AMD Bulldozer Eng. Sample leaked, benched

[MilkyWay]

Theres no way they would release a processor that was worse in performance than thier older processors.

I think i will wait for some real world performance tests in things like games or media encoding which could be boosted due to the extra cores.

 

[NdMk2o1o]

Theres no way they would release a processor that was worse in performance than thier older processors.

I think i will wait for some real world performance tests in things like games or media encoding which could be boosted due to the extra cores.

+1

Suggesting such a thing is idiotic imo

 

[twilyth]

Everybody needs to at least look at the article Velvet posted here

The System managed to obtain a score of 26676 marks (CPU – 81917 |GPU – 19718). I have to say that the CPU score shows the power of the Extra Cores. The site managed to compare the Score with an Intel based Rig (i7 2600K + GTX 580) which managed to get a score of 29125 however when we look at the details the score shows that the GPU was the only reason for the overall score being higher, The GTX 580 being a faster card than the 560 scored higher while the CPU (i7 2600K) was a little less than 20,000 marks away from the Fx-8110 performance level with a score of 64146. Here let me make it all easy for you guys to understand:

3D Mark Vantage CPU Score:

AMD FX-8110(3.8Ghz): 81917 CPU Marks
Intel Core i7 2600K(Stock): 64146 CPU Marks



Read more: http://wccftech.com/2011/05/30/amd-...dmark-vantage-cpu-test-details/#ixzz1Os6xdFgt

 

[MilkyWay]

A 2600k is 3.4ghz stock and the bulldozer they used was 3.8ghz i wonder if it made any difference if they where the same speeds? Maybe not a massive difference im guessing.

If the above post is anything to go by the FX chips look decent, maybe should be compared to socket 2011 whenever those are released.

 

[twilyth]

I want to point out that the BD's score was 27.5% higher than sandy's but the clock speed was only 11% higher for BD.

I do the math so you don't have to.

 

[SaiZo]

Could the entire thing be photoshopped? I actually looked at the fifth picture (from top), where the CPU is seated on the socket. They forgot to have that black thing they had on the first picture. So I opened up PSD, and started to mess around.. Can't read the exact figures but if I manage to get it readable - should I post it here??

 

[Wyverex]

If the Vantage score is anything to go by, (at least in multi-threaded applications) 8-core AMD FX is 14% faster than 4-core Sandy Bridge clock-for-clock.

Quite impressive and terrible at the same time.
Impressive due to cost and AMD finally, once again, being competitive.
Terrible due to AMD needing 8 cores to compete with a quad (true, the quad does have HT so it "acts like it has more cores")



But, I don't really believe neither of the tests. I refuse to get hyped and/or trust any pre-release info.

*Wyv impatiently waits for official release and legit reviews*

 

[Velvet Wafer]

If the Vantage score is anything to go by, (at least in multi-threaded applications) 8-core AMD FX is 14% faster than 4-core Sandy Bridge clock-for-clock.

Quite impressive and terrible at the same time.
Impressive due to cost and AMD finally, once again, being competitive.
Terrible due to AMD needing 8 cores to compete with a quad (true, the quad does have HT so it "acts like it has more cores")



But, I don't really believe neither of the tests. I refuse to get hyped and/or trust any pre-release info.

*Wyv impatiently waits for official release and legit reviews*

To be true, the Technology implemented is similar to Hyperthreading, in that there are not really 8 cores, but rather 4 modules, with each 2 "cores"...
so i guess its not too surprising, and also not too terrible,regarding, that even real 6 cores have problems competing with SB

 

[BarbaricSoul]

I own't believe any pre-release benches until they are backed up by a reputable website like TPU or HARD, or they come from AMD themselves

 

[H82LUZ73]

I call fake on 2 reasons
1 CPUZ needs to be beta 1.58 for BD to be read ,and look at the vids and pics at the FX logo see the greyish shadows on it,

2 the most overlooked of the pics and vids of this so called BD in action why the F is the tpd at 186 when we ll know the chips will be 125 tpd and 95 tpd for the lower A series.

Well if you looked at the Asrock extreme 5 990fx manual you will see a pre shot of the chip and it also has a notch near the gold triangle ,This so called chip is a x6 thuban ES chip.

 

[Zen_]

I was reading the other day about how AMD were not happy with the performance of BD on the early sample chips, After reading that link in the OP my guess is that the CPU shown is one of the first (engineering samples) that AMD were not happy with.

My understanding is AMD are working on new stepping (B2 or something) that should be in the retail CPU's.

This is true and was already reported awhile back.

You guys can just take this with a grain of salt from random guy on the internet, but I don't think it's a big secret that earlier this year AMD summoned all their senior engineers to an emergency meeting in Austin to discuss a serious kink in the Zambezi architecture (hence the B2 stepping). For the casual observer that's not an electrical engineer it's almost impossible to fathom how complicated these things are.

 

[CrackerJack]

That's not helping me or anyone else understand.

Sorry, thought I was pretty clear. Wasn't trying to be rude

Check here, those results are from x2 5k blacky...

BD= 23 sec
x2 5000= 29 sec

So either it's a fake... or a fail.. or just maybe a horrible bug...

 

[JrRacinFan]

I don't call this fake, I call it moreless early sample BS. Sure this may be a leaked preview but too early of a stepping as mentioned earlier.

 

[twilyth]

Sorry, thought I was pretty clear. Wasn't trying to be rude

Check here, those results are from x2 5k blacky...

BD= 23 sec
x2 5000= 29 sec

So either it's a fake... or a fail.. or just maybe a horrible bug...

It probably was for someone who benches and knows what a good 1M score is. But even here there are different levels of geekdom. I'm sorry for being terse, but in situations like that, some of us need at least a clue. I don't really mean that as a criticism either, just an observation.

To show the depth of my ignorance, I have this idea that super pi is single threaded. If that happens to actually be true, then there are probably a couple of possible explanations. Maybe the pgm was only use one core of one module rather than the whole module or maybe it needs to be modified given how different BD's architecture is. IDK. I could just be pissing in the wind here. It's just hard for me to belief it would suck so bad on super pi and be so much better on 3D Mark Vantage

 

[cadaveca]

Maybe the pgm was only use one core of one module rather than the whole module or maybe it needs to be modified given how different BD's architecture is

I hope this is the case. Otherwise...UGH.

Pricing that we "know about" now kinda has me hopeful, though.

 

[cheesy999]

I hope this is the case. Otherwise...UGH.

Pricing that we "know about" now kinda has me hopeful, though.

if that is the case we have more performance per core at a lower ghz so its less of a 'i hope this is the case' and more of A 'If it is AMD will rule supreme)

 

[twilyth]

AMD will rule supreme

Mmmmm. I like the sound of that.

Yes. Very nice. Very, very nice.

 

[TheMailMan78]

Waste of time until we see real benches.

 

[cheesy999]

Waste of time until we see real benches.

for once in my life i agree with mailman, tonight we take a visit to AMD headquarters to 'look' at the engineering samples

 

[Wyverex]

To be true, the Technology implemented is similar to Hyperthreading, in that there are not really 8 cores, but rather 4 modules, with each 2 "cores"...
so i guess its not too surprising, and also not too terrible,regarding, that even real 6 cores have problems competing with SB

I know that, but AMD insists on calling those real cores... that's why I'm calling them cores too

 

[cheesy999]

I know that, but AMD insists on calling those real cores... that's why I'm calling them cores too

well they share a lot less then a lot of people think, its less 'fake cores' and more of 'scales of economy' getting rid of things that don't need to be duplicated

 

[Damn_Smooth]

I find it a good sign that a chip that was never meant for consumer hands, a chip that has problems booting windows, can outperform Sandy in anything.

It definitely makes waiting for real performance scores a lot easier.

 

[seronx]

Just going to help you guys burn it in


1, CPU-Z is a fake, the CPU-Z guy hasn't even gotten a FX Chip yet(and if it was done regardless it would say AuthenticAMD)

2. If he was using a B0 chip it is 1 year old and in no way reflects the September Chips

3. AMD ES chips show up as AuthenticAMD and no model number is shown clock speed is also not shown when done in Cinebench


--------------------
The main differences why the AMD CPUs thread amount is consider core amount is because it is not bottlenecked via resources

i7 2600k
L1.
4 x 32 KB instruction caches
4 x 32 KB data caches
L2.
4 x 256 KB
L3.
8 MB

Latency:
4 (L1 cache)
11 (L2 cache)
25 (L3 cache)

AMD Bulldozer 8-core
L1.
4x64KB instruction caches
8x16KB data caches
L2.
4 x 2MB
L3.
8MB

I only know the L3 cache latency for AMD Bulldozer
Latency:
19 (L3 Cache)

Bulldozer also has 2 schedulers vs the 1 scheduler in SB

 

[cdawall]

i have had major issues with prerelease chips and current software or BIOS's

 

[cadaveca]

BIOS's

Best i can come up with. Unless you are tlaking drivers, software age doesn't matter...it either goes faster, slower, or just the same. It's not there will magically appear new versions of EVERYTHING for Bulldozer...