPS much anyways because the results will still be the same irrespective, as for transition times neither is noticeable so the latency phenomena is still rather moot at this juncture.

It is small irrelevancies that have drawn people needlessly away from AMD, despite the mootness of AMD taking on a architecture that is probably two years before its time, despite all the single threaded trials and tribulations the bottom line is a AMD processor still remains excessive, I don't really ever recall so much negitive sentiment about AMD trying something when Intel basically screwed around with P4 and P-D. AMD did it clean and on resources less than intel pays its high ranking officers what did you expect?

I am starting to get irritated at how everything is stark minimal in this thread and when you jump to the other thread the figures are 15% improvement and 2x improvement to iGPU, both spin off fanfares we will let that disappointment sink in next year.

Now I had my rant for the day and feel better thanks for asking 😀

2X improvement and 15% across the board some people are living in a dream lol Trinity is only around 20-30% better on their graphics and 10-15% better on the CPU which is a great improvement not to mention power consumption is quite well.

so 8350 is only 15% improvement over 8150 ?

base clock of 8350 with resect to 8150 is ~11% higher so is that mean only 4% gain via ipc gains

my point is 8350 clearly sounds ~20-25% gain over 8150

ivy was a die shrink with minimal improvement and got a ~5-10% performance boost crown

and bd is much improveable and is only getting ~10%


which apu is going to/using pcie instead of ht?

How is this so when Amd only said 10-15% boost(in IPC+Power savings) people are pulling numbers in the wrong places not to mention in some of the rarest benchmarks i'm sure it will be 20% better. Even a 15% boost in just one year(don't forget they had samples in June 2011 right before Llano hit newegg) is impressive. I'd say well get a 15% boost in performance while having some improvements in power consumption i just hope it comes at the right price.

 

[sarinaide]

The way I see it is;

AMD adopting a radical approach to CPU architecture introducing modulization which in principle is very sound, but like anything radical it takes time to iron out so the designated evolutions starting from BD ending at Excavator which was bracketed as Phase 1 HSA. This suggests that with BD being step 1 basically being a unknown quantum, this does now leave AMD with significant improvement path.

Conversely

Intel have been flogging its efficiency and IPC laden architecture since Core2, the evolutionary process is getting less, it is now harder to improve per core performance without alleviating the potential bottlenecks allowing for this. This was not dissimilar from AMD flogging K architecture until every cow in the barn had been milked.

 

[-Fran-]

In all fairness to AMD, they're improving power numbers in the same 32nm (crappy) node from GloFo. That has to be some sort of feat by itself, haha.

Trinity is quite a good showing of what AMD can squeeze out of what it has. The same as with VLIW5 to 4.

So that 15% is a very good one. Intel got a big power reduction and a modest 10% increase (mostly due to the 100Mhz bumps) and it was "good".

Cheers!

 

[Cazalan]

I recall the Bulldozer overclocked power consumption was significantly higher than stock. By 100-150 Watts.

Trinity is showing only a 50W increase to 4.5Ghz with GPU overclocked as well. Could be a good sign for Vishera.

http://www.tomshardware.com/reviews/a10-5800k-trinity-efficiency,3315-10.html

 

[sarinaide]

When I hear Global Foundries, I see Hector Ruiz and I see me sticking knives in him with a great big smile on my face.....

Well AMD basically wrote off GloFo debts which whacked them hard but splitting from GloFo is the best decision in a long while.

 

[gamerk316]

You're looking at it the other way around.

The CPU will now stay more at 4Ghz than at 4.2Ghz. Remember Turbo is a very situational speed, when you're not taxing every module and the CPU has room to speed up.

So, it's not the "lack of turbo" you should look at, but the increase in "base speed". Overall, since PD won't improve that much in the 32nm process, you'll probably will be within striking distance of Zambezi's overclocking potential at 140W or 150W. I really want to know what speed you'll get when reaching that power consumption, since PhIIs are comfortable in those ceilings.

Cheers!

Look at it the other way again though: Is there a reason why the Turbo isn't slightly more aggressive? Heat/power draw problems, prehaps? Very interested to see how high a normal chip can be OC'd to normally...

 

[gamerk316]

The way I see it is;

AMD adopting a radical approach to CPU architecture introducing modulization which in principle is very sound, but like anything radical it takes time to iron out so the designated evolutions starting from BD ending at Excavator which was bracketed as Phase 1 HSA. This suggests that with BD being step 1 basically being a unknown quantum, this does now leave AMD with significant improvement path.

Conversely

Intel have been flogging its efficiency and IPC laden architecture since Core2, the evolutionary process is getting less, it is now harder to improve per core performance without alleviating the potential bottlenecks allowing for this. This was not dissimilar from AMD flogging K architecture until every cow in the barn had been milked.

But lets take that discussion even farther: What happens when Intel decides to slap more cores on their CPU? They win again, because of IPC. So AMD's entire architecture relies almost entirely on programs scaling to "x" CPU cores, which is unrealistic for most workloads. So if Intel moved to 8 cores, and AMD decided to move to 16 to retaliate, guess what? Intel wins because the S/W won't scale that high, and IPC pulls intel ahead.

My point being, its easier to scale up a design that already has high IPC, as adding more cores, from a technical perspective, is easy to accomplish compared to increasing IPC.

 

[esrever]

But lets take that discussion even farther: What happens when Intel decides to slap more cores on their CPU? They win again, because of IPC. So AMD's entire architecture relies almost entirely on programs scaling to "x" CPU cores, which is unrealistic for most workloads. So if Intel moved to 8 cores, and AMD decided to move to 16 to retaliate, guess what? Intel wins because the S/W won't scale that high, and IPC pulls intel ahead.

My point being, its easier to scale up a design that already has high IPC, as adding more cores, from a technical perspective, is easy to accomplish compared to increasing IPC.

8 core sandybridge will have TPD of like 160W, its still a double edged sword. Ivy might shave off some but even then it would be unlikely that the cpu will be below 130W with the same frequency. I don't know about you but most workloads that can be threaded can be subdivided very well across more and more cores. Amdual's law doesn't kick in until more than 64 cores. By the time AMD offers 16 core consumer CPUs, AMD would have hoped to have pushed HSA to the point where they won't have to do it. Higher IPC also requires more transistors and die area as well as other things, don't think its "free" like you assume.