AMD - Bulldozer Chitchat

[CAT-THE-FIFTH]

the reviewers conclusion sums it up for me:


It is good that there is finally an official announcement of a patch which may improve the Win 7 Scheduler, cant see it making groundbreaking changes to anyones decision in buying a CPU though.

TBH,it would mean improvements in lightly threaded games though. This would mean that in lightly threaded games the whole range will look better when compared to the Phenom II. It also means Trinity will also end up looking better now with regards to gaming.

[Biscuit]

Well i guess that will depend on how drastic the improvements actually are, if they are significant then perhaps it might just about bring it inline in overall performance. That still doesn't tackle the insane power usage though, i dont see the current FX ever really fitting properly in line with the Phenoms.
If the above mentioned scheduler patch is as we hope, then, in conjunction with improved silicone (which is undoubtedly on the way for the next enthusiast chip and trinity) then yeah trinity should be good, especially if it there is a good boost to the GPU aswell. Llano is a hard acto to follow though, i still think it offers a near perfect blend of performance, power use and cost.

[CAT-THE-FIFTH]

Microsoft has released a Windows hotfix for BD:

http://www.techpowerup.com/156844/AM...icrosoft..html

"This article introduces an update that optimizes the performance of AMD Bulldozer CPUs that are used by Windows 7-based or Windows Server 2008 R2-based computers. Currently, the performance of AMD Bulldozer CPUs is slower than expected. This behavior occurs because the threading logic in Windows 7 and in Windows Server 2008 R2 is not optimized to use the Simultaneous Multithreading (SMT) scheduling feature. This feature was introduced in the Bulldozer family of AMD CPUs.

Note This issue may occur when you use applications that run in multiple threads."

[watercooled]

So much for the rumour saying it wasn't going to happen. I wonder who started that?

[Noxvayl]

Now we need some reviews done to see how well the hotfix works.

[Biscuit]

Microsoft has released a Windows hotfix for BD:

http://www.techpowerup.com/156844/AM...icrosoft..html

"This article introduces an update that optimizes the performance of AMD Bulldozer CPUs that are used by Windows 7-based or Windows Server 2008 R2-based computers. Currently, the performance of AMD Bulldozer CPUs is slower than expected. This behavior occurs because the threading logic in Windows 7 and in Windows Server 2008 R2 is not optimized to use the Simultaneous Multithreading (SMT) scheduling feature. This feature was introduced in the Bulldozer family of AMD CPUs.

Note This issue may occur when you use applications that run in multiple threads."

Good news indeed, very curious to see just how much difference it actually makes. Im amazed that there isnt any benchmarks yet, i would have thought people would have got straight on it!

So much for the rumour saying it wasn't going to happen. I wonder who started that?

I dont think there was a rumour saying it wouldnt happen, more that it took months for anything official from MS or AMD to actually confirm that there was a patch in the works. I genuinely didnt think it would actually come but im pleased it has.

[CAT-THE-FIFTH]

Two games are tested in this thread so far:

http://www.overclock.net/t/1185039/t...re-in-have-fun

[watercooled]

No there was actually a rumour I read, or rather someone claiming that the patch itself was a rumour.

Edit:

We're tuning things, and we'll be right back

All the best,
Overclock.net

That's about right.

[Biscuit]

Seems the patch was released by accident. What we have seen is only half of it:

http://techreport.com/discussions.x/22179

We've spoken with an industry source familiar with this situation, and it appears the release of this hotfix was either inadvertent, premature, or both. There is indeed a Bulldozer threading patch for Windows in the works, but it should come in two parts, not just one. The patch that was briefly released is only one portion of the total solution, and it may very well reduce performance if used on its own. We're hearing the full Windows update for Bulldozer performance optimization is scheduled for release in Q1 of 2012. For now, Bulldozer owners, the best thing to do is to sit tight and wait.

[watercooled]

Yeah I've just been reading about that, it seems it's causing some programs to treat the modules like cores with HT / too keen to use a new module rather than the other core.

[watercooled]

Something I've just spotted on xbitlabs, their power consumption measurements are far lower than Anandtech posted. I mean it could be down to them getting a high VID die but it's quite a difference.
http://www.xbitlabs.com/articles/cpu..._13.html#sect0
http://www.anandtech.com/show/4955/t...x8150-tested/9

[Biscuit]

Something I've just spotted on xbitlabs, their power consumption measurements are far lower than Anandtech posted. I mean it could be down to them getting a high VID die but it's quite a difference.
http://www.xbitlabs.com/articles/cpu..._13.html#sect0
http://www.anandtech.com/show/4955/t...x8150-tested/9

That's quite odd, they do seem to be the odd ones out though so i would tend to go along with everyone else on real world figures.

I'm genuinely intrigued as to where AMD go with the 8170, I have a feeling (albeit hesitant) it will end up on a much better die with serious clocking ability as a test bed ready for piledriver.

Wishfull thinking?

[watercooled]

Well the reviewed power draw does seem quite high based on transistor count/process so it must be down to yield. I'm thinking the xbit numbers aren't erroneous though, rather they got a later batch or something. If that's the case, the future CPUs look promising in terms of performance/watt. GloFo's yield problems really aren't helping AMD though.

[CAT-THE-FIFTH]

The first picture of a Trinity die:

http://semiaccurate.com/2012/01/05/e...about-trinity/

Here is a comparison with Llano:

http://forum.beyond3d.com/showpost.p...&postcount=166

[CAT-THE-FIFTH]

It seems the quad core Core i7 3820 has 1.27 billion transistors and a 294MM2 die size:

http://www.anandtech.com/show/5276/i...sandy-bridge-e

The socket 2011 six core CPUs(which are actually eight core CPUs) have 2.27 billion transistors and a 435MM2 die size. It seems the quad core socket 2011 CPUs are much less transistor dense than the eight core version. This would put transistor density about 13% to 14% higher than the FX8100 series with a much more mature 32NM process. However,power consumption at load seems much less than the FX8100 series.

Hopefully,Piledriver will benefit from a more mature 32NM process too.

[watercooled]

It looks like there is quite a lot of wasted die space on Zambezi, maybe the transistor mis-count has something to do with that? The die space on Trinity looks to be much more efficiently used; combined with a more mature process, lessons learned from Zambezi and better OS support, it should be much more efficient.