Fudzilla: Bulldozer performance figures are in

[3DVagabond]

This is interesting. From a thread at Semi Accurate. The thread topic is a pole on what people think the release stepping of Bulldozer will be. The choices are, B0, B1, B2, B3, C0, other. Rich Wargo, the person responding works for Global Foundries. His response, "other, not guessing", is very interesting to me. So, what will the release spin be? B4, C1?

 

[drizek]

E1

 

[3DVagabond]

E1

And your information comes from...?

 

[greenhawk]

the person responding works for Global Foundries. His response, "other, not guessing", is very interesting to me. So, what will the release spin be? B4, C1?

more likly he thinks (or the information is) company private and so does not want to run the risk of loosing his job over it.

 

[3DVagabond]

more likly he thinks (or the information is) company private and so does not want to run the risk of loosing his job over it.

Why respond at all if he's concerned with getting in trouble? Also, this guy is very credible. I would be shocked if he gave out incorrect info.

 

[drizek]

so don't encrypt it

the encryption performance is not really in 99% of the buyers' minds, nor will it be.

It is in the minds of 100% of enterprise purchasers. It is in Apple's mind, as they completely reworked encryption in Lion. It is in Microsoft's mind since they developed BitLocker for windows.

Oh, and I want it.

 

[drizek]

And your information comes from...?

Oops, sorry, replied to the wrong thread. Someone was asking me where I wanted to move my Knight.

 

[-Slacker-]

An overclocking of BD with what seems to be an early ES,
by the same polemical guy...
We can estimate that Superpi 1M is executed in a time
that is between 10 and 14s , although the way the cpu
start the computation show that core affinity is not
set to core 1 , wich is the one actually executing
the soft.
http://www.youtube.com/watch?v=IO7NcbcUjEM&feature=player_detailpage#t=4s

6483mhz?

 

[3DVagabond]

Oops, sorry, replied to the wrong thread. Someone was asking me where I wanted to move my Knight.

:thumbsup:

 

[BlueBlazer]

The original phenom debuted at a price higher than the Q6600.

Thuban 1090T also debut at around $300. Initial launch pricing are always higher, but it does give some indication (that BD is not going to compete with 6 core Gulftown).

According to Intel , for professionnal purposes were FP capabilities were important, a C2Q6600 was no match for a 2g Barcelona.
How much was the cost of a xeon.????....

"Professional purposes" you've mentioned are usually HPC type applications where FPU is heavily utilized (which is why K10 was used in Cray). However most server applications rely heavily on integer operations (this is where majority of server market depends on). Early FSB based Core 2 Quad has memory bandwidth limitations, which plays some role in HPC applications. Then there are those (HPC applications) that depend less on memory bandwidth as shown here.

6483mhz?

Equivalent to 6.483GHz (roughly 6.5GHz). Another one here also shows clock speed in MHz units.

 

[Abwx]

"Professional purposes" you've mentioned are usually HPC type applications where FPU is heavily utilized (which is why K10 was used in Cray).

According to Intel , for professionnal purposes were FP capabilities were important, a C2Q6600 was no match for a 2g Barcelona.

However most server applications rely heavily on integer operations (this is where majority of server market depends on). Early FSB based Core 2 Quad has memory bandwidth limitations, which plays some role in HPC applications. Then there are those (HPC applications) that depend less on memory bandwidth as shown here.

Perhaps less in bandwith but most surely in optimisations provided
mainly with such tools as an Intel compiler...

 

[BlueBlazer]

Perhaps less in bandwith but most surely in optimisations provided
mainly with such tools as an Intel compiler...

If you want to blame everything on Intel compiler then check what Anandtech review says over here, the same E5472 in the LS Dyna chart >>

Intel has made some successful changes to the quad-core Xeon that have helped it achieve as much as a 56% lead in performance over the 2.0GHz Barcelona part.

That's server benchmarks (software used check here) and are mostly integer operations (note that HPC applications account for tiny percentage of the overall segment). That is the overall picture when it comes to server chips.

 

[Abwx]

If you want to blame everything on Intel compiler then check what Anandtech review says over here, the same E5472 in the LS Dyna chart >> That's server benchmarks (software used check here) and are mostly integer operations (note that HPC applications account for tiny percentage of the overall segment). That is the overall picture when it comes to server chips.

Why not quote the full sentence ?...()

Intel has made some successful changes to the quad-core Xeon that have helped it achieve as much as a 56% lead in performance over the 2.0GHz Barcelona part. Of course this is mostly due to the fact that the Harpertown part has a 1GHz clock speed advantage, and the various micro-architecture tweaks surely help fill in the rest.

As well as some benches from the article ...?..

 

[Arzachel]

pretty sure we'd have leaked legit numbers if BD were worth buying.

AMD have a dog and we all know it deep down in our bones. They'd be leaking benchmarks by now if they have a winner.

Yup, just like with the HD69XX series, search for the thread, its almost identical to this one right down to people crying lemon over no leaks. Or might I remind you of K8 with the only leaks being made with a crippled ES?

Bulldozer won't be the second coming of Jesus nor will it be as bad as some are making it out to be. It will probably beat similarly priced SB's at highly threaded tasks, lose at single threaded tasks. Just by how much is the interesting part

 

[podspi]

I expect it to be very reminiscent of the Thuban launch, except I honestly expect it to do better w.r.t. singlethread performance vs. S.B. than Thuban did vs. Nehalem. (Stock).



Overclocked, we'll see. Hopefully most BD will be able to break 5ghz, otherwise I just don't see how they could compete with SB in ST performance (unless we've all just really underestimated IPC).

 

[Abwx]

I expect it to be very reminiscent of the Thuban launch, except I honestly expect it to do better w.r.t. singlethread performance vs. S.B. than Thuban did vs. Nehalem. (Stock).



Overclocked, we'll see. Hopefully most BD will be able to break 5ghz, otherwise I just don't see how they could compete with SB in ST performance (unless we've all just really underestimated IPC).

With those 315mm2 and about 40% higher density , it has about
double the transistors count of a 4C/8T IGP less SB.

So assuming a very unfavourable cubic root law in respect of
perfs increasement/transistor count increasement , it should
provide about 30% better perf , wich would be quite disappointing..

 

[radaja]

some new results from chiphell

truecrypt updated, 990x performance is not normal.
Official website of the Intel Xeon W3680 AES is 4.7G.
990x should be higher than this.

FX8150P at 4.2GHz

i7 990X at 3.4GHz

 

[-Slacker-]

^I'm guessing that truegrypt is a heavily multithreded application.

 

[frostedflakes]

I think it can take advantage of as many threads as you can throw at it (within reason), so yeah it's probably utilizing 8 threads on the FX-8150P and 12 on the 990X.

Also what's the deal with AMD's turbo anyway? Is the 4.2GHz in the above screenshot the all core turbo, or was the CPU OC'd to 4.2GHz non-turbo clock? If the former, what does it turbo to when fewer cores are being utilized? Does it go even higher than 4.2?

 

[Riek]

With those 315mm2 and about 40% higher density , it has about
double the transistors count of a 4C/8T IGP less SB.

So assuming a very unfavourable cubic root law in respect of
perfs increasement/transistor count increasement , it should
provide about 30% better perf , wich would be quite disappointing..

What would be interesting is how would BD performance be affected with half the l2 and l3 cache sizes.. given that those 2 are real die space eaters. They used to do this with previous cpu's also. Launch with bigger cache and the next revision lower the cache.

 

[drizek]

some new results from chiphell




FX8150P at 4.2GHz

This seems consistent with no real increase in IPC over Phenom II.

 

[podspi]

This seems consistent with no real increase in IPC over Phenom II.

How do you figure? Phenom II doesn't have AES. I assure you my X6 would do all of those much, much slower

 

[drizek]

Just comparing Twofish results with what my Phenom II gets.THis test might not be representative of overall performance, but there is nothing in it that would suggest that AMD is any more competetive with Intel now clock for clock.

 

[allies]

Just comparing Twofish results with what my Phenom II gets.THis test might not be representative of overall performance, but there is nothing in it that would suggest that AMD is any more competetive with Intel now clock for clock.

8 core (AMD) vs 12 core (Intel): Intel with 50% more cores
4.2GHz (AMD) vs 3.4GHz (Intel): AMD 23.5% faster than Intel

Scores seem to be a little give or take. This doesn't paint a favorable light in AMD's favor (if the benchmarks are true)?

 

[Cogman]

Do people REALLY use AES encryption all that often to justify the new extension? I can understand FMA, but AES just seems a little meh for most.