AMD Bulldozer performance benchmarks?

[ambam]

AMD's new "bulldozer" chips are allegedly on par with the performance of an overclocked Sandy Bridge. If this is actually true, then huge Kudos to AMD for trying to take back the performance throne from Intel. Just like how the Athlon 64 smoked the Pentium 4 in almost every benchmark.

Are there any Bulldozer benchmarks to back this up?

Also, the enthusiast-level bulldozer chips have 8 physical cores and 16 physical threads. Each core has two CPU modules in them. Meaning that AMD's approach at hyper-threading should be faster than Intel's logical threads.

 

[popatim]

There are none yet. You can visit the Bulldozer rumors thread at the top of this section for lots of info and or speculation.

 

[jaguarskx]

wait...

 

[ambam]

wait...

The bulldozers will probably be closer in performance to the Core i7 Nehalems than the Sandy Bridge or Ivy Bridge.

 

[hunter315]

Says who? A random anomous poster on a chinese forum? Thats where the "leaked" benchmarks usually come from

None have been leaked yet, so my guess that its going to be as fast as me using an abacus is just as valid as a guess that it will be 100x faster than the i7-2600K

 

[mdd1963]

AMD's new "bulldozer" chips are allegedly on par with the performance of an overclocked Sandy Bridge. If this is actually true, then huge Kudos to AMD for trying to take back the performance throne from Intel. Just like how the Athlon 64 smoked the Pentium 4 in almost every benchmark.

Are there any Bulldozer benchmarks to back this up?

.

No....

Not a single reputable benchmark has been released/leaked....; I would love them to even be able to match SB clock for clock or even in performance for the dollar, but it is probably wishful thinking to think it will exceed SandyBridge...

 

[buzznut]

AMD's new "bulldozer" chips are allegedly on par with the performance of an overclocked Sandy Bridge. If this is actually true, then huge Kudos to AMD for trying to take back the performance throne from Intel. Just like how the Athlon 64 smoked the Pentium 4 in almost every benchmark.

Are there any Bulldozer benchmarks to back this up?

There is plenty of dubious benchmarks out there, but we really don't know. AMD claims "greater than 2X Ph II X6 1100T" performance, but we don't know in what workload they're talking about.

However it does seem likely that the 8 core CPU's will match Nehalem performance, which is pretty damn good.

As far as hyperthreading goes, I don't think its similar at all. Hyperthreading takes a physical core and allows it to work on two threads as two logical cores. The result is not quite as good as a physical core can produce, but its pretty damn fast.

The Zambezi CPU's will have 8 integer processors but only 4 floating point units. How this will work in real world performance is what is unknown, but I wouldn't count on it to be better than current core i5/i7. AMD is looking forward with this architecture, it may be some time for it to mature but they are gambling that it will pay off in the long run.

So the simple answer is no, AMD's architecture is not known to be faster than hyperthreading.

 

[megamanx00]

Meh, you'll see what bulldozer can do when it launches and it's already pretty darn late in my opinion.

 

[pjmelect]

On a per core bases and at the same clock speed on multi-threaded benchmarks then Bulldozer may well be faster than Sandy bridge. However on single threaded programs I expect Bulldozer to be behind Sandy bridge.

 

[dalauder]

I was hoping someone actually had new information. So far all we have are the same facts from three months ago:

-Bulldozer is late and has been delayed too many times. Who knows when it will actually arrive?
-AMD is having trouble with the manufacturing process.
-There are no performance benchmarks, but speculations is that it will lose clock-for-clock to SB so it uses more cores.
-At least AMD believes in keeping sockets so you can upgrade them.

 

[ambam]

AMD's strategy is to make decent-performing chips for a relatively low price.

AMD has never charged >$1,000 for a CPU. Like Intel's unnecessarily expensive $1,000+ "extreme edition" processor lineup.

The bulldozers will probably be priced +/- $300 and offer performance slightly faster than the Nehalems. Also, they will work with AM3+ sockets.

 

[buzznut]

I was hoping someone actually had new information. So far all we have are the same facts from three months ago:

-Bulldozer is late and has been delayed too many times. Who knows when it will actually arrive?
-AMD is having trouble with the manufacturing process.
-There are no performance benchmarks, but speculations is that it will lose clock-for-clock to SB so it uses more cores.
-At least AMD believes in keeping sockets so you can upgrade them.

thats a reasonable summary. Who knows where this will lead though? A couple of years ago everyone was saying what a phenomenally stupid idea it was for AMD to aquire Ati. They said AMD was sure to fail in a year, that they couldn't compete with both nVidia and Intel, and that they would simply go out of business.
At first, AMD's new graphics division was not putting out stellar products. It looked as though the naysayers were right, that AMD had failed in the video card business.
Now, they seem to be firing on all cylinders. They have some compelling products out now and more coming in the future.

AMD still needs market share in graphics cards, but they have provided top notch integrated solutions. I think they have a solid strategy, and yeah its David vs goliath in the CPU arena but already nVidia was forced to give up on chipsets. This is an area where AMD can excel. I think you can make the argument that AMD is more diversified than Intel, and that maybe a better strategy long term.

 

[dalauder]

Well, I left AMD's grand strategy out of my bulldozer summary because it's not relevant. Bulldozer = Enthusiast. Currently if you are an enthusiast, you go with Intel.

Conveniently for AMD, Enthusiasts are not your standard computer user. For most users (like my parents, wife, and brother), an AMD Llano platform would be perfect. The integrated GPU is enough for everything but gaming and adding a discrete card on the 5670 level will allow hybrid crossfire (assuming drivers improve) to provide good gaming performance in all current games.

Intel has no perfect solution for the low budget mainstream user who dabbles in gaming or wants accelerated video playback. An i3-2100 is good, but it's gonna be $50 more than Llano if you want any gaming capability. When we're talking sub $400 prices for entire systems, $50 a big deal.

But we're supposed to be talking about Bulldozer. Bulldozer's a bag of unknowns with a promise to not strand users who want an upgrade path. If everything pans out and Bulldozer's 8 can match Sandy Bridge's 4 in normal applications, then Trinity should be able to displace i3's and i5's with high-end CPU performance and everything the non-hardcore gamer needs in graphics.

Of course, AMD also has a much larger "unlocked" product line than Intel--overclocking might be the difference between a low end Bulldozer and an i3.

Note: I mean good as in medium settings in Just Cause 2 at 1680x1050, possibly no AA.

 

[loneninja]

AMD's strategy is to make decent-performing chips for a relatively low price.

AMD has never charged >$1,000 for a CPU. Like Intel's unnecessarily expensive $1,000+ "extreme edition" processor lineup.

The bulldozers will probably be priced +/- $300 and offer performance slightly faster than the Nehalems. Also, they will work with AM3+ sockets.

AMD's FX processors were priced $1000 before Core2Duo launched in 2006. :lol:

 

[jj463rd]

^ +1
Exactly.
I remember those exorbitantly expensive AMD CPU's.
Couldn't afford one myself.

 

[geekapproved]

This thread makes about as much sense as asking " are we all gonna die on 12-20-12"?

 

[keithlm]

I was hoping someone actually had new information. So far all we have are the same facts from three months ago:

-Bulldozer is late and has been delayed too many times. Even though AMD has said it will be on retail shelves 60-90 days from June 1; Intel fans keep feigning ignorance and asking the same question: Who knows when it will actually arrive?
-Even though AMD mentioned that the last delay was because they wanted the fabs available for Llano, which is a logical and rational explanation: Intel fans keep saying that AMD is having trouble with the manufacturing process.
-There are no performance benchmarks, but speculation by Intel fans with absolutely no data to back up their opinionsis that it will lose clock-for-clock to SB so it uses more cores.

I added some missing details (in bold) to your statements to give you the information you requested in your first sentence.

 

[mdd1963]

Of course, AMD also has a much larger "unlocked" product line than Intel--overclocking might be the difference between a low end Bulldozer and an i3.

.

This 'ease of overclocking' almost allows the X4s clocked near 4 GHz to compete with a stock clocked i5-2300, even when giving the X4 a 33% clock speed advantage...

(I sincerely hope BD is able to close this obvious IPC disparity; better for all of us!)

 

[jaguarskx]

This thread makes about as much sense as asking " are we all gonna die on 12-20-12"?

Well, are we all gonna die?!?!?!

Please, please, please tell me!!!!

This may be my last chance to loose my virginity!!!

 

[dalauder]

I added some missing details (in bold) to your statements to give you the information you requested in your first sentence.

Apparently anyone who's growing skeptical of Bulldozer is an Intel fan? I love AMD and wanted to go K10 with my build until I found out about Microcenter i7 prices.

-Even though AMD has said it will be on retail shelves 60-90 days from June 1; Intel fans keep feigning ignorance and asking the same question: Who knows when it will actually arrive?

After 3 delays, I'm surprised you're certain it will launch on AMD's vague date. I believed it would launch in June when AMD told me that before. They changed the date at the end of May. Who's to say it won't change yet again? That's all I'm saying. I don't want my heart broken again.

-Even though AMD mentioned that the last delay was because they wanted the fabs available for Llano, which is a logical and rational explanation: Intel fans keep saying that AMD is having trouble with the manufacturing process.

If they weren't having trouble with manufacturing, it would've been launched on 22nm last November. If it's conflicting with Lano on the 32nm process, doesn't that mean they're having trouble with the fact that Bulldozer's even being manufactured at 32nm? Is the current stepping error a manufacturing problem or architecture error?

-There are no performance benchmarks, but speculation by Intel fans with absolutely no data to back up their opinionsis that it will lose clock-for-clock to SB so it uses more cores.

True, I do lack data, which is why I qualify that by calling it speculation. Intel promised a 25% per-clock performance boost with SB. They brought it. What performance promises has AMD given us with Bulldozer? None--except for a promise of more cores than what Intel has at the same price points. If Thuban was any indication, AMD needs those cores to keep up. It's speculation based on the knowledge available. That, and that AMD has stated the most recent delay was due to performance not being what was intended.

Do you have a reason that you speculate BD will match SB clock-for-clock? Don't be blinded by being an AMD fan. I'm just saying AMD might only match/beat Intel on the enthusiast end during highly parallel work loads. I'm sad about that too, but I'm trying to look at it realistically.

 

[keithlm]

Apparently anyone who's growing skeptical of Bulldozer is an Intel fan? I love AMD and wanted to go K10 with my build until I found out about Microcenter i7 prices.

Only the people that post opinions and present them as facts without any concrete sources. It's very common on this forum.

After 3 delays, I'm surprised you're certain it will launch on AMD's vague date. I believed it would launch in June when AMD told me that before. They changed the date at the end of May. Who's to say it won't change yet again? That's all I'm saying. I don't want my heart broken again.

There was never any official announcement before it was announced recently that it would be on-the-shelves 60-90 days from the first of June. There was some leaked "internal only" documents that mentioned 2Q2011; but that is not a truly reliable source or quotable source.

If they weren't having trouble with manufacturing, it would've been launched on 22nm last November. If it's conflicting with Lano on the 32nm process, doesn't that mean they're having trouble with the fact that Bulldozer's even being manufactured at 32nm? Is the current stepping error a manufacturing problem or architecture error?

Do you have any sources for the comments I just quoted above? Until we have an indisputable source for these comments then they are only opinions of posters on forum(s) or writers of blogs which have no real credibility.

True, I do lack data, which is why I qualify that by calling it speculation. Intel promised a 25% per-clock performance boost with SB. They brought it. What performance promises has AMD given us with Bulldozer? None--except for a promise of more cores than what Intel has at the same price points. If Thuban was any indication, AMD needs those cores to keep up. It's speculation based on the knowledge available. That, and that AMD has stated the most recent delay was due to performance not being what was intended.

AMD officially stated that they have delayed due to the Llano release/rollout. Any speculation that it was because of performance was made from unconfirmed and questionable sources that have never been confirmed by anybody.

Do you have a reason that you speculate BD will match SB clock-for-clock? Don't be blinded by being an AMD fan. I'm just saying AMD might only match/beat Intel on the enthusiast end during highly parallel work loads. I'm sad about that too, but I'm trying to look at it realistically.

If I had a reason I might do some speculation on this forum. But it would be pointless to speculate for or against BD at this point in time. Although that apparently doesn't stop many people who continually post that BD will never be able to compete.

Personally what I can state with complete confidence is that if AMD created a chip that completely tied Intel in 100% of all available benchmarks, there would be many people that would use that data to declare that AMD just can't compete with Intel in any way.

 

[dalauder]

Only the people that post opinions and present them as facts without any concrete sources. It's very common on this forum.

Look, if you can tell me Bulldozer's launch date and personally guarantee it, I'll start telling people that launch date. But maybe you're right, I should have said, "I have no idea when it will finally arrive" instead of presenting it as a consensus that most people don't know the actual launch date. (Although I think most people don't know the actual launch date).

There was never any official announcement before it was announced recently that it would be on-the-shelves 60-90 days from the first of June. There was some leaked "internal only" documents that mentioned 2Q2011; but that is not a truly reliable source or quotable source.

Do you have any sources for the comments I just quoted above? Until we have an indisputable source for these comments then they are only opinions of posters on forum(s) or writers of blogs which have no real credibility.

Credible sources? Nope, just leaked "internals" that plenty of articles were written about. Don't tell me you never heard that Bulldozer was supposed to skip 32nm altogether and do twenty-something nanometer fabrication. That was back in the start of 2010 when they were talking about quad channel memory and FMA. Not all those rumors turned out to be true.

AMD officially stated that they have delayed due to the Llano release/rollout. Any speculation that it was because of performance was made from unconfirmed and questionable sources that have never been confirmed by anybody.

I was basing it off of articles I could find about Bulldozer like this one: http://www.xbitlabs.com/news/cpu/display/20110601180002_AMD_Confirms_Delay_of_FX_Series_Bulldozer_Microprocessors.html
Google News is the best I've got to find out about Bulldozer. It's probably more than unbased speculation though--I'm as well informed as it's possible to be without working within AMD (or having a good friend who does?).

If I had a reason I might do some speculation on this forum. But it would be pointless to speculate for or against BD at this point in time. Although that apparently doesn't stop many people who continually post that BD will never be able to compete.

Personally what I can state with complete confidence is that if AMD created a chip that completely tied Intel in 100% of all available benchmarks, there would be many people that would use that data to declare that AMD just can't compete with Intel in any way.

You're taking this the wrong way. I'm not saying AMD sucks or anything to that effect. What I am saying is that Bulldozer is starting to feel a lot like Duke Nukem Forever and that has to sip away at confidence in the product.

I will be the first one on this forum to recommend a Bulldozer build AFTER it launches IF it does well in gaming benchmarks. But until then...all I can do is let my confidence in Bulldozer slowly wane as AMD tests everyone's patience. All I can do is make guesses with the information at hand and summarize what articles about Bulldozer say.

Go ahead and disagree with me and say I've summarized articles incorrectly--but don't label me as an Intel fanboy.

EDIT: To be clear, I expect Bulldozer to release in less than 14 years of development and be a MUCH better product than Duke Nukem Forever.

 

[keithlm]

Don't tell me you never heard that Bulldozer was supposed to skip 32nm altogether and do twenty-something nanometer fabrication. That was back in the start of 2010 when they were talking about quad channel memory and FMA. Not all those rumors turned out to be true.

I never read any rumors about 22nm. I have read articles about the fact that the rush to smaller and smaller die sizes won't always create benefits; that there the processes will "hit a wall".

I was basing it off of articles I could find about Bulldozer like this one: http://www.xbitlabs.com/news/cpu/display/20110601180002_AMD_Confirms_Delay_of_FX_Series_Bulldozer_Microprocessors.html
Google News is the best I've got to find out about Bulldozer. It's probably more than unbased speculation though--I'm as well informed as it's possible to be without working within AMD (or having a good friend who does?).

The xbitlabs article quotes no sources for their speculation. The only known source for that particular speculation was the supposed engineer at a motherboard company. But nobody can really confirm the information so it is heresay. There was also supposedly a disgruntled AMD employee that "spilled the beans". But I don't take the word of disgruntled employees as being credible.

 

[dalauder]

I never read any rumors about 22nm. I have read articles about the fact that the rush to smaller and smaller die sizes won't always create benefits; that there the processes will "hit a wall".

The xbitlabs article quotes no sources for their speculation. The only known source for that particular speculation was the supposed engineer at a motherboard company. But nobody can really confirm the information so it is heresay. There was also supposedly a disgruntled AMD employee that "spilled the beans". But I don't take the word of disgruntled employees as being credible.

Well, I tried to look up the 22nm stuff. Apparently it was 28nm and was actually just people confused when Global Foundries was shutting down some 32nm stuff in favor or 28nm stuff for graphics cards--which will apparently be Radeon's "Souther Islands"--not Bulldozer. I think it made its way into the Wikipedia article long ago, but that misinformation is long gone.

Disgruntled people aren't credible, no--but press releases have their bias as well. Maybe a more fair summary of Bulldozer would be:
-I don't know when it's launching, but all the delays have worn out my patience.
-There's a lot of speculation that AMD needs more cores to match Sandy Bridge's performance with fewer cores because Bulldozer loses clock-for-clock, but there are no benchmarks so nobody outside of AMD really knows.
-It will be socket AM3+, continuing AMD's history of allowing CPU upgrade paths in their socket development.

 

[Super XP]

I was hoping someone actually had new information. So far all we have are the same facts from three months ago:

-Bulldozer is late and has been delayed too many times. Who knows when it will actually arrive? Delayed 2 official times, will get released in Sept to Oct 2011...
-AMD is having trouble with the manufacturing process. NO, notihng was ever wrong with the 32nm Process. The problem is Bulldozer's design is dam complex and therefore requires massive Fine Tunning. Once they fine tune the CPU, it should blow past anything Intel has to offer IMO.
-There are no performance benchmarks, but speculations is that it will lose clock-for-clock to SB so it uses more cores.
-At least AMD believes in keeping sockets so you can upgrade them. Once again, NO. Bulldozer was delayed and designed to compete with SB and beyond. They should either be equal or faster "Clock for clock".

AMD Bulldozer 2011/2012