Bulldozer, Massively unoptimized

[Steelski]

Obviously I was a little disappointed at the bulldozer architecture debut. But there were cases where the benchmarks really did not make sense, in one instance it scales, another it doesn't. Its great, its not and all that mixed stuff. Like a rollerskating CPU. And it runs hot as well as inefficient. (although one wanders what most of us are doing thats keeping out CPU's away from an idle state most of the time).
Anyway. Then there was the little increase in Windows 8 performance over at toms.
And other things that seem to be popping out of the woodwork. It led me to keep looking for reasons. (WHY AMD WHYYYYYY... I cried)
I stumbled upon the Phoronix.com website. Which tests Bulldozer under Linux.... A LOT!!!!

Even in the initial review you get the idea that Bulldozer is better under linux than Wondows.
http://www.phoronix.com/scan.php?page=article&item=amd_fx8150_bulldozer&num=1

I was thinking its not all bad... And then I stumbled upon the second article about performance using different Compilers. Being a bit of a unknown to me as to what compiling is I was surprised by the sheer amount of performance to be gained by doing this.
http://www.phoronix.com/scan.php?page=article&item=amd_bulldozer_compilers&num=2
The conclusion to the article is that Linux users will get these benefits before Windows because the kernel is always the same for windows (whatever a kernel is )
My thoughts are the following, it looks like its actually a great architecture once applications take advantage of the AVX and other extensions available. Which simply are not coded for at the moment, and when they are compiled right... Bingo.
Take Cinebench 11.5, Most people seem to take this benchmark as the defacto " how fast is the CPU benchmark" in heavy threaded apps. simply because its the one thats conveniently available. If it was optimized for AMD and Intel alike, would the Intel CPU's be showing better. C-RAY for example takes a massive performance lead for Bulldozer once its compiled with their compiler.

I am not trying to say that Bulldozer is a magic chip, but there is more to it than just "it sux". Windows 8 will bring a new kernel, other patches and chip optimization that are simply not available to Win7, New programs will certainly take advantage in time, and I'm guessing that at the end of 2012, people will be praising the chip once power usage goes down and clock speeds increase even more. I know this because I own a Phenom I processor, and its really no different to a Phenom II other than a patch I disable and I'm unable to go beyond 2.6GHZ. (I got it cheap).

What do you guys think, does the long term performance benefit look better than the one available now ( I acknowledge that Intel has a considerable lead in normal programs now)

 

[Vic Vega]

I think that, at least on this forum, there are too many people who like to hate things and very few people who actually try to understand them, good or bad. I predict you will be flamed by these people. I agree with you though. This is basically a design which our software people had not considered before because it didn't exist - we don't really know how it will perform on software written for it.

I agree with you though. It's far to early to declare this CPU a dud and frankly I'm surprised Anand would. I think he's just pandering the above mentioned people on this forum. That's fine, he knows his audience. These people want a "yes or no" answer. Thumbs up or thumbs down. It's more complicated than that.

Just watch the replies roll in as normal.

 

[Ancalagon44]

I think its a design dud for the desktop. It simply is not well suited to desktop tasks. It is better suited to server tasks, but even then, 2 billion transistors for such mediocre performance is embarrassing.

A compiler problem? I dont doubt that compilation optimizations could improve performance, but if thats the case why didnt AMD let everyone know and work on helping compiler writers optimize their code? Why didnt they just release the server version now and then release the improved desktop version next year after Windows 8 hits?

Or how about designing a CPU that doesnt require compiler tricks for it to function well? Like Intel is somehow able to do.

 

[Steelski]

I think its a design dud for the desktop. It simply is not well suited to desktop tasks. It is better suited to server tasks, but even then, 2 billion transistors for such mediocre performance is embarrassing.

A compiler problem? I dont doubt that compilation optimizations could improve performance, but if thats the case why didnt AMD let everyone know and work on helping compiler writers optimize their code? Why didnt they just release the server version now and then release the improved desktop version next year after Windows 8 hits?

Or how about designing a CPU that doesn't require compiler tricks for it to function well? Like Intel is somehow able to do.

One, the performance is simply not embarrassing when the chip performs well,
Two, AMD are working on compilers for the chip, and they have been for some time now. But its not like machine language is the simplest thing to interact with and say that its " there, thats your problem right there, that line of code" its not like calling a plumber round to look at the pipes.
Three, Compiler tricks. Because Intel designed a chip that was not really that new architecturally. And when we talk of compiler tricks..... lets not neglect the past rubbish Intel has pulled in this department.
Simply put I am not here to argue that intel has a great chip, but I am here to argue that so does AMD, but from a future perspective, and I have lined up evidence to support this.

 

[TheRyuu]

And when we talk of compiler tricks..... lets not neglect the past rubbish Intel has pulled in this department.

You can overwrite (page 134) the cpu dispatching that Intel's compiler uses so it's fair regardless of the cpu it gets run on for both the compiler and most if not all of it's supplied libraries (this has worked on all versions back to version 7 or something). It may just be because of lack of knowledge but I'm not a fan of the way Intel does the dispatch (every time the function is called) instead of just using function pointers.

 

[jmanny]

I predict you will be flamed by these people

This

 

[Don Karnage]

Server processor that has no business on the desktop

 

[sm625]

It is a dud. I've seen quite a few benchmarks with AVX and XOP and it doesnt improve performance that much. The L3 cache is basically useless. A new uArch should have produced at least 50% better performance per clock vs a Thuban. They could have easily surpassed BD by simply adding another int execution unit to Thuban and 2 more cores. They could have sewn together two thubans and still kept the die size below what it is now.

 

[podspi]

I think its a design dud for the desktop. It simply is not well suited to desktop tasks. It is better suited to server tasks, but even then, 2 billion transistors for such mediocre performance is embarrassing.

A compiler problem? I dont doubt that compilation optimizations could improve performance, but if thats the case why didnt AMD let everyone know and work on helping compiler writers optimize their code? Why didnt they just release the server version now and then release the improved desktop version next year after Windows 8 hits?

Or how about designing a CPU that doesnt require compiler tricks for it to function well? Like Intel is somehow able to do.

Bingo, from the desktop market prospective, BD just isn't that great of a product. Consider the itanium. There are lots of great things about it, but that doesn't change the fact that for home systems it just isn't a great fit. I believe that BD will do much better in the server market, where apps can be recompiled and tweaked for the architecture. In the desktop world, that just doesn't happen. It really doesn't matter what "could be".



That being said, it isn't a worthless product, and I think for people with specialized applications that can be recompiled and scale well BD could be a fantastic value, but that is something that will almost never show up in the reviews -- and that's ok because by definition not many people run specialized apps :biggrin: (And those that do can eat the cost of testing it themselves).

 

[Steelski]

Server processor that has no business on the desktop

Not being funny, But I thought this forum was above this level of conversation. Not that its rude, just really annoying. What about some EXTREME editions, are they just server chips. Of course they are, but it does not stop them topping some charts.
Mindless drivel seems to be your best contribution as it seems. Its certainly not a pure desktop chip, but then again no chip really is.
ATOM and ARM servers anyone. Sandy is more balanced for a desktop, does that mean it can not be on a server?

I don't even know why I am answering your post as it just seems to be there for pushing buttons. I'm off to chop some wood for fire..... FIRE FIRE.

 

[StrangerGuy]

very few people who actually try to understand them.

Most here do. The conclusion is BD suck balls.

 

[infoiltrator]

Sandy Bridge cost less and outperforms the previous generation, using existing software.
Whether this continues, and it may with Ivy Bridge, it is here now and operational. IS it perfect, no, but performance per dollar works.

Bulldozer is not "all that" out of the box. The previous generation is scheduled to disappear, and value is largely a maybe right now.
To date, previous generation chips are better in most cases, and power demands for overclocking high.

Just how it looks to me.

 

[LOL_Wut_Axel]

Most here do. The conclusion is BD suck balls.

QFT.

It has horrible performance/watt and costs significantly more than Intel's competition while delivering overall less performance. Even if we were to look at multi-threaded only, you're looking at currently paying $60 more for a meager 5% while still having the problem of power consumption and heat. If you were to overclock both to 4.5-4.6GHz Bulldozer's advantage in MT would be essentially erased as the FX-8150 has a 300MHz advantage stock that when both are OCed would vanish. As we all know, BD doesn't OC better than SB.

Let's not forget about the follow-ups. Ivy Bridge will bring more than 25% higher performance/watt than Sandy Bridge, a bit coming from slightly higher IPC and clock speeds, and the other part coming from the 22nm die shrink plus new 3D transistors. AMD is aiming for 10-15% higher performance/watt with Piledriver, so what are they gonna do then?

 

[mikeymikec]

Server processor that has no business on the desktop

Being pitched to the wrong audience is probably one problem with Bulldozer, but with the "Win8 handling BD better than Win7", couldn't AMD have worked with MS to do something that would help BD work better on Win7? A processor driver perhaps, or something similar to AMD's "dual-core optimizer"?

I think NVIDIA have the right idea in working with other software makers to ensure that their products work as well as possible. AMD/ATI seems to drop the ball more often than not with product support, such as tuning and bug-fixing drivers, IMO.

 

[zsdersw]

For servers it's all about performance per watt per dollar... in which Bulldozer is not at all compelling.

Do all the weaseling you want around it, with talk of optimizations and specialized scenarios, but none of that changes the PPWPD fact that Bulldozer is not good.

 

[tulx]

Most here do. The conclusion is BD suck balls.

Really? Thank you for adding value to this conversation.

I suppose some people are just happy that they can hate something and don't care about anything beyond that.

Bulldozer is a very promising CPU architecture that can only become more relevant with time - software will get more threaded, not less. I will probably not buy an FX now and will wait for a stepping or generation update, but I'm very much looking forward to what AMD can make of it. You, meanwihle, hate on!

 

[Ancalagon44]

It never ceases to me amaze me the lengths that people will go to defend something they feel loyalty to.

I have no emotions regarding Bulldozer - no hate, no malice, no disappointment. I'm not posting against it because of spite. Okay, I was angry and disappointed when it first came out, but I've had time to think about that. And now my position on Bulldozer has no emotion in it. So dont accuse us of hating it, or being unfair. We are simply stating things how they are.

And how they are is that a 2 billion transistor CPU is unable to beat a much older 900 odd million transistor design (Thuban). That is a fail in any language and in any test.

One, the performance is simply not embarrassing when the chip performs well,

Which is how often exactly?

Two, AMD are working on compilers for the chip, and they have been for some time now. But its not like machine language is the simplest thing to interact with and say that its " there, thats your problem right there, that line of code" its not like calling a plumber round to look at the pipes.

Oh I know believe me. Do you know how old Bulldozer is? Are you telling me in all those years, AMD didnt think to visit the subject of compilers? How come the AMD graphics division is able to get new drivers on time for a new graphics card release, but the AMD CPU division only bothers to think of compilers after their product gets slandered in the press? Even if you believe the negative press that the chip got was undeserved, dont you think, maybe, AMD could have done something about the compilers? Its not like its a 2 man team doing all the work. They arent as big as Intel but tell me they could afford to pay one or two guys to look at compilers.

Three, Compiler tricks. Because Intel designed a chip that was not really that new architecturally. And when we talk of compiler tricks..... lets not neglect the past rubbish Intel has pulled in this department.
Simply put I am not here to argue that intel has a great chip, but I am here to argue that so does AMD, but from a future perspective, and I have lined up evidence to support this.

A) Whether or not Sandy Bridge is "new" or not doesnt mean its not a great product. Its a flipping amazing product, and its even better for the fact you dont need any special compilation to really take advantage of it.
B) Oh yeah I know Intel is guilty of making its compiler produce sub optimal code for AMD architectures. That still doesnt excuse AMD from having done NOTHING about this. I mean there are other compilers out there, like the gcc, and the Borland compiler (if its still used?) that AMD could have worked with. And I remember installing some kind of AMD dual core patch when I got my first AMD dual core CPU - are you telling me AMD and MS could not have visited such an idea? A change to the scheduler only for BD CPUs? I'm sure MS would have done so if AMD had paid for it.

Not being funny, But I thought this forum was above this level of conversation. Not that its rude, just really annoying. What about some EXTREME editions, are they just server chips. Of course they are, but it does not stop them topping some charts.
Mindless drivel seems to be your best contribution as it seems. Its certainly not a pure desktop chip, but then again no chip really is.
ATOM and ARM servers anyone. Sandy is more balanced for a desktop, does that mean it can not be on a server?

I don't even know why I am answering your post as it just seems to be there for pushing buttons. I'm off to chop some wood for fire..... FIRE FIRE.

You're really taking this too seriously and too emotionally. He said its a server chip and I completely agree. Why dont you read the ArsTechnica article on the subject for a more detailed analysis of the design tradeoffs that AMD made in BD, instead of getting all offended when we just mention facts.

Really, take a chill pill, this is a processor which no one is forcing you to buy, and unless you are an AMD shareholder, it really should not matter to you if someone doesnt like it.

 

[exar333]

Server processor that has no business on the desktop

It sucks as a server chip too, what's your point? Intel markets pretty-much the same chips (on the low-end) between enterprise and consumers. A quad-core Xeon is pretty much the same as the equivalent i7 offering.

Performance/watt is where server value lies, BD definitely does not have this.

BD is fail, plain and simple. As stated in so many other threads, so saying otherwise is really ignorant.

 

[gorydetails]

Haters gona hate,,,

 

[HopJokey]

It sucks as a server chip too, what's your point? Intel markets pretty-much the same chips (on the low-end) between enterprise and consumers. A quad-coreb Xeon is pretty much the same as the equivalent i7 offering.

Performance/watt is where server value lies, BD definitely does not have this.

BD is fail, plain and simple. As stated in so many other threads, so say otherwise is really ignorant.

QFT. Someone show me where BD is a "win" in the server space. It may just be "less fail" than in the desktop segment.

 

[StrangerGuy]

Really? Thank you for adding value to this conversation.

I suppose some people are just happy that they can hate something and don't care about anything beyond that.

Bulldozer is a very promising CPU architecture that can only become more relevant with time - software will get more threaded, not less. I will probably not buy an FX now and will wait for a stepping or generation update, but I'm very much looking forward to what AMD can make of it. You, meanwihle, hate on!

Until then, BD still sucks. So your point is?

AMD fanboys always use the same old lame arguments.

 

[LOL_Wut_Axel]

It sucks as a server chip too, what's your point? Intel markets pretty-much the same chips (on the low-end) between enterprise and consumers. A quad-coreb Xeon is pretty much the same as the equivalent i7 offering.

Performance/watt is where server value lies, BD definitely does not have this.

BD is fail, plain and simple. As stated in so many other threads, so say otherwise is really ignorant.

QFT. Someone show me where BD is a "win" in the server space. It may just be "less fail" than in the desktop segment.

Remember this?

Definitely doesn't look good for servers, either. The upcoming 8-core/16-thread LGA 2011 server CPUs would probably be comparable in those types of workloads to the 16-core Interlagos while consuming a huge amount less power. They'd cost more upfront, but they'd easily recoup those costs in very little time in electricity bills because of both lower power consumption and lower cooling requirements. They'd pay for themselves.

 

[Ferzerp]

Please stop saying it's "a server processor". It has no value in that space either. The workloads aren't suddenly magically something it is better suited to than the intel competition that it is a generation or so behind.

 

[PowerYoga]

Server farms are all about energy efficiency. Imagine you have a whole farm of bulldozers vs xeons or sandybridges. Eeeeesh. You'd lose hundreds of thousands a year for less performance. It makes absolutely no business sense to switch to bulldozer.

 

[tulx]

Until then, BD still sucks. So your point is?

AMD fanboys always use the same old lame arguments.

Last two PC's I've build have been Intel.
My point is that it's often worth looking at things from a wider perspective and not limit ones opinion to "OLOLOLOL FAIL". The value of Bulldozer architecture will only be apparent once it has matured - like with Phenom. Agena definitely wasn't a good start, as isn't FX, yet Deneb made the Phenom the a very, very popular CPU series, which is why I'm excited about Pilediver.

The 1st plane flew only a couple of meters - EPIC FAIL, right?