AMD Zambezi news, info, fans !

[crazydiamond]

I see it as a 4 core part that can execute up to 8 threads, An engineer at AMD while I was there tended to agree with me, but what does he know.........

We obviously have members here that know more about BD than AMD.

not bolded part is stating what was said in a factual discussion (even though there could be another engineer out there who sees it as an 8 core part, again it makes no difference!) bolded part i saw/see as smart ass comment/reply

saying a smart ass reply =/= calling you an ass

srsly done now, thx dirtyd for actually reading my posts and understanding them!

best wishes to all my amd brethren!

[chew*]

I merely stated multiple facts one being from a conversation, one from observation in this thread, read the thread, we have many self proclaimed experts and analysts.

You have guys here analyzing the chip and spouting out performance numbers all willy nilly as if they designed the chip, Duron said it best.

I merely offered some insight on how it works, I obviously can not go into detail.

As far as your comment crazy, you have been combative with me for a while now and obviously have personal issues with me for whatever reason.

I almost did not reply to your PM becasue of this but still gave you what I would consider "sound" advice.......if at least to save you money in the long run becasue that is what I do, help people.

This is how i'm repaid. <---now i'm bing a smartass.

I've done alot for all of you whether it be behind the scenes or publically. What do I get in return?

On that note you guys are probably right I don't know a (f)lucking thing.

Attachment 119392

[Daveburt714]

Wow.... It's sad to see good/contributing members of XS upset with each other.

CrazyD: Chew* can be blunt and honest alot of times, don't take it as an insult. I'm sure he didn't intend it that way...

Chew*: I don't see where he called you an ass, he was just offended by your by your tone. You can be a little impatient sometimes.

I'm sure I shouldn't even get involved in the quip, but I like both you guys....

Let's take a deep breath OK?

[PerryR]

Why do you think Zambezi scores so low in cinebench?

Huh? All the Zambezi scores I've seen have been all over the place. Plus, JF-AMD, recently, pretty much said all the leaks are inaccurate.

[ilkkahy]

I think arguing whether 4-module BD is 4 or 8 core cpu is quite pointless semantics. Its bit like two colourblind persons would argue over some colour. Amd obiously calls it 8-core cpu because it sounds more powerful. In the end its the performance/price that counts.

[chew*]

I think arguing whether 4-module BD is 4 or 8 core cpu is quite pointless semantics. Its bit like two colourblind persons would argue over some colour. Amd obiously calls it 8-core cpu because it sounds more powerful. In the end its the performance/price that counts.

There is no argument over whether it is a 4 or 8 core part......AMD calls it an 8 core part, it is what it is.

Whether it performs/acts like an 8 core or 4 core 8 thread part is the matter at hand.......

Why does that matter or should it matter?

I leave you with this. Almost same speeds.

http://www.hwbot.org/submission/2194...t_be_4sec_78ms 6 threads ( 6 real cores )

http://www.hwbot.org/submission/2181...600k_4sec_77ms 4 core 8 thread (4 real cores )

[freeloader]

Too bad you had to sign a NDA Chew. But we all know you need to, to play with the latest hardware.

[Dumo]

Why does that matter or should it matter?

I leave you with this. Almost same speeds.

If the "new" chip averaging @ 53X then its obvious.....gotta get the other "new" chip that can go 6Ghz

Dry dewar and venom...for almost a year. Lol

[Solus Corvus]

Chew is right, the performance in the applications you use is what matters. The number of cores doesn't tell you much without knowing what the rest of the picture looks like. It's no more useful than knowing GHz without any additional information. Don't get hung up on marketing terminology since the technical definition of cores has been growing increasingly blurred since the beginning of the multicore era.

[nex_73]

Ok, i stick my chin out with this prediction, feel free to smash it once real numbers are out.
Zambezi FX (BD) thread VS Phenom II (Thuban) thread performance: (not including support for new instruction sets)
Attachment 119446

[drfedja]

Ok, i stick my chin out with this prediction, feel free to smash it once real numbers are out.
Zambezi FX (BD) thread VS Phenom II (Thuban) thread performance: (not including support for new instruction sets)

Add to that calculation BD clock advantage. My prediction is that the per module IPC could be 15-25% better than Thuban depends of sotware. Eight BD threads can be faster from 35% to 80% than Thuban six threads.

[Manicdan]

Ok, i stick my chin out with this prediction, feel free to smash it once real numbers are out.
Zambezi FX (BD) thread VS Phenom II (Thuban) thread performance: (not including support for new instruction sets)

Heres what i got:
the IPC scaling varies alot based on if threads are using a module by them selves (1.25x IPC compared to thuban) or if they are the second thread on the module for a .95x IPC. which averages out to 10% better IPC (which can be seen at 4 and 8 threads). and btw these are after we compare differences in clock rates and may be a little on the high end.

when it comes to 1-2 threads BD will be very fast since it can run each thread on different modules, and still use full turbo
3 threads seems pretty low since its not sharing resources, and 4 threads is fast again because thuban loses its turbo
5-6 have increased IPC vs 4 threads because now that it cannot use the faster turbo its spreading threads across all modules first
between 5-8 threads i had it use less and less turbo since its tough to tell how often it will really use it.

i think the most interesting part is how smooth the curve is, because of different turbos, and different utilization, its almost a perfect arc

*this was all done out of boredom while my steam game finished downloading
Attachment 119411

[nex_73]

Add to that calculation BD clock advantage. My prediction is that the per module IPC could be 15-25% better than Thuban depends of sotware. Eight BD threads can be faster from 35% to 80% than Thuban six threads.

Yupp, this is calculated @ same core clockspeed (turbo off). Average IPC about 18%...(This is also about the same price per core increase %)
@Manicdan: Interesting numbers

[drfedja]

Heres what i got:
the IPC scaling varies alot based on if threads are using a module by them selves (1.25x IPC compared to thuban) or if they are the second thread on the module for a .95x IPC. which averages out to 10% better IPC (which can be seen at 4 and 8 threads). and btw these are after we compare differences in clock rates and may be a little on the high end.

I agree! I think that the average IPC per module will be at average 15-25% better than Thuban. That is fairly enough to compete with Sand B if clock speeds will be another 10-15% higher.

when it comes to 1-2 threads BD will be very fast since it can run each thread on different modules, and still use full turbo
3 threads seems pretty low since its not sharing resources, and 4 threads is fast again because thuban loses its turbo
5-6 have increased IPC vs 4 threads because now that it cannot use the faster turbo its spreading threads across all modules first
between 5-8 threads i had it use less and less turbo since its tough to tell how often it will really use it.

i think the most interesting part is how smooth the curve is, because of different turbos, and different utilization, its almost a perfect arc

*this was all done out of boredom while my steam game finished downloading
Attachment 119411

I have told that to people many times, BD will be very strong up to 4 threads. Clock rate and turbo will depend on type of workload. The question is how to determine what core to use for two different threads. If you have only two active threads, that two threads can be handled by two cores in one module, or two different modules. With 2 threads on two modules thats run faster, but what about turbo? If you have two threads on single module, that module will be clock gated at faster clocks, than two modules.
With turbo off, probably with up to 4 threads, per clock BD will be faster up to 30%, with turbo core 2 that could be 45% of difference, but with 6 threads will be something about 10-15% faster per clock and another 10% with turbo core and with 8 threads will be much faster per clock, something about 35-40%. Of course, IPC with 8 threads will be only 5-10% faster than Thuban, but with more cores and higher clock, who cares about IPC?

IPC increase of 20% combined with high frequency design could be very strong point for future BD based products. Also, multicore design with relativly small area for execution resources is also strong point for BD. That is very important for server products. Because of that, Opterons can have low TDP, high clock and very strong performance.

IBM 48 core blade server
Opteron 6180SE 2.5 GHz 48core
specint rate: 837
specfp rate: 661

Unisys ES7000 E7-8870 2.4 GHz TB 2.8 GHz 40 core
SpecInt rate: 1080
SpecFP rate: 1180

AMD says that Interlagos will have 35% better performance than MC. Does it means that SpecInt Rate could be something about 1130? FP rate could be also higher something arround 1000.

[FlanK3r]

Im curious about single thread performance (so, active in one module)...Hope, 4200 MHz cant be slower than 3700 MHz PII core. What do u think guys?

[Formula350]

I wonder where did you read the '80-90% of the efficiency of a true dual core' statement, if it's true the Zambezi 8 core will only act like 8 * 80% = 6.4 cores in worst scenario?

AMD's Bulldozer Blog (centralized around Server, though). The exact entry I had linked to about 1/3 or 1/2 way through the thread, but for a different reason. I'd have to read it again to make sure this is accurate (which I've not the time to do atm), but I believe it said something on the order that if each module were working on one thread, it would yield quite a bit more performance than a single Magny-Cours core (heh); however, a single module would only be about 90% of the performance of two of Magny's cores.

Since I'm 98% sure what I read was by JF (John) anyways, this is just as valid as whatever blog post I'm referring to (though I think it might be one of the 20 Questions posts found here):
http://www.xtremesystems.org/forums/...1&postcount=67

[informal]

however, a single module would only be about 90% of the performance of two of Magny's cores.

Where did you read that exactly?

[Apokalipse]

however, a single module would only be about 90% of the performance of two of Magny's cores.

No.

If both cores in each module are being used, AMD says it would get 80% higher performance (i.e 180% of single core performance).
That percentage had nothing to do with K10.5 - it's looking solely at one BD module, with single thread performance vs two thread performance.

If you ran each thread in separate modules, you'd get better scaling than that though.

[BeepBeep2]

Im curious about single thread performance (so, active in one module)...Hope, 4200 MHz cant be slower than 3700 MHz PII core. What do u think guys?

Dunno how many times John has said this but IPC increases.

[jimbo75]

Dunno how many times John has said this but IPC increases.

It would be nice to hear it once more, recently.

[FlanK3r]

beep: of course, but what is exactly IPC? Maybe here is not talking about core to core, but about cores to cores (6 vs 8 etc)

[Formula350]

Where did you read that exactly?

What I wrote might not be verbatim, hence my disclaimer that it might not be accurate and that I am going off memory lol Regardless of if I was close, or wayyyy off, I read pretty much all of the Bulldozer related blog entries at AMD. All three of the 20-Questions entries are a good read and where I think John may have said what I was going off of, but I was on Vacation for all of July (without internet) so my brain has been making room for new info while playing catch-up haha If I get time after editing reviews I'll try to dig for the specific blog link.

[drfedja]

however, a single module would only be about 90% of the performance of two of Magny's cores.

JF doesn't said that. He said that the one module has 1.8x of performance scaling with two threads on it. Single module with two threads would be 90% of the performance of two BD modules with two threads. In other words CMT - clustered multithreaded module has 90% of chip multiprocessed two cores with same type of microarchitecture. There isn't comparison with Magny Cours or 10h.

Since I'm 98% sure what I read was by JF (John) anyways, this is just as valid as whatever blog post I'm referring to (though I think it might be one of the 20 Questions posts found here):
http://www.xtremesystems.org/forums/...1&postcount=67

He talks about mythical 6-core bulldozer:

Mythical 6-core bulldozer:
100% + 95% + 95% + 95% + 95% + 95% = 575%

Orochi die with 4 modules:
180% + 180% + 180% + 180% = 720%

What if we had just done a 4 core and added HT (keeping in the same die space):
100% + 95% +95% +95% + 18% + 18% + 18% + 18% = 457%

What about a 6 core with HT (has to assume more die space):
100% + 95% +95% +95% +95% +95% + 18% + 18% + 18% + 18% + 18% + 18% = 683%

And someone concluded that the IPC decreases. :d
And, from that moment, there is so much people who think that BD has decreased IPC from 10h.

beep: of course, but what is exactly IPC?

IPC = Instructions Per Cycle. This is a number who's tells us how many instructions CPU can retire per one cycle. Today modern Out Of Order processors have many execution units, they can execute more than one instruction per cycle, parallel, and out of order. Because program code is in order, there is quite difficult to make such machine to execute out of order, because of data dependencies, too many branches in code, etc.
IPC is software performance measuring unit. If code is properly optimised, it can run on CPU, or even GPU with more instructions per cycle. However, also if CPU is faster, than it can run same software with higher IPC number.

Maybe here is not talking about core to core, but about cores to cores (6 vs 8 etc)

If BD has 50% more throughput than 6-core 10h, at probably same clock, that means the server workload uses all cores. That means the core per core BD vs 10h, BD has 12.5% more throughput than 10h, but single module can do 25% more serialized, single thread jobs than one single core.

[undone]

Mythical 6-core bulldozer:
100% + 95% + 95% + 95% + 95% + 95% = 575%

Orochi die with 4 modules:
180% + 180% + 180% + 180% = 720%

What if we had just done a 4 core and added HT (keeping in the same die space):
100% + 95% +95% +95% + 18% + 18% + 18% + 18% = 457%

What about a 6 core with HT (has to assume more die space):
100% + 95% +95% +95% +95% +95% + 18% + 18% + 18% + 18% + 18% + 18% = 683%

Interesting, this calculation take HyperThreading for 18% increase into account, which I often think HT may bring 25-30% additional throughput.

[drfedja]

Interesting, this calculation take HyperThreading for 18% increase into account, which I often think HT may bring 25-30% additional throughput.

I don't think so that HT brings 25-30% on throughput on average.... 18% is more realistic number.