AMD - Bulldozer Chitchat

[Biscuit]

It is?

http://forums.hexus.net/pc-hardware/...over-soon.html

[CAT-THE-FIFTH]

This chap raises an interesting point about AMD moving to a bulk process:

http://semiaccurate.com/forums/showthread.php?t=6242

[DanceswithUnix]

Not even sure its running at 1333, il have to check when im back home. I did notice the timings it was running were quite low though, like cas6.
It was like £5 more and i figured it worth it if RAM Is set to get way more expensive in the coming months, if not me for a friend or family.

Well the E450 is supposed to be capable of 1333 at least.

My E350 has 1333 ram in it, but only runs up to 1066. Again, wasn't any cost difference at the time and it might get used somewhere else in the years ahead that *can* make use of the speed.

[Biscuit]

Well the E450 is supposed to be capable of 1333 at least.

My E350 has 1333 ram in it, but only runs up to 1066. Again, wasn't any cost difference at the time and it might get used somewhere else in the years ahead that *can* make use of the speed.

Exactly, im not sweating over the ~4 quid extra it was, not harming anything.

I also have a Samsung 830 SSD to stick in there... stupid lenovo and their 7mm slots. Looks sexy though!

[badass]

Interesting. It would appear that the Trinity performance testing by AMD labs used DDR3-1600, whereas the Llano testing used DDR3-1333. So the 56% graphics performance improvement will in part be down to the increased memory bandwidth available ... sneaky AMD...

Worst case scenario for AMD, that's a 27% improvement at 1333 MHz. Still very good considering they are on the same 32nm process and Power envelope.

[scaryjim]

Worst case scenario for AMD, that's a 27% improvement at 1333 MHz. Still very good considering they are on the same 32nm process and Power envelope.

Oh, I agree completely. I just think it's snaeky of AMD to use non-comparable platforms. Also make me wonder if Trinity will benefit even more from faster RAM - 2GHz+ DDR3 is still pretty reasonably priced, AFAIK... could be possible to get even more performance out of it...

[badass]

That'd be a suitable conclusion if that review actually had any price comparable processors in it It has neither an E5-2630 nor an Opteron 6274, which is the midrange server point where the two processor ranges coincide, price-wise. They've attempted to produce those CPUs by disabling and downclocking higher-spec CPUs, but that's never going to give an exact representation.

And looking at the linked chart, the Opterons are hardly incredibly slow - they turn in 93.5% of the (mocked up) roughly price equivalent Xeon's throughput performance.

Maybe not incredibly slow then. Still not great for virtualisation since it's slower than the price equivalent processor from Intel.

(I note the response times detailed in the power testing section of the review are far shorter - and far closer to each other - than those in the benchmarking section, indicating that the benchmarking section does *not* represent a typical usage pattern). It seems to me you're being a little unfair in your assessment...

http://www.anandtech.com/show/5553/t...-for-servers/7

That shows a more realistic usage pattern. The response times for the 6274 are way longer than the Xeon Equivalent. Power consumption of the platform is a lot higher as well.

On the page where the response times are similar, the price equivalent Xeon is putting more throughput out as well as having better response times.

So, higher power consumption, lower max throughput and much worse response times than the price equivalent Xeon.
I stand by my assertion. Bulldozer is rubbish for virtualisation.
It's a shame really as pre Nehalem, the Opterons utterly destroyed the Equivalent Intel Processors at virtualisation. Barcelona was way ahead of its time virtualisation wise. Unfortunately for AMD, that was also it's Achilles heel. One of the things that helped its virtualisation performance was it's large, complex TLB. Unfortunately, the TLB bug caused them problems. An unfortunate situation that could have just as easily happened to Intel (and has done in the past - FDIV bug, anyone?)

In fact, Bulldozer is pretty useless for most things. On the desktop, the cheaper to buy, much cheaper to manufacture Intel 2500k is a huge percentage faster in all but the most heavily threaded workloads, and where the FX-8150 does win, it only just wins. More than 4 cores is generally a waste of time on desktop workloads.
In the server side, where performance can scale almost linearly with cores, the Bulldozer cores are so much slower than the Intel ones that Intel CPU's with less than half as many cores are as fast or faster.
So if you're an extremely niche desktop user that has a workload that can properly saturate more than 6* threads, bulldozer is for you. Any other user and it's too expensive for it's performance and power consumption.

* I chose 6 threads because in any workload with 6 or less saturated threads, the 2500k will probably still be faster due to its much faster cores.

It may seem that i'm being inconsistent in that I think Trinity may have a chance to shine whilst I think Bulldozer is rubbish but Trinity is designed for an entirely different market and doesn't waste die space for more cores than almost any desktop user will ever use. It also looks like trinity's in for a chance of really improve lightly threaded performance - something all desktop workloads experience on a regular basis.

Add to that the rumored quicksync equivalent and you've got something vastly better.

[Biscuit]

It's a shame really as pre Nehalem, the Opterons utterly destroyed the Equivalent Intel Processors at virtualisation. Barcelona was way ahead of its time virtualisation wise. Unfortunately for AMD, that was also it's Achilles heel. One of the things that helped its virtualisation performance was it's large, complex TLB. Unfortunately, the TLB bug caused them problems. An unfortunate situation that could have just as easily happened to Intel (and has done in the past - FDIV bug, anyone?)

Could it be a case of the BD architecture being ahead of its time? Perhaps virtualisation on BD is something which which is yet to be optimised for?

[scaryjim]

*shrug* I'll happily give you that it's not as good as the Xeon. But, since you've accepted that they're not incredibly slow, to also call then "rubbish" at virtualisation seems unfair too. They're actually perfectly acceptable at virtualisation. The thing is the Xeon's are better than acceptable, at about the same price. That makes the Opterons poor value, but it doesn't make them "rubbish".

I think we've all accepted that Bulldozer failed to acheive it's target of making AMD properly competitive again, but branding it "rubbish" and "useless" seems unnecessarily harsh. An FX4100 has more than enough power to meet the computing needs of 99% of computer users. It's not necessarily the best choice, but that makes it neither "rubbish" nor "useless".

As you say, it looks like Trinity might be a decent improvement, and if it is that bodes well for the future Piledriver-based server CPUs. It's rare for a new architecture to be a massive improvement over the incumbent (think Netburst P4!) and all the signs are that AMD can address the deficits in bulldozer and make it competitive in the future. How effectively they do that only time will tell, but they have a decent track record of architecture tweaks in the recent past (i.e. Phenom I/II) and given that bulldozer already excels in a (very!) small number of workloads (something Phenom I didn't) they may yet produce something special.

[CAT-THE-FIFTH]

Intel seems very annoyed with Sea Micro:

http://www.wired.com/wiredenterprise...l-took-a-pass/

It is now basically calling them useless!! Bahahahaha! Considering that Intel rarely comments on these sorts of things it does sound out of character.

What a load of BS,Intel was very interested in working with the very same company due to its use of Intel Atom CPUs:

http://www.tmcnet.com/usubmit/2011/09/12/5768899.htm

http://blogs.wsj.com/digits/2011/02/...ew-intel-chip/

So,if they are so useless why did they ask them to participate in IDF and give them first access to certain CPUs??

[CAT-THE-FIFTH]

I just saw this:

http://www.forbes.com/sites/briancau...market-slumps/

So the worldwide server market is slumping and if anything a move to more cost effective and cheaper servers might be quite important.

The problem is Intel does not do cheap,they do high margin. Their "cheap" CPU is Atom. I wonder what Sea Micro was using!! Hmmmm??

[Sputnik]

I noticed some of the Athlon quad cores are now socket FM1 are the Athlons all going to become FM1 ?

[watercooled]

The FM1 Athlons are simply Llano with failed/disabled IGP. There is no specific Athlon die in production to my knowledge.

[Sputnik]

That makes sense. I just wondered as they were socket FM1.

[scaryjim]

Since we're on the topic of FM1 Athlons, anyone know if removing the IGP helps overclocking at all, or do they still have the problems of increasing clocks giving general system instability?

[CAT-THE-FIFTH]

Interesting news about the IB Core i3 CPUs:

http://www.donanimhaber.com/islemci/...e-ediliyor.htm

http://en.wikipedia.org/wiki/Christm...holiday_season

It seems that Trinity will be competing against the SB Core i3 CPUs for the timebeing.