AMD Piledriver rumours ... and expert conjecture

Page 99 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
Reynod

Reynod

Administrator
Mar 26, 2007
14,921
16
42,215
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 ...
 
Sort by date Sort by votes
Hard to bench reliably. You'd basically need your own server and have a group of people carry out a pre-planned routine in order to get something that was even somewhat reliable as a benchmark.
Click to expand...
While this is true, personally I want to know how your cpu will handle unknown variables. The problem is to get an accurate picture on MP runs, you would probably need to play for about 30 minutes on each test to get all the highs and lows for each system.

Last thing I want to do is go buy a cpu and game (bf3) on a review like this, granted they did a better job than any others due to their "cpu usage" chart. http://www.techspot.com/review/458-battlefield-3-performance/page7.html

CPU_02.png


According to just the chart, the Phenom X2 560 can run just fine (granted its pegged at 100% on the cpu chart, kudos to having that on reviews) ... but come home and end up with this

2507


As for multiplayer vs single player, I don't know of very many games that don't take a huge performance hit with more and more outside players. WoW for one was unplayable on a low end system in a 40 player MC raid, but you could run around all day long leveling up with 100 people in a single zone (just not in your area). That doesn't mean WoW is better programmed, it all depends on where and what your doing vs everyone else. Thats just it, there are multiple levels of multiplayer. BF 3 would probably run nearly identical to single player on a 8-player map, but 64 player absolutely kills a dual core cpu.
 
Upvote 0 Downvote
Trinity has really cut the power draw of the CPU. That was one of the biggest drawbacks of Bulldozer. Granted most of the CPU performance increase has come from the increased clock speed.

Good to see and hope hey can keep improving on that.

 
Upvote 0 Downvote
Hard to bench reliably. You'd basically need your own server and have a group of people carry out a pre-planned routine in order to get something that was even somewhat reliable as a benchmark.
Click to expand...

Capture and replay the network packets, that should be indistinguishable from true multiplayer.
 
Upvote 0 Downvote
Well, that would eliminate the need for several people for multiple runs and give accurate and, maybe, repetitive data.

You'll still need the server though, hehe.

Cheers!
Click to expand...

Yeah, I was thinking of working with the code I guess. The game would have to be modified to run in a "replay" mode. I suppose if you spent enough time with it you could make a fake server, you would essentially have to hack the network packets. Not going to happen.
 
Upvote 0 Downvote
Trinity has really cut the power draw of the CPU. That was one of the biggest drawbacks of Bulldozer. Granted most of the CPU performance increase has come from the increased clock speed.

Good to see and hope hey can keep improving on that.
Click to expand...

So far as they say. We haven't seen that in action yet though and thats what we are waiting for. No sites probably have it yet to test with either.
 
Upvote 0 Downvote
So far as they say. We haven't seen that in action yet though and thats what we are waiting for. No sites probably have it yet to test with either.
Click to expand...

Actually, if it was a clock friendly arch (with all the good and bad it might have) and they solved the power problems and low clock-ability of the 32nm process, that's a very good sign for Trinity IMO.

Now, it seems the clock mesh tech might be paying off as well.

At this point, the glass seems half full 😛

Cheers!
 
Upvote 0 Downvote
Go back a page there is a screen shot from a China site.
Could be fake of course but seems in line with the other leaks.
Click to expand...

You mean the one showing it idling at 800MHz? That wont mean much for load power though. Idle it could be good unless GFs 32nm is very leaky, which it could be considering the power draw a BD CPU has when overclocked (almost double).
 
Upvote 0 Downvote
JS still pushing the "AMD is liez!!!" about their power draw?

Its already been released, 100TDP for the above mentioned chip, less for lower chips. That means not 120, 130, 140 or 150, but 100W of thermal power that is required to be disseminated (or whatever amount is specified). AMD would be liable for a class action lawsuit from every single OEM and purchaser of their CPU if they were lying, their not lying. Do I need to repost the exact definition of TDP for both AMD and Intel? Trying to cast doubt onto their power figures is exactly like trying to cast doubt on Intel's power figures.

Recently AMD purchased power recycling technology that, according to the creator, could easily be implemented into any design with minimal effort. Tweaks to the BD uArch (Trinity isn't PD) combined with Cyclos's technology could account for the sudden jump in clock speed and the drop in power usage.

http://www.engadget.com/2012/02/22/amd-piledriver-cores-will-employ-resonant-clock-mesh/

http://gigaom.com/2012/03/09/meet-cyclos-a-startup-that-cuts-chip-power-consumption/

Also under no circumstance should anyone be comparing clock speeds from a Llano to a Trinity, their different uArchs. And while I cry a little to see the little "Phenom III" be put to pasture, if all AMD's future CPU's use this clock technology then this is a good thing.

So ask yourselves, if their able to get these clocks / power levels with a APU, what could they get with a full up PD CPU without the GPU component.
 
Upvote 0 Downvote
Well, that would eliminate the need for several people for multiple runs and give accurate and, maybe, repetitive data.

You'll still need the server though, hehe.

Cheers!
Click to expand...

Network data tends to have time stamps and synchronizing between different systems so that the packets make sense is difficult. Typically better to build an automated simulator that simulates client interactions. Would probably need vender assistance or documentation on their network structure for that. This way you don't need 64 different clients, just one big one that is simulating 64 different clients.

I know the big industrial testing suits have simulators for different kinds of loads. This wouldn't be much different.
 
Upvote 0 Downvote
JS still pushing the "AMD is liez!!!" about their power draw?

Its already been released, 100TDP for the above mentioned chip, less for lower chips. That means not 120, 130, 140 or 150, but 100W of thermal power that is required to be disseminated (or whatever amount is specified). AMD would be liable for a class action lawsuit from every single OEM and purchaser of their CPU if they were lying, their not lying. Do I need to repost the exact definition of TDP for both AMD and Intel? Trying to cast doubt onto their power figures is exactly like trying to cast doubt on Intel's power figures.

Recently AMD purchased power recycling technology that, according to the creator, could easily be implemented into any design with minimal effort. Tweaks to the BD uArch (Trinity isn't PD) combined with Cyclos's technology could account for the sudden jump in clock speed and the drop in power usage.

http://www.engadget.com/2012/02/22/amd-piledriver-cores-will-employ-resonant-clock-mesh/

http://gigaom.com/2012/03/09/meet-cyclos-a-startup-that-cuts-chip-power-consumption/

Also under no circumstance should anyone be comparing clock speeds from a Llano to a Trinity, their different uArchs. And while I cry a little to see the little "Phenom III" be put to pasture, if all AMD's future CPU's use this clock technology then this is a good thing.

So ask yourselves, if their able to get these clocks / power levels with a APU, what could they get with a full up PD CPU without the GPU component.
Click to expand...

Please don't put words in my mouth. I didn't say AMD lied, considering thats not even a AMD produced image not sure how I could. All I said was that a lower idle clock does not mean load power draw has been lowered, and a TDP can only do so much if the process is flawed. GF 32nm is obviously not the best 32nm out there and its possible that it may have leakage issues that cause BDs current high power draw, as I said before when overclocked 30ish %, it almost doubles the power draw on a BD CPU.

There are no "rumors" right now as to actual power draw of the APU. A TDP only says so many things about the APU itself, as does the arch on paper. BD looked great on paper but in reality it falls very short of it, as did K10 or NetBurst.

And we can compare clock speeds. If the only gains are made via clock speed increases then IPC has not increased which is still very important. So why not? As I said, the rumors posted thus far show a 31% higher clock speed gain for 12% better CPU performance in 3DMark. Sure its not the end all but its not impressive to see 31% only equate to 12%, especially when AMD is claiming 30% better CPU performance.

And still its all just my theory. Not sure why it bothers you this much but my theory is that Trinity will mainly only get performance gains over Llano due to higher clock speeds, not because the arch has IPC improvements.

As for the higher clock speeds, wasn't it just 10 years ago that AMD was the one who was praised for having high IPC and low clocks? Not sure why its being switched around. Intel could probably match the clock speeds and perform better, in the end AMD wont be able to keep their prices higher and make any money. AMD needs to produce another Athlon 64. Thats how they will make the money to expand even further.
 
Upvote 0 Downvote
A TDP only says so many things about the APU itself, as does the arch on paper. BD looked great on paper but in reality it falls very short of it, as did K10 or NetBurst.
Click to expand...

TDP says exactly what TDP is, nothing more, nothing less. It's the amount of heat that is required to be removed for sustained operations. If we know the performance characteristics of a uArch then we can use TDP to predict it's upper limit. That is why the TDP number is important, it shows a BD uArch running at higher clock speeds and having a GPU attached while still being under their previous TDP. Immediately this means here has been improvements in leakage and power consumption.

Not once have I argued performance of said APU, only the truthfulness of AMD's posted specs. This is something you've tried to murk up several times. Before with your "they liez with ACP" statements, then with your "so I was wrong, they don't liez with ACP but they might" and now with your "TDP is just a number, it tells us nothing, they could still be liez to us".

I have an issue with you claiming to be non-biased yet constantly posting biased comments and trying to further bias other posters. It was bad enough as a CR, but now with mod rights it's damaging to the credibility of the site. You can't have your mods posting like that and not turn into another Anandtech. You deliberately phrase your statements to insinuate your bias. Example is your positive statements in the IB thread yet your negative statements in this thread, about the same thing which is the reduction in power. Intel reduces power in IB and it's a great sunny day, AMD reduces power in BD/PD and they must be deceiving us. It's the very definition of bias.

Trinity is BD-e, Llano was a "Phenom II enhanced", radically different IPC, you can't compare clock speeds vs performance growth between them. Instead you compare performance growth with price and heat.

You got stuck on the whole "30% more clock but only 12% more performance" when you should be focusing on the 12% more performance at same cost and power part. Of course this is rumor and we need to wait for official bench's, I trust Toms will do a thorough review of it. Your attempt at insinuating that it was a failure, even before it launched, due to it's clock speed increase is another example of flawed reasoning. We all know BD design's are clock speed favorable similar to the P4. The P4 never reached planned clock speeds and thus Intel scrapped the idea. It seems AMD has found a way to reach those clock speeds, or rather Cyclos has.

Maybe trinity / PD will suck, maybe it won't, I don't know and neither does anyone else here. We know the Llano Sabine APU's were a success, no matter how hard some people try to discredit it. AMD gained market share from those. If they can maintain the same performance and drop price / power, then they'll be even more of a success. If they can increase performance while dropping price / power then it'll be a very large success. Those are awfully big "IF's" and I don't bank on maybes.
 
Upvote 0 Downvote
A TDP only says so many things about the APU itself, as does the arch on paper. BD looked great on paper but in reality it falls very short of it, as did K10 or NetBurst.
Click to expand...

TDP says exactly what TDP is, nothing more, nothing less. It's the amount of heat that is required to be removed for sustained operations. If we know the performance characteristics of a uArch then we can use TDP to predict it's upper limit. That is why the TDP number is important, it shows a BD uArch running at higher clock speeds and having a GPU attached while still being under their previous TDP. Immediately this means here has been improvements in leakage and power consumption.

Not once have I argued performance of said APU, only the truthfulness of AMD's posted specs. This is something you've tried to murk up several times. Before with your "they liez with ACP" statements, then with your "so I was wrong, they don't liez with ACP but they might" and now with your "TDP is just a number, it tells us nothing".

I have an issue with you claiming to be non-biased yet constantly posting biased comments and trying to further bias other posters. It was bad enough as a CR, but now with mod rights it's damaging to the credibility of the site. You can't have your mods posting like that and not turn into another Anandtech. You deliberately phrase your statements to insinuate your bias. Example is your positive statements in the IB thread yet your negative statements in this thread, about the same thing which is the reduction in power. Intel reduces power in IB and it's a great sunny day, AMD reduces power in BD/PD and they must be deceiving us. It's the very definition of bias.

Trinity is BD-e, Llano was a "Phenom II enhanced", radically different IPC, you can't compare clock speeds vs performance growth between them. Instead you compare performance growth with price and heat.

You got stuck on the whole "30% more clock but only 12% more performance" when you should be focusing on the 12% more performance at same cost and power part. Of course this is rumor and we need to wait for official bench's, I trust Toms will do a thorough review of it. Your attempt at insinuating that it was a failure, even before it launched, due to it's clock speed increase is another example of flawed reasoning. We all know BD design's are clock speed favorable similar to the P4. The P4 never reached planned clock speeds and thus Intel scrapped the idea. It seems AMD has found a way to reach those clock speeds, or rather Cyclos has.

Maybe trinity / PD will suck, maybe it won't, I don't know and neither does anyone else here. We know the Llano Sabine APU's were a success, no matter how hard some people try to discredit it. AMD gained market share from those. If they can maintain the same performance and drop price / power, then they'll be even more of a success. If they can increase performance while dropping price / power then it'll be a very large success. Those are awfully big "IF's" and I don't bank on maybes.
Click to expand...

A TDP cannot make up for a bad process. That is all I was saying.

And in my previous post, I said nothing about TDP you stated I did when I did not, I said the idle clock speed does not state anything about load power draw which is one of the major falling points of BD.

And i feel that clock speed is still important. One of the biggest reasons why the 2500K is the best CPU right now is because it takes less power and less clock speed to do the same job as a higher clocked and higher power draw BD CPU.

I still don't get why people will only focus on one part of a CPU. When it comes down to it, every part is still important, IPC clock speed, TDP everything. Focusing on the parts that make the CPU seem better is not the best way to do it. If it takes a 3.8GHz Trinity CPU to outperform a 2.9GHz Llano, and not by that much, that means IPC has not improved. If it uses less power at a higher clock, thats a benefit. But there will always be ups and downs to a CPU and so long as the downs don't outweigh the ups its a good CPU. BD had more downs than ups, therfore it is not priced higher or the prefered CPU currently.

If IB has a higher power draw, I would state that as a downfall as well since it takes more power to do the same job. Or if it took a higher clock speed as well.

But before, as I said don't put words in my mouth.
 
Upvote 0 Downvote
Here let me demonstrate how raw clock speed is irrelevant when competing two different uArch

And i feel that clock speed is still important. One of the biggest reasons why the 2500K is the best CPU right now is because it takes less power do the same job as higher power draw BD CPU.
Click to expand...

Your statement says the exact same thing with clock speed removed. That is because clock speed is a factor of performance, thus any comparison of performance has it built in by default. Clock speed is only relevant when comparing CPU's within the same uArch to each other.

Your focusing on the differences to spread misinformation again. Trinity is BD, Llano is Phenom, might as well rehash this exact same comparison with desktop BD vs Phenom II. Except now AMD has found (bought) a way to reduce power consumption such that the higher clocked BD / Trinity will outperform the lower clocked Phenom uArch. And you don't like that. Otherwise you'd be speaking positively. Of course what's really going on is that you hate BD and everything related to it, thus BD can never do anything good.

The only things that matter is cost and power usage when determining relative performance. (Assuming compatible uArch / ISA / customer warrantee and so forth).

CPU A can accomplish 10,000 wingle operations (made up unit of measurement) per second at 100W TDP, $100 USD and 3.0Ghz
CPU B can accomplish 10,000 wingle operations at 100W TDP, $100 USD and 10.0Thz (yes a T).

They cost the same, use the same energy and produce the same heat. They are functionally the same.

Your entire argument is classic example of a red herring.

http://en.wikipedia.org/wiki/Ignoratio_elenchi

"red herring" is a debating tactic that seeks to divert an opponent. A digression can, similarly, be a verbal tactic of diversion, but has no place in a serious debate; and the diversion of digression may also be in play.
Click to expand...

Your absolute focus on "Clock Speed" / IPC and abandonment of power requirements / cost is an example of you attempting to divert attention away from the true argument (performance of Trinity vs Llano / SB / IB) and towards the irreverent argument (Trinity has 30% higher clock speed!!, it must suck!!).

It could have 3,000% higher clock speed, or even 3,000,000% higher (laws of physics not withstanding) clock speed, wouldn't make a difference. What matters is cost and heat production (just another factor of cost) vs performance obtained, or the number of "wingle operations" you get for that cost / heat.

Please, refute the logic.

At this point in time your down to personal preference, you don't prefer BD or it's derivatives because you don't like BD. Just like you only like ATI GPU's when their not made by AMD. There is nothing wrong with personal preference, just don't try to pass a subjective measurement off as an objective measurement then defend it.
 
Upvote 0 Downvote
Here let me demonstrate how raw clock speed is irrelevant when competing two different uArch

And i feel that clock speed is still important. One of the biggest reasons why the 2500K is the best CPU right now is because it takes less power do the same job as higher power draw BD CPU.
Click to expand...

Your statement says the exact same thing with clock speed removed. That is because clock speed is a factor of performance, thus any comparison of performance has it built in by default. Clock speed is only relevant when comparing CPU's within the same uArch to each other.

Your focusing on the differences to spread misinformation again. Trinity is BD, Llano is Phenom, might as well rehash this exact same comparison with desktop BD vs Phenom II. Except now AMD has found (bought) a way to reduce power consumption such that the higher clocked BD / Trinity will outperform the lower clocked Phenom uArch. And you don't like that. Otherwise you'd be speaking positively. Of course what's really going on is that you hate BD and everything related to it, thus BD can never do anything good.

The only things that matter is cost and power usage when determining relative performance. (Assuming compatible uArch / ISA / customer warrantee and so forth).

CPU A can accomplish 10,000 wingle operations (made up unit of measurement) per second at 100W TDP, $100 USD and 3.0Ghz
CPU B can accomplish 10,000 wingle operations at 100W TDP, $100 USD and 10.0Thz (yes a T).

They cost the same, use the same energy and produce the same heat. They are functionally the same.

Your entire argument is classic example of a red herring.

http://en.wikipedia.org/wiki/Ignoratio_elenchi

"red herring" is a debating tactic that seeks to divert an opponent. A digression can, similarly, be a verbal tactic of diversion, but has no place in a serious debate; and the diversion of digression may also be in play.
Click to expand...

Your absolute focus on "Clock Speed" / IPC and abandonment of power requirements / cost is an example of you attempting to divert attention away from the true argument (performance of Trinity vs Llano / SB / IB) and towards the irreverent argument (Trinity has 30% higher clock speed!!, it must suck!!).

It could have 3,000% higher clock speed, or even 3,000,000% higher (laws of physics not withstanding) clock speed, wouldn't make a difference. What matters is cost and heat production (just another factor of cost) vs performance obtained, or the number of "wingle operations" you get for that cost / heat.

Please, refute the logic.

At this point in time your down to personal preference, you don't prefer BD or it's derivatives because you don't like BD. Just like you only like ATI GPU's when their not made by AMD. There is nothing wrong with personal preference, just don't try to pass a subjective measurement off as an objective measurement then defend it.
Click to expand...

Thanks for putting words in my mouth again. Love that I have said I "hate BD".

Yep.
 
Upvote 0 Downvote
Thanks for putting words in my mouth again. Love that I have said I "hate BD".

Yep.
Click to expand...

What else is it then? You've been pretty adamant on being negative about anything dealing with BD or it's derivatives.

AMD has reduced temperatures and increased clock speeds, resulting in better performance. That is a good thing, yet you've not said anything positive in relation to it. You've even tried to spin it negatively.

Caz

Trinity has really cut the power draw of the CPU. That was one of the biggest drawbacks of Bulldozer. Granted most of the CPU performance increase has come from the increased clock speed.

Good to see and hope hey can keep improving on that.
Click to expand...

You

So far as they say. We haven't seen that in action yet though and thats what we are waiting for. No sites probably have it yet to test with either.
Click to expand...

AMD's info about processor lineup

http://www.fudzilla.com/home/item/25919-amd-trinity-lineup-detailed

The full lineup starts with the flagship A10-5800K Black Edition quad-core APU that features HD 7660D graphics, has 4MB of L2 cache, and works at 3.8GHz base and 4.2GHz boost CPU clock. The GPU part of this APU, named HD 7660D features 384 GCN stream processors (or Radeon Cores 2.0, as AMD calls them now) and works at 800MHz
Click to expand...

That's 100W TDP,

Then
A10-5700 quad-core APU that features the same GPU but has slightly lower clocks for both CPU, 3.4GHz/4.0GHz, and GPU part, set at 760MHz. This one, on the other hand, also has a lower 65W TDP
Click to expand...

http://www.donanimhaber.com/islemci/haberleri/DH-Ozel-AMDnin-Trinity-kod-adli-yeni-nesil-Fusion-islemcileri.htm

Trinity's reduced their power draw, at the very least from the technology AMD purchased / licensed from Cyclos.

http://www.extremetech.com/computing/119507-amd-to-use-resonant-clock-mesh-to-push-trinity-above-4ghz

The amount they've cut is above what Cyclos claims to be capable of, so it then follows that they've also reduced leakage and tuned the die a bit. This isn't PD, but it's an indication of what PD ~might~ be capable of, a the very least.

Some places saying it's GCN, others VLIW4. At 32nm it would be kinda hard to do GCN, so I'm gonna bank on a VLIW4 HD6x GPU.
 
Upvote 0 Downvote
Anyone else notice the Trinity stepping was A1? Seems low for a final release. Probably old engineering samples.

This S|A article mentioned A1 silicon back in November 4. I doubt they're still at A1 silicon 8+ months later.

http://semiaccurate.com/2011/11/04/getting-ready-for-trinity-yet/

 
Upvote 0 Downvote
and what's even more disappointing is that we were promised a stepping revision for the FX-Bulldozer.
I thought it was to be the FX-4170, I was wrong.
now word of three or four more FX chips coming out and still no stepping revision.
so basically it's the same chip @ different speeds.
same crap.

When is Piledriver coming again.? :??:
Click to expand...
afaik, there is technically nothing wrong with zambezi. the sb-e cpus (stated only for reference, not comparison) got a stepping revision because of the vt-d bug. nothing like that was inside zambezi.
but i think amd has started releasing the 'real' zambezi lineup with the 125w tdp cpus.
i agree with you though. i also thought zambezi would get a revision. i guess, with trinity (summer 2012) and pd (october-ish, with win 8 release?) on the way and amd's new roadmap, zambezi might not be getting a revision.
 
Upvote 0 Downvote
Pallidin, if you want to bank on what AMD is saying, go for it. I stated there are no real review samples ready yet. We have no idea what the power draw will be for Trinity.

I would ratehr wait for an actual review than take slides from AMD, or slides from any company for that matter, because they will always give the cherry picked numbers, not the real numbers.

As for clock speed, as I said I think its still important to see if the arch is actually improving or if they are just going around it. You can only improve clock speed so much before hitting a wall.

and what's even more disappointing is that we were promised a stepping revision for the FX-Bulldozer.
I thought it was to be the FX-4170, I was wrong.
now word of three or four more FX chips coming out and still no stepping revision.
so basically it's the same chip @ different speeds.
same crap.

When is Piledriver coming again.? :??:
Click to expand...

I wonder if they may skip the revision for BD and go straight to PD.

As for when, either Q2, Q3 or Q4 depending on the source.
 
Upvote 0 Downvote
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom