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]

We know the slight tweaks and timings made to Piledriver to alleviate choke points, the bumped clocks also help in that regard, basically 200mhz on a PD module is around 6-700 mhz on a BD module. Though the real changes are made for steamroller but one would say that AMD have abandoned modular design with SR, individual FPU's with its own front sides effectively change it from the idea behind BD, for all intents and purposes these will be individual cores on a shared module with adjusted L2 Cache.

 

[de5_Roy]

Exclusive: AMD to Announce Another 5% Employee Reduction?
http://vr-zone.com/articles/exclusive-amd-to-announce-another-5-employee-reduction-/17317.html

AMD APUs at Amazon
http://www.fudzilla.com/home/item/28930-amd-apus-at-amazon
AMD sticks 500,000 Android apps onto Windows 8
http://www.fudzilla.com/home/item/28929-amd-sticks-500000-android-apps-onto-windows-8

 

[mayankleoboy1]

AMD to partner with Oracle to accelerate Java (for servers) on their APU's.
If this doesnt help HSA, nothing will.

 

[gamerk316]

Except Java is going the way of the dodo now that its clear its a bug riddled mess. HTML5 is going to replace Java in the long run.

 

[cgner]

Nice at times Amd even had the intel chip beat but like the comments i want to see it benchmarked fully at the I3. If Trinity is priced at the I3 its going to be quite a deal.

Yea I totally agree. Why do they compare it to a higher tier chip that costs more? Typical online benchmark intel bias bs

 

[sarinaide]

My gripe about benches is that they rarely ever manifest in real life, the performance difference between intels top chip and bulldozer, faildozer, bullcrap whatever people called it is still not distinguishable, if it were 55% faster it would be noticeable even at the basic level of booting up, responsiveness.

I don't advocate for this or that choice, I do believe each chip has a market and user that it suits, I also don't deny that AMD with BD apart from being late is basically the last relic of the old regime, a marked reminder that bad management and pocket lining left loyal fans with a product barely better than the ancient architecture it replaced, its a solemn reminder to the new guard that such disappointments will not be tolerated again.

 

[mayankleoboy1]

Except Java is going the way of the dodo now that its clear its a bug riddled mess. HTML5 is going to replace Java in the long run.

not for servers, AFAIK.i dont see Java dying in 3-4 years. Thats all the time AMD needs to make HSA big.

 

[mayankleoboy1]

My gripe about benches is that they rarely ever manifest in real life, the performance difference between intels top chip and bulldozer, faildozer, bullcrap whatever people called it is still not distinguishable, if it were 55% faster it would be noticeable even at the basic level of booting up, responsiveness.

true for synthetics definitely.
But for real benchmarks, like x.364, R11.5, CS6 and other productivity softwares, the results of benchmarks are real. If they show a CPU as 20% slower, then it is 20% slower when you will do work in that SW.

 

[sarinaide]

My gripe about benches is that they rarely ever manifest in real life, the performance difference between intels top chip and bulldozer, faildozer, bullcrap whatever people called it is still not distinguishable, if it were 55% faster it would be noticeable even at the basic level of booting up, responsiveness.

true for synthetics definitely.
But for real benchmarks, like x.364, R11.5, CS6 and other productivity softwares, the results of benchmarks are real. If they show a CPU as 20% slower, then it is 20% slower when you will do work in that SW.

That is true, it depends on the users orientation, AMD still very much can offer a set offering to a end user and in instances it can be a good choice, but obviously I would like AMD to make a definite go to chip again for all circumstances.

 

[guskline]

My gripe about benches is that they rarely ever manifest in real life, the performance difference between intels top chip and bulldozer, faildozer, bullcrap whatever people called it is still not distinguishable, if it were 55% faster it would be noticeable even at the basic level of booting up, responsiveness.

I don't advocate for this or that choice, I do believe each chip has a market and user that it suits, I also don't deny that AMD with BD apart from being late is basically the last relic of the old regime, a marked reminder that bad management and pocket lining left loyal fans with a product barely better than the ancient architecture it replaced, its a solemn reminder to the new guard that such disappointments will not be tolerated again.

Excellent post. My 4.5Ghz Buldozer 8150 grows on me day by day. sarinaide, I take it from your posts you have had access to the PileDriver?

Again, being one who has 3 main rigs(see member configuration), 2 of which are i5 2500ks @4.5 Ghz and one of which is a 8150 @ 4.5 Ghz the difference is measurable but in most gameplay not noticeable. The 8150 is incredible smooth.

 

[sarinaide]

Access yes, mine no...the other thing to bare in mind is that these are engineering samples which are within 10% margin of error.

 

[gamerk316]

^^ But the point, as you noted, was if the 2500k is faster, why bother with BD since both cost about the same?

 

[sarinaide]

The BD is marginally cheaper and the platform is cheaper, the performance is about within margin of error of a i5 anyways to, there are instances where a i5 is a better option, notably MATX and ITX builds.

 

[guskline]

Between the time of the ES release and the commercial release, is there sufficient time to make hardware tweaks in the actual cpu, or is the feedback from the ES used to make tweaks for the mb BIOS only ? i.e. is it too late to change the actual CPU until the next stepping?

 

[sarinaide]

It depends on what level tester one is, some get chips throughout the process to beta test and evolve the process. Others get chips prior to release or because they are in good regards. Most ES chips though are older samples.

 

[esrever]

I just really don't get software like PCmark, sysmark and passmark which all give really bad estimations of performance but are ran by tech sites. Most of these tests seem extremely intel biased. Even worse are tests like superpi, using outdated instructions and don't even scale with cores. Bulldozer added a lot of instructions and extensions, it should be much better than phenom when newer softwares come and make use of them.

 

[esrever]

not for servers, AFAIK.i dont see Java dying in 3-4 years. Thats all the time AMD needs to make HSA big.

http://www.techpowerup.com/173011/AMD-and-Oracle-to-Explore-Heterogeneous-Computing-for-Java.html

AMD has been doing extensions of HSA for many environments.

 

[noob2222]

My gripe about benches is that they rarely ever manifest in real life, the performance difference between intels top chip and bulldozer, faildozer, bullcrap whatever people called it is still not distinguishable, if it were 55% faster it would be noticeable even at the basic level of booting up, responsiveness.

true for synthetics definitely.
But for real benchmarks, like x.364, R11.5, CS6 and other productivity softwares, the results of benchmarks are real. If they show a CPU as 20% slower, then it is 20% slower when you will do work in that SW.

Cinebench isn't a "real" benchmark. its somehow calculating a score, but its not based on completion times. I can't find the link now, but someone tested recently, BD finished before the phenom X6 at an equal clock speed, but scored lower. How do you get a lower score by finishing first?

btw, the link is somewhere in this thread, probably somewhere between page 40 and 80, if i get bored, i may try to find it again.

 

[Dangi]

Well, now there is no need for benchies (why bother with ES releases) until PD release at 23 october where all flaws are supposed to be fixed

 

[jdwii]

My gripe about benches is that they rarely ever manifest in real life, the performance difference between intels top chip and bulldozer, faildozer, bullcrap whatever people called it is still not distinguishable, if it were 55% faster it would be noticeable even at the basic level of booting up, responsiveness.

I don't advocate for this or that choice, I do believe each chip has a market and user that it suits, I also don't deny that AMD with BD apart from being late is basically the last relic of the old regime, a marked reminder that bad management and pocket lining left loyal fans with a product barely better than the ancient architecture it replaced, its a solemn reminder to the new guard that such disappointments will not be tolerated again.

But when a True program shows it has more performance it is faster on lots of games Amd is behind by 20-60FPS i saw Benchmarks that shows Amd to get around 50FPS on average vs 70-80FPS on Intel and this is at comparable prices.

 

[jdwii]

Well, now there is no need for benchies (why bother with ES releases) until PD release at 23 october where all flaws are supposed to be fixed

Piledriver is just a 1.1 bulldozer its better but nothing to special people are saying Steamroller is more a 2.0.

 

[obsama1]

The A10 did pretty well compared to the i5. It doesn't even have L3. The 4350 is going to be a pretty good budget CPU.

 

[esrever]

But when a True program shows it has more performance it is faster on lots of games Amd is behind by 20-60FPS i saw Benchmarks that shows Amd to get around 50FPS on average vs 70-80FPS on Intel and this is at comparable prices.

A lot of programs are specialized and optimized that way. Games are generally some of the most optimized software everyday people run. Im not sure where AMD is dropping FPS in those games but there are probably specialized code for intel that aren't implemented for AMD systems in there.

 

[Dangi]

Piledriver is just a 1.1 bulldozer its better but nothing to special people are saying Steamroller is more a 2.0.

Well yes, PD is supposed to have minor tweaks here and there and SR being the huge improvement, but BD was released over a year ago and software is slowly ( ver slowly) catching up, not too long ago Tom's made a review http://www.tomshardware.com/reviews/amd-llano-opencl,3284.html and well while I don't think BD will be that huge improvement over a year, it did improve a bit with different patches and newer soft

Also if I'm not mistaken SR will be 22nm ( or 28 ? ) that also will improve the chip, not a huge improvement but ~5% is quite easy to achieve with new fab proces

 

[esrever]

SR will be 28nm. Piledriver is bulldozer 1.1 but they did fix a lot of the small issues. The performance gain is significant since the many problems with bulldozer.

Also Piledriver gets the RCM and improved power gating so its power consumption should be much better. From the previews of trinity, the idle power consumption is amazing, probably a much more useful metric in desktop than loaded power consumption imo.