Bulldozer Folding @ Home Performance Numbers

[hamunaptra]

OK, talked to my FPU guy and this is the scoop:

NICE thanks for confirming that! So those 2 other pipes arent just MMX INT, its SSE INT just as HW2050 stated! Woot. So, BD FPU could effectively be one hell of an SSE crunching monster. If the workload is a nice mix of FP SSE and INT SSE.

The 4 SSE moves / cycle is nice as well, although I dont have anything to relate that too .. to compare how significant / insignificant this is...

 

[JFAMD]

And to further clarify directly:

Also there are some features AMD downplayed so far in my opinion. It is because obviously AMD has not only 2 FPU pipes and 2 MMX pipes. Those MMX pipes don't do MMX they are full 128 Bit integer SSE pipelines
(true).

So all register moves and load/stores can be executed also in those two pipelines
(not really, reg-reg moves for SSE and AVX-128 can be done with mov-elimination

Load – doesn’t actually require an execution pipe in the FP at all – but is limited to 2 128b loads/cycle max throughput.
Store – does take an execution pipe, but can only execute down 1 of the pipes. That & LS restrictions limit it to 1 128b store/c throughput)

I recently read a source that those two don't do 64 Bit MMX but 128 Bit SSE! Really don't know why AMD was so quiet about that so far and obfuscated that by using the wrong term "MMX". Therefore AMD can do 4 * 128 Bit SSE/cycle!

(yes, “MMX” is likely a bad name to use in describing the BullDozer micro-architecture and is somewhat misleading. Yes, we can do 4 128b arithmetic operations/cycle: 2 “floating-point” and 2 “SSE/AVX-128 integer”. Or/instead/in-combination we can also do 2 x87 “floating point” and 2 mmx “integer” per cycle – and by mmx I really mean the architected “mmx&#8221.



And that is the sound of me clapping my hands like a blackjack dealer and saying "all done", can't get any further into this topic.

 

[hamunaptra]

Thanks for the clarifications JF =)

 

[Arkadrel]

it slices! It dices!.... it ll cut a pineapple in half mid-air!


Yes yes... youve convinced me already, when can I buy it?

Man the guys that thought up this cpu design, deserve some praise, thing is looking to turn out neat.

 

[hamunaptra]

Apparently June 20th is the launch date for desktop BD's

 

[PreferLinux]

In the SATA and PCI Express world, Gbit to GByte relation is 10:1.

So is USB, but not PCIe 3. I thought it was for SATA, but I wasn't sure. Thanks for clarifying.

 

[IntelUser2000]

So is USB, but not PCIe 3. I thought it was for SATA, but I wasn't sure. Thanks for clarifying.

You are right about the PCI Express 3.0. Thanks for pointing it out. The transfer rate per pin only increases to 8GT/s from 5GT/s but by reducing the impact of encoding they achieve twice the bandwidth.

 

[Idontcare]

Apparently June 20th is the launch date for desktop BD's

They should make it be June 23, just for the symbolism.

(Athlon K7 was released on the 23rd of June)

 

[SickBeast]

Apparently June 20th is the launch date for desktop BD's

Is there any word on how many cores they will have? I've heard that there will be 4, 6, and 8 core versions, but only two of those CPUs will come out in June, with the 3rd not coming out until November.

 

[Joseph F]

@SickBeast I've heard that 4 8-Core, 2 6-Core and 2 4-Core BD CPUs are coming out on June 20th.

 

[PlasmaBomb]

I see the 4k random read coming in at 0.82MB/s.
If you look at the brand new generation of sandforce drives, they beat a raptor by ~100x in some things, but overall speed (and the speed the user feels) is more like 5-10x.

The Crucial C300 for example comes in at 215.16MB/s

So 215/0.82 = ~262x

Therefore the statement of up to 100 times faster is indeed correct.

 

[PlasmaBomb]

They should make it be June 23, just for the symbolism.

(Athlon K7 was released on the 23rd of June)

Nice touch... but they are probably running with the 20th because it's the Monday.

 

[eternalone]

Any chance of getting some $99 dollar dual core bulldozers?? Because im willing to bet they destroy my phenom x4. Im in it for the architecture of BD, so I hope they do price some disabled cores that would make me very happy.

 

[JFAMD]

Is there any word on how many cores they will have? I've heard that there will be 4, 6, and 8 core versions, but only two of those CPUs will come out in June, with the 3rd not coming out until November.

I can't understand why that would happen.

 

[Edrick]

I am willing to bet a lot of people will be surprised and disappointed when BD prices turns out to be not as "value" orientated as AMDs current lineup. I do not see a 8 core BD being under $500. But that is just my opinion.

 

[Joseph F]

^I'd be surprised to see a BD for more than $500. (assuming they're good) what makes you think that they'll be so expensive?

 

[hamunaptra]

Well, I think at first BD may be expensive so AMD can make some huge profit on it. Because the die size will be pretty small for packing 8 cores. So, if their yeilds are good and performance is good...they will want a higher price for them.

As time goes one when yeilds are even better and intel releases faster processors they will lower their price accordingly.

What would be nice though is if the 4 core and 6 core variants, if they are crippled BD's..it would be nice if they were unlockable like their current chips =)

That would be the budget OC pathway I think!

 

[Edrick]

^I'd be surprised to see a BD for more than $500. (assuming they're good) what makes you think that they'll be so expensive?

If the 8 core BD is beating a 2600 ($350) and matching a 6 core 980x ($900), why would AMD ever sell it for less than $500? Ok, maybe $479, but I doubt any lower. AMD is a company and needs to make money. And if they offer the performance king, expect a kings price.

Remember, AMD used to have $1000 CPUs when they were beating Intel. I would not be surprised to see their top of the line BD being near $700.

Too many people are used to AMD being a "value" line, simply because they had no choice for some years now. Expecting a BD that beats a 2600 being priced below $200 is unrealistic.

 

[Martimus]

^I'd be surprised to see a BD for more than $500. (assuming they're good) what makes you think that they'll be so expensive?

If they are good, then I would expect them to be expensive. If they aren't good enough to justify a higher cost, then I expect them to be priced more affordably.

One thing that gets to me is that we are seeing Llano demonstrations and no Zambezi demonstrations even though Zambezi is rumored to be released first. The only reason I can see for this is that Llano is going into a new market where AMD doesn't need to worry about hurting current sales, but Zambezi is replacing current products and AMD doesnt want people to avoid buying current products to wait for Zambezi.

Or performance isn't very good, or silicon isn't really ready yet and BD will be delayed more, or maybe something else I haven't thought of.

 

[Edrick]

The only reason I can see for this is that Llano is going into a new market where AMD doesn't need to worry about hurting current sales, but Zambezi is replacing current products and AMD doesnt want people to avoid buying current products to wait for Zambezi.

Good point. And could be the case. But on the other hand, showing a strong Zambezi demo would cut into Intel's SB sales more so than current AMD sales in my opinion. Which makes me think it is either one of the other 2 possibilities you mentioned "performance isn't very good, or silicon isn't really ready yet and BD will be delayed more".

Even though I am not an AMD fan, I am very excited to see BD performance.

 

[Joseph F]

One sad thing is that even when AMD was repeatedly sodomizing intel's CPUs (First with a Sledgehammer and then with a Clawhammer!) intel never lowered its prices for their top-tier products. I truly hope that BD gains AMD a lot of respect from the enthusiast community.

 

[Idontcare]

Good point. And could be the case. But on the other hand, showing a strong Zambezi demo would cut into Intel's SB sales more so than current AMD sales in my opinion. Which makes me think it is either one of the other 2 possibilities you mentioned "performance isn't very good, or silicon isn't really ready yet and BD will be delayed more".

Even though I am not an AMD fan, I am very excited to see BD performance.

Considering that AMD is sorta paralyzed at the top-most level from a decision making standpoint (there are ramifications to having your CEO suddenly and immediately depart, and merely having an interim acting CEO take their place versus a long-term permanent hire) I really don't think we can divine anything from the tea leaves when it comes to actions taken, and those that are not, by AMD at this time.

Llano demos are probably happening because that was the plan set in motion months before Dirk vacated AMD.

It may have been too early still for any solid plans to gel regarding bulldozer demos, and once it became time to create those plans there was confusion at the top regarding the strategy for bulldozer demos.

Double whammy for servers with both Dirk and the server head-man stepping out in the same business quarter. I'm sure everyone puts on a brave face and does the daily routine hoping the situation pans out for the better instead of the worst, but there is a reason these sorts of unprecedented upheavals in executive circles are without precedent.

 

[AnandThenMan]

One sad thing is that even when AMD was repeatedly sodomizing intel's CPUs (First with a Sledgehammer and then with a Clawhammer!) intel never lowered its prices for their top-tier products. I truly hope that BD gains AMD a lot of respect from the enthusiast community.

Intel didn't have to lower prices because the playing field was not level. We all know why. The very same thing is happening with Zacate, and will happen with Bulldozer. Many are still afraid of retribution from Intel, unfortunately.

 

[HW2050Plus]

Good point. And could be the case. But on the other hand, showing a strong Zambezi demo would cut into Intel's SB sales more so than current AMD sales in my opinion. Which makes me think it is either one of the other 2 possibilities you mentioned "performance isn't very good, or silicon isn't really ready yet and BD will be delayed more".

Even though I am not an AMD fan, I am very excited to see BD performance.

Why should the company go for "we hurt ourselfes but we hurt the competition more". That really makes no sense. AMD is not to hurt competition but to make business - there own business. Only after that prime maxime is fully satisfied, there might be some thinking about how to hurt competition. It is clear that they want to get the maximum of their 45 nm sales as they can get.

 

[Edrick]

It is clear that they want to get the maximum of their 45 nm sales as they can get.

I highly doubt that. They are not making much money from their 45nm sales since they have to price them so low. Many people have said that AMD does not want to sell this cheap, they just have no choice at the moment.