Fudzilla: Bulldozer performance figures are in

[ydnas7]

http://obrovsky.blogspot.com/2011/07/you-were-punkd.html

OBR fooled everyone.

'i can post here anything: emails ... differently edited pictures, and more and more proofs, ive manipulated scores and sent them to DH ... STOP dreaming kids, Bulldozer is CRAP!

OF Course, all results was FAKEs, i know it, ive faked them No real numbers before NDA

PS. Overclocking is not FAKE! Only results!'

 

[LOL_Wut_Axel]

Why is this thread still alive? All these benchmarks are either useless or made up crap. Wait for the official reviews on August or September.

 

[bridito]

From a business standpoint yes, why bother spending large on R&D for high end chips when you are so far behind. Would make sense to aim for the lower end only parts.

From a enthusiast standpoint i hope AMD breaks back into the Entusiast market so Intel doesnt have a monopoly on it and has reason to push there R&D.

Right now intel has no competetion, they have no push to advance there hardware. If AMD had a high end chip out beating a 2600k you can bet Intel would have pushed higher end SB 1155 chips out by now(we all know intel has tons of clockspeed they arnt using) and probably launched SB-E by now. The way it is now they spend as little as possible on R&D and wait for AMD to almost catch up before launching anyhting new, not a ideal situation for us enthusiast buyers.

So in other words, welcome to the monopoly. CPU progress is being held back by AMD's inability to compete with a rival 27x its size. It's not the first time that monopolies have proven counterproductive and I guess CPUs won't be the last. It's almost a shame that AMD continues to be around as otherwise some judicious anti-trust action could break up the monopoly into Baby Intels and we might actually see some real competition in this extremely critical market.

No, because right now for both AMD and Intel it is simply an extension to the low end server/workstation market, in effect.

Yes, but only to a point, since the GHz sweepstakes are primarily enthusiast-focused while the multicore-thread is a server advantage.

http://obrovsky.blogspot.com/2011/07/you-were-punkd.html

OBR fooled everyone.

What a #$&%ing moron. I've been saying it all along that there were NO benchys and that if it can't be verified, it doesn't exist.

They might as well have just stuck up a large sign saying "Bulldozer is slower then sandy bridge". If you had the faster chip then when you brought it to a major geek event you'd be praising it to the rafters with lots of real time testing showing how much faster it was then the opposition.

[H] was such a foreseeable debacle, I'm honestly surprised that AMD even let them have the CPU! This is what happens when you have beheaded companies making decisions! No accountability to the top exec and people start doing stupid things. [H] has caused damage to AMD. Maybe only from a handful of various forum enthusiasts, but damage nonetheless.

Why is this thread still alive? All these benchmarks are either useless or made up crap. Wait for the official reviews on August or September.

I think that threads like this serve a purpose. We know that various execs from both AMD and Intel read these comments (or at least some of them). So it's a way for enthusiasts to express their serious frustration at what seem to be AMD's ham handed inability to get BD to perform at an "expected" level and Intel from sitting on its monopolistic laurels and pushing back Moore's Law. CPU development doesn't just benefit powerusers, it boosts all of computing and its applications which are spread throughout every facet of science, research, engineering, etc.

Monopolies are BAD. AMD seems to be well on the way to guaranteeing us an effective monopoly if one does not exist already. Sure, AMD can launch a killer BD with phenomenal power and speed... it can still happen, although I realistically have to say that it's continuing to get more unlikely by the day!

 

[ydnas7]

i used to use an ARM based desktop 20years ago (Archimedes) you are welcome to use one of a hundred different ARM manufacturers today if you want, I' prefer X86-64 myself

 

[LOL_Wut_Axel]

[H] was such a foreseeable debacle, I'm honestly surprised that AMD even let them have the CPU! This is what happens when you have beheaded companies making decisions! No accountability to the top exec and people start doing stupid things. [H] has caused damage to AMD. Maybe only from a handful of various forum enthusiasts, but damage nonetheless.

Um, what about [H] harmed AMD?



I think that threads like this serve a purpose. We know that various execs from both AMD and Intel read these comments (or at least some of them). So it's a way for enthusiasts to express their serious frustration at what seem to be AMD's ham handed inability to get BD to perform at an "expected" level and Intel from sitting on its monopolistic laurels and pushing back Moore's Law. CPU development doesn't just benefit powerusers, it boosts all of computing and its applications which are spread throughout every facet of science, research, engineering, etc.

Monopolies are BAD. AMD seems to be well on the way to guaranteeing us an effective monopoly if one does not exist already. Sure, AMD can launch a killer BD with phenomenal power and speed... it can still happen, although I realistically have to say that it's continuing to get more unlikely by the day!

We did that already in June. I don't know why these threads keep popping up when we already divulged our concerns and anger about Bulldozer being delayed.

As for Intel, yes, they do have a monopoly on the Performance and Enthusiast market for now, but those are a lot less relevant to consumers, especially Enthusiast. Bulldozer is a Performance CPU because AMD probably doesn't want to spend the little liquid assets it has on R&D for an Enthusiast CPU that will not sell in volume.

 

[Phynaz]

It has 4 FlexFP's that can each execute 2 128 bit instructions, or 1 256 bit instruction.

It is not half of anything on the intel side.

I refer you to the AMD provided slide at the bottom of this page.

http://www.anandtech.com/show/3863/amd-discloses-bobcat-bulldozer-architectures-at-hot-chips-2010/4

Note the one FPU shared amongst two integer cores.

 

[bridito]

We did that already in June. I don't know why these threads keep popping up when we already divulged our concerns and anger about Bulldozer being delayed.

As for Intel, yes, they do have a monopoly on the Performance and Enthusiast market for now, but those are a lot less relevant to consumers, especially Enthusiast. Bulldozer is a Performance CPU because AMD probably doesn't want to spend the little liquid assets it has on R&D for an Enthusiast CPU that will not sell in volume.

There were a goodly number of people who were really looking forward to "something anything" at [H]. When they got twaddle, then it's completely normal for them (and me) to want to vent. As long as we're not violating forum procedure, I see no reason to censor this frustration.

As for your statement that AMD might not want to compete in the top end, then that is an assertion that we do have a monopoly on CPU progress. Pushing the envelope is one thing, filling in the background is quite another.

 

[podspi]

I can't remember a time when Intel was more on the defensive than it is now.

Even at the height of AMD's achievement in the Hammer days, Intel was not bothered by AMD. Sure, they had lost on the forums, but AMD never had the capacity or OEM support to threaten Intel.


Now, not only has ARM seized the low-power market (the OEMs there are not friendly to x86) but it is threatening to move upwards into servers with the emerging popularity of cloud-computing.



Moving back to their traditional markets, we've seen more OEM support for AMD now than ever before, and if Fusion actually catches on, could be a game-changer. Llano is actually a very powerful chip, unless you are running SuperPI

 

[Edrick]

Llano is actually a very powerful chip, unless you are running SuperPI or anything CPU intensive.

Fixed.

I would not want to do my development on it, or photoshop for that matter.

 

[Ares1214]

I refer you to the AMD provided slide at the bottom of this page.

http://www.anandtech.com/show/3863/amd-discloses-bobcat-bulldozer-architectures-at-hot-chips-2010/4

Note the one FPU shared amongst two integer cores.

1 AMD Flex FPU can do 2 128bit, 4 64bit, 8 32bit and so on. 1 Intel FPU can do 1 128bit, and scale down so on. The only place where AMD is "half" of Intel is 256bit AVX, which is basically not seen. And even then its 4 vs 4.

 

[bridito]

Fixed.

I would not want to do my development on it, or photoshop for that matter.

Photoshop on a Llano. Now that's funny. I can really see me doing that too... Nah... Might as well get the netbook out and do it on that blazing Atom N450 with that rip roaring 1GB RAM.

Intel is on the defensive? Maybe, maybe not. If it is it's surely on the low and mid range end and all of that is outside of the area of my slightest concern. I couldn't give a good rodent's booty about what AMD or Intel do or don't do as long as one of them gives me a rocketship to plunk in my next system! :thumbsup:

 

[Phynaz]

1 AMD Flex FPU can do 2 128bit, 4 64bit, 8 32bit and so on. 1 Intel FPU can do 1 128bit, and scale down so on. The only place where AMD is "half" of Intel is 256bit AVX, which is basically not seen. And even then its 4 vs 4.

Look at the slide again.

According to AMD, Intel does 2x128 bit per core, while AMD does 2x128 bit per two cores.

That's why there is all the talk about the FPU being shared.

Edit:
Or read here: http://www.realworldtech.com/page.cfm?ArticleID=RWT082610181333&p=2

Each module or compute unit (i.e. a pair of cores) share an L1I cache, floating point unit (FPU) and L2 cache

 

[Arzachel]

This thread is identical to this http://forums.anandtech.com/showthread.php?t=2119332&highlight=cayman thread, just swap Cayman and Fermi for Bulldozer and Sandy Bridge, Nvidia with Intel.

 

[drizek]

Look at the slide again.

According to AMD, Intel does 2x128 bit per core, while AMD does 2x128 bit per two cores.

That's why there is all the talk about the FPU being shared.

Edit:
Or read here: http://www.realworldtech.com/page.cfm?ArticleID=RWT082610181333&p=2

AMD has 8 cores but Intel only has 4.

 

[hooflung]

Fixed.

I would not want to do my development on it, or photoshop for that matter.

That is a bit baseless unless you tell us what you develop for. It can easily run Eclipse for J2EE on a host OS while providing guest OS's for deployment server testing instances. All while running guests for running another guest for a clean room or specific Eclipse install ( or embedded IDE's that only like win32 ). Not to mention it can be a chip 16GB ram machine and cooled sufficiently with a H50.

My PII 810 runs a similar setup and the 2.9ghz Llano would be even faster. Ofc, I probably won't buy one because I just invested in upgrading my old 940 on a 780i mobo to a PII 1075 and raid 5. Two PC upgrades to EOL hardware ain't gonna happen.

PS: I do exactly what I said you could do above with my X4 and X6's.

 

[Edrick]

That is a bit baseless unless you tell us what you develop for.

Fair enough.

I am a .NET/SQL developer for a large financial institution. Some of our jobs take hours to run due to the large amounts of data we need to manipulate. To even compile some of our solutions take up to 15 minutes on Nehalem Xeon workstations.

 

[hooflung]

Could always ask for some sort of continuous integration solution for that instead of doing it all locally?

 

[LOL_Wut_Axel]

There were a goodly number of people who were really looking forward to "something anything" at [H]. When they got twaddle, then it's completely normal for them (and me) to want to vent. As long as we're not violating forum procedure, I see no reason to censor this frustration.

Yeah, but this thread isn't about [H] and their hype about a Bulldozer meeting/event and then not meeting expectations. Not that I know of, anyway.

As for your statement that AMD might not want to compete in the top end, then that is an assertion that we do have a monopoly on CPU progress. Pushing the envelope is one thing, filling in the background is quite another.

And to be honest, I couldn't really care less if they control the Enthusiast market. The Performance market, which is already much smaller than Mainstream, will give them more profits and more affordable (under $350) CPUs. Very few people, even here, want to purchase extremely expensive platforms with features almost no one will use and CPUs costing upwards of $500. As long as they're very competitive in the Performance market, I'm happy.

I can't remember a time when Intel was more on the defensive than it is now.

Even at the height of AMD's achievement in the Hammer days, Intel was not bothered by AMD. Sure, they had lost on the forums, but AMD never had the capacity or OEM support to threaten Intel.


Now, not only has ARM seized the low-power market (the OEMs there are not friendly to x86) but it is threatening to move upwards into servers with the emerging popularity of cloud-computing.



Moving back to their traditional markets, we've seen more OEM support for AMD now than ever before, and if Fusion actually catches on, could be a game-changer. Llano is actually a very powerful chip, unless you are running SuperPI

It already caught on, I think. Ontario and Zacate are selling in the millions and are taking away valuable market share from Atom. Llano has a lot of design wins with many OEMs and almost all of the main ones have announced new laptops based on it.

Fixed.

I would not want to do my development on it, or photoshop for that matter.

Why are you complaining when you're not the intended market? If your job depends on it you're supposed to buy a CPU in the Performance market, not the Mainstream. Also, Llano (the desktop version, that is) is very competitive with Core i3 on multi-threaded apps. If you're talking the mobile version then that's another story since Llano is around 20% slower in multi-threaded apps. With both versions, though, one thing you do get is 2x higher GPU performance.

 

[PreferLinux]

Yeah, but this thread isn't about [H] and their hype about a Bulldozer meeting/event and then not meeting expectations. Not that I know of, anyway.



And to be honest, I couldn't really care less if they control the Enthusiast market. The Performance market, which is already much smaller than Mainstream, will give them more profits and more affordable (under $350) CPUs. Very few people, even here, want to purchase extremely expensive platforms with features almost no one will use and CPUs costing upwards of $500. As long as they're very competitive in the Performance market, I'm happy.



It already caught on, I think. Ontario and Zacate are selling in the millions and are taking away valuable market share from Atom. Llano has a lot of design wins with many OEMs and almost all of the main ones have announced new laptops based on it.



Why are you complaining when you're not the intended market? If your job depends on it you're supposed to buy a CPU in the Performance market, not the Mainstream. Also, Llano (the desktop version, that is) is very competitive with Core i3 on multi-threaded apps. If you're talking the mobile version then that's another story since Llano is around 20% slower in multi-threaded apps. With both versions, though, one thing you do get is 2x higher GPU performance.

Yes, it should be: 4 cores vs. 2 cores with HT.

 

[GaiaHunter]

When they announced that BD would be at [H]/AMD event what I was interested was the [H] staff reaction to it. Some has finally arrived to their forums.

http://hardforum.com/showthread.php?t=1623007&page=3

It played the games smoothly. Then again that doesn't tell us much because you can do that on a Phenom II X4 just about as well as Sandy Bridge for the most part. The times where you will see a difference are few and far between outside of simple benchmark runs. The silicon is certainly stable enough for general usage based on what I saw but there could still be errata that needs to be addressed before release. In fact I'd bet that the reason for Bulldozer's delay is either due to errata which the B2 silicon should resolve or there could be manufacturing issues which still need to be ironed out.

I think the inclusion of Zambezi parts in the gaming machines at the event is to let people know that the release isn't too far off and you'll get decent performance out of it when it does arrive. It's really just to whet the apetite for it and little more. The reality of it is, those could have been Phenom II X6 machines or (with windowless cases) Core i7 990X's. We'd never have known the difference the way they were being used.

They did not have a comparison setup going between the Phenom II X6, Intel's Core i7 or AMD's Zambezi processors. They had Phenom II X6 and Zambezi based machines, but not necessarily doing the same things.

As for what I said regarding it's performance, that's about all that can be said. There are instances where one might notice the difference between an AMD Phenom II X6 and a Core i7 2600K machine clocked as high as each can reasonably go, but those scenarios are few and far between. In fact really, until you get into multi-GPU and high resolution multi-monitor gaming, there really isn't a difference that can be seen outside of benchmarks.

For some people this will be no information, for others it will tell something.

There has been some talk about AMD having problem with the cache, something that seems to have quite an impact in games. The fact it ran the games smoothly it is a strong indication those particular samples didn't have that problem.

The fact that the system was stable is also a good indication.

About benchmarks, I doubt very much AMD will release any for the general consumer before the official release - it hasn't been there MO for their last product releases, including the very successful ones.

 

[Phynaz]

AMD has 8 cores but Intel only has 4.

Ummm...What? There are (will be) 4, 6 and 8 core Bulldozer CPUs. And you can buy 4, 6, 8 and 10 core Intel CPUs today.

None of which has anything to do with the discussion of shared FPUs.

 

[LOL_Wut_Axel]

Yes, it should be: 4 cores vs. 2 cores with HT.

Except AMD is using a slighly tweaked architecture from 2009 while Intel is using an architecture that's two years newer. Less impressive to me is the fact that while the Core i3 has two fewer cores and a lot less powerful IGP it only consumes a bit less power even though they're on the same process node.

 

[nonameo]

Ummm...What? There are (will be) 4, 6 and 8 core Bulldozer CPUs. And you can buy 4, 6, 8 and 10 core Intel CPUs today.

None of which has anything to do with the discussion of shared FPUs.

In the end, it's all about what you can get for your money.

 

[Edrick]

Why are you complaining when you're not the intended market? If your job depends on it you're supposed to buy a CPU in the Performance market, not the Mainstream. Also, Llano (the desktop version, that is) is very competitive with Core i3 on multi-threaded apps. If you're talking the mobile version then that's another story since Llano is around 20% slower in multi-threaded apps. With both versions, though, one thing you do get is 2x higher GPU performance.

I am not complaining at all. I am simply making a counter point to someone saying llano is a very powerful chip for anything other than SuperPI. There is no arguement it has a great GPU, and it great for the mobile market, but to call it a great cpu performer is a little extreme in my opinion.

 

[LOL_Wut_Axel]

I am not complaining at all. I am simply making a counter point to someone saying llano is a very powerful chip for anything other than SuperPI. There is no arguement it has a great GPU, and it great for the mobile market, but to call it a great cpu performer is a little extreme in my opinion.

He said it was a very fast chip. And as an overall chip, it definitely is very fast. You wouldn't want to run extensive Photoshoping on either the Core i3 or the A8.

With almost everything apart from audio encoding going multi-threaded, it's definitely not bad in the CPU performance sector. From the reviews what I could conclude overall is that CPU performance overall is 10-15% lower comparing the 2105 and 3850 because of audio encoding and it being the same speed in multi-threaded ones. As for laptops, there's definitely a bigger discrepancy. We're looking then at a difference in CPU overall of 35-40%.

What I want to get at: it's definitely not an extremely fast CPU, but neither is the Core i3, even if it's 10-40% faster. The IGP is definitely something to be impressed with, being only 5-10% slower than a Radeon HD 5570. Overall I think it makes for a better tradeoff and overall package.