Rumour: Bulldozer 50% Faster than Core i7 and Phenom II.

[grimpr]

Rumour: Bulldozer 50% Faster than Core i7 and Phenom II.

http://www.techpowerup.com/138328/Bulldozer-50-Faster-than-Core-i7-and-Phenom-II.html

Here, take some salt. AMD reportedly gave out performance figures in a presentation to its partners, performance figures seen by DonanimHaber. It is reported that an 8-core processor based on the "Bulldozer" high-performance CPU architecture is pitched by its makers to have 50% higher performance than existing processors such as the Core i7 950 (4 cores, 8 threads), and Phenom II X6 1100T (6 cores). Very little is known about the processor, including at what clock speed the processor was running at, much less what other components were driving the test machine.

Taking this information into account, the said Bulldozer based processor should synthetically even outperform Core i7 980X six-core, Intel's fastest desktop processor in the market. Built from ground-up, the Bulldozer architecture focuses on greater inter-core communication and reconfigured ALU/FPU to achieve higher instructions per clock cycle (IPC) compared to the previous generation K10.5, on which its current Phenom II series processors are based. The processor is backed by new 9-series core logic, and a new AM3+ socket. AMD is expected to unveil this platform a little later this year.

http://www.donanimhaber.com/islemci...er-islemcisi-Core-i7-950den-50-daha-hizli.htm

As mentioned above, but not yet test results are detailed in Hardware News AMD's official documentation of performance we were able to reach. In this document, AMD Bulldozer 8-core processor (model name and clock speed performance segment, but that information has not been specified), 6-core Phenom and Core i7 950 and compares II X6 1100T. The estimated results in the document, processors, three different categories (media, rendering and game) in comparing the overall assessment of the Bulldozer processor 8-core Core i7 950 is 50% faster stressed. Carefully examined the performance table in the 8-core processor, AMD's Bulldozer makes a difference, especially game and rendering tests.

Closed on account of fanboys and stupid bickering. Half you are brain-dead, the other half of you are also brain-dead. Learn to live with each other.

Please feel free to start a new thread.
-ViRGE

 

[Riek]

Rumour: Bulldozer 50% Faster than Core i7 and Phenom II.

http://www.techpowerup.com/138328/Bulldozer-50-Faster-than-Core-i7-and-Phenom-II.html
http://www.donanimhaber.com/islemci...er-islemcisi-Core-i7-950den-50-daha-hizli.htm

well, JF once said they expect 50% more throughput for interlagos compared to MC 12core.

Now a rumour is BD 8core is 50% faster then Thuban x6 1100T.
Seems to me to be an incorrect interpretation of previous known information. (statement 33% more cores for 50% more performance).

 

[eternalone]

This would explain them bringing back the FX Line up, which would indicate performance confidence we havent seen in along time.

http://www.nordichardware.com/news/...brings-back-fx-processors-with-bulldozer.html

 

[Nemesis 1]

Ya AMD said the same about Phenom I didn't they . Slides are still in those old topics.

 

[HW2050Plus]

This comes without surprise for all who studied the Bulldozer architecture, as it will give a 70-100% performance boost over previous generation (Phenom) by design.

However it will still be intersting in which range of the 70-100% they will be.

If they are still 50% faster than a Phenom with 2 more cores that could indicate that they are near the upper range of the expected margin, however that must be backed by benchmark results.

 

[IntelUser2000]

My 2 cents:

Phenom II X4 965 vs. Phenom II X6 1100T
http://www.anandtech.com/bench/Product/102?vs=203

On the extremely well threaded applications like x264 HD Encode 2nd Pass, Cinebench Multi-thread, or POV-ray 3.7 Beta 23 SMP benchmark, and Microsoft Excel 2007 - Monte Carlo simulation the performance gains are very close to 50%, or the ratio of increase in core count.

(Results go like: Phenom II X6 1100T(Core i7 980X))

x264 HD 2nd pass: 47.5%(144%)
Cinebench R10 Multi-thread: 34.3%(93.1%)
POV-ray 3.7 Beta 23 SMP: 43.9%(137%)
Microsoft Excel 2007 - Monte Carlo Simulation: 46.3%(180%)

My first ever prediction for Bulldozer was that AMD would try to meet its even earlier estimations regarding "Clustered Multi-threading", where the "2nd thread/core" gains 80% of performance. On the heavily threaded applications mentioned above, the 80% gains just enabling Clustered MT should be correct.

(The following estimates assume Bulldozer is a 4 module 8 core)

Clustered Multi-threading: 1.8x
Single thread advantage of Nehalem/Westmere over Phenom II "Stars": 1.15x*
Common knowledge gain for Hyperthreading: 1.3x
Nehalem vs Stars: ~1.5x
Moving from 4 to 6 cores: 1.32x
AMD's claim for Interlagos: 1.5x
Interlagos estimate + moving from 4 to 6 cores: 2x
Westmere 6 core vs Stars 4 core: 2x
Westmere to Sandy Bridge: 1.15x
Conclusion 1, Bulldozer vs Westmere 6 core: 1.1-1.15x in favor of Westmere
Conclusion 2, Bulldozer vs. Westmere 6 core: 1-1.05x in favor of Westmere, the differences from #1 being Bulldozer includes a more aggressive Turbo Mode. Of course there will be applications that go either camp because of the different architecture between the two.

Bulldozer vs. Sandy Bridge 4 core: 1.15x in favor of Bulldozer
Bulldozer vs. Sandy Bridge 6 core: 1.15x in favor of Sandy Bridge


*those that do not believe me compare Cinebench R10 Single threaded results for Phenom II X6 1090T and Core i7 975X. Both have roughly equal single core Turbo speeds. Plus or minus 2-3% variation might occur because Core i7's all core Turbo sometimes engages to 1 step lower. Likely a minor scheduling and benchmark issue.

 

[JMapleton]

I'll say exactly what the AMD fanboys said before Conroe came out.

"I'll believe it when I see it."

Albeit, the AMD fanboys were proven wrong. *scratches head*

 

[IntelUser2000]

I don't believe AMD will gain the performance crown either. But they will be far closer thanks to proliferation of multi-threaded apps. Die size/performance wise they'll squeeze in between 4 core and 6 core Sandy Bridge chips. It's also likely even Nehalem will have some single thread advantage.

 

[Axon]

I don't believe AMD will gain the performance crown either. But they will be far closer thanks to proliferation of multi-threaded apps. Die size/performance wise they'll squeeze in between 4 core and 6 core Sandy Bridge chips. It's also likely even Nehalem will have some single thread advantage.

I expect this as well. Nothing to do but wait on the benches. AMD needs to move already.

 

[grimpr]

It's also likely even Nehalem will have some single thread advantage.

So what?, noone buys 6 and 8 core monsters to run excel and play solitaire, if the architecture is designed to more multithreading throughput at a smaller die size and a decent price tied to a mainstream platform then its a winner.

 

[IntelUser2000]

Grimpr, that's not what I am saying, not to mention Solitaire would be fine on a Pentium MMX, heh. That's just a side point to illustrate the comparison isn't black and white. I do know 5% or so per core advantages Nehalem might have over Bulldozer isn't anything worth to talk about when Bulldozer has 33%-100% more cores.

 

[FalseChristian]

The real point is that both are gonna be uber performers so you can't go wrong with either. This'll make AMD more expensive and Intel less so.

 

[HW2050Plus]

Bulldozer vs. Sandy Bridge 4 core: 1.15x in favor of Bulldozer
Bulldozer vs. Sandy Bridge 6 core: 1.15x in favor of Sandy Bridge

Though I am not saying I agree on those numbers that could be a possible outcome, however you should write it like that:

2011:
Bulldozer 4 module (8 core) vs. Sandy Bridge 4 core (8 threads): 1.15x in favor of Bulldozer
Bulldozer 4 module (8 core) vs. Sandy Bridge 6 core (12 threads): 1.15x in favor of Sandy Bridge
Bulldozer 6 module (12 core) vs. Sandy Bridge 6 core (12 threads): 1.15x in favor of Bulldozer
Bulldozer 8 module (16 core) vs. Sandy Bridge 8 core (16 threads): 1.15x in favor of Bulldozer

2012 (die shrink of Intel):
Bulldozer 8 module (16 core) vs. Sandy Bridge 12 core (24 threads): 1.15x in favor of Sandy Bridge

where die size of 4 module Bulldozer is roughly equal to 4 core Sandy Bridge (note: both are on 32 nm), though it actual depends on the exact amount of L3 cache of the part. With a die shrink in 2012 Intel could reduce the gap / improve die sizes and therefore increase core count.

So in 2011/2012 you will have really well performing chips at a low price range. The crown will go to Bulldozer FX which will be likly priced very high, since Intel will compete with lower prices (the lower Sandy Bridge prices by now already indicate that), more production capacity and selling power.

Intel will likly answer by adding more cores, so you can be sure to see at least a 8 core / 16 thread Sandy Bridge and with die shrink in 2012 a 12 core part doesn't seem unlikly.

All said for overall performance, since there will be lot of performance differences, e.g. chess programs which even drop with hyper threading or on the other side other programs which scale not very good because of bad programming. Therefore the above regarding all benchmark results taken together to an overall performance.

Still open and interesting will be average power dissipation. This could be another large gain for AMD if you take less modules (and therefore a bit less performance) but you gain a huge power saving. I mean Bulldozer can give 4 core stars performance at the power dissipation of a dual core.

With AMD Bulldozer you can choose between (nearly) double performance (over Phenom) or less than half the power (over Phenom) as of today.

AMD will have a window of ~1.5 years until Intel incorporates the "module technology" in their chips.

 

[Riek]

I don't believe AMD will gain the performance crown either. But they will be far closer thanks to proliferation of multi-threaded apps. Die size/performance wise they'll squeeze in between 4 core and 6 core Sandy Bridge chips. It's also likely even Nehalem will have some single thread advantage.

I do not have the time to look into the math you gave, so i will not comment on that. However if we assume the statement about BD are correct including the performance figures:
BD 8core is 50% faster then i7 950. 50% was an average about stressed test with rendering, video and gaming. The statement was that the biggest differences were in rendering and gaming.

So to get an 50% average every change ha to be around 50% or their is a fluctuation between low and highs. They specifically mention gaming and rendering are better. That means those both are above 50% or on is close to 50%. We also know rendering and gaming are total different markets. rendering = multithreaded, gaming = single threaded. For rendering this means BD will be very close to gulftown. For gaming however the difference between i7 950 and 2600K (highest performing part) is less than 50%. So single threaded/dual threaded seems better on BD then SB (SB @ 3.8Ghz, BD unknown).

 

[IntelUser2000]

Games are persistently low-threaded. Unless they are talking about 3DMark CPU score or ones like Lost Planet running on all low with a GTX580, I wouldn't expect anywhere near 50%.

 

[wwswimming]

i would like to obtain about a 100% increase in performance from the 9550 quad core - for $500 ... for multi-threaded aps.

the 980X would make the upgrade worth it.

here's hoping the Sandybridge release and/or Bulldozer will bring us a $500 980X, or equivalent.

 

[Maximilian]

Exciting Hope this is all true and we return to the days of pentium 4/athlon XP neck and neck competitiveness.

 

[Phynaz]

It is reported that an 8-core processor based on the "Bulldozer" high-performance CPU architecture is pitched by its makers to have 50% higher performance than existing processors such as the Core i7 950

An eight core cpu is faster than a four core cpu.

Huh, who woulda guessed?

 

[exar333]

An eight core cpu is faster than a four core cpu.

Huh, who woulda guessed?

LOL, you said it!

 

[Idontcare]

An eight core cpu is faster than a four core cpu.

Huh, who woulda guessed?

has begun, these core wars

 

[Rubycon]

Exciting Hope this is all true and we return to the days of pentium 4/athlon XP neck and neck competitiveness.

Sadly there was no real competition with Athlon vs. P4 (netburst).

Conroe was a true breakthrough as was the original Slot-A Athlon was for AMD back in '99.

If this future chip from AMD can run or be o/c to 4-5GHz then sure its performance should be quite good.

 

[sawtx]

An eight core cpu is faster than a four core cpu.

Huh, who woulda guessed?

8 thread vs. 8 thread, seems like a reasonable match up.

 

[Maximilian]

Sadly there was no real competition with Athlon vs. P4 (netburst).

Conroe was a true breakthrough as was the original Slot-A Athlon was for AMD back in '99.

If this future chip from AMD can run or be o/c to 4-5GHz then sure its performance should be quite good.

Aw come on it went back and forth quite a bit back then:

athlon > P4
P4B > athlon
athlon XP > P4B
P4C > athlon XP
to athlon 64/x2 dominating for a few years until conroe where intel has dominated since.

 

[Rubycon]

Aw come on it went back and forth quite a bit back then:

athlon > P4
P4B > athlon
athlon XP > P4B
P4C > athlon XP
to athlon 64/x2 dominating for a few years until conroe where intel has dominated since.

The differences between those were not that great when the P4 was slightly better. (overclocking excluded) Athlon64 certainly pulled ahead considerably until Conroe debuted.

What's crazy is I have a Dell XPS notebook from 2004 that has a 3.4GHz Northwood and 2GB ram. It was very fast (and heavy!) for its time. Today that computer can't even play a 1080P youtube video without stuttering! The video card is no dog (for its time) either - Radeon Mobility 9800 w/256MB. It's amazing how heavy the footprint of even surfing the web with multimedia content has become.

 

[zsdersw]

This comes without surprise for all who studied the Bulldozer architecture, as it will give a 70-100% performance boost over previous generation (Phenom) by design.

However it will still be intersting in which range of the 70-100% they will be.

If they are still 50% faster than a Phenom with 2 more cores that could indicate that they are near the upper range of the expected margin, however that must be backed by benchmark results.

You seem very certain about something that hasn't been released, much less benchmarked.