Why all the AMD hate, guys?

[werxen]

Let us stop the fanboyism and acknowledge AMD's market strategy - the AVERAGE consumer not troubled by synthetic benchmark scores or maxing our Crysis. I see too much hate going on from both camps - let us make love not war. But in reality anyone who 'hates' AMD needs a friggin reality check. Intel will not be on top forever - anyone remember the Athlon series? People were sure AMD would lead the way forever after that.

 

[yannifb]

Why spend a $1000 on a CPU, when you can spend $1000 on a computing card that is much, much faster? There's lots of software that take advantage of them now. I mean seriously if you have ever run particle collision simulations (im using this as an example since its the only heavy calculations i ever really do) they run so much faster, and even can shorten the time from a day to an hour or two.

 

[cjl]

Because Solidworks flow simulations don't take advantage of any of that kind of thing, as far as I can tell. What kind of computing card are you thinking of? Tesla?

 

[Chad Boga]

And I thought I was the only one who thought that way, or talked that way.

Lots of people who stop taking their medication think and talk that way.

 

[sighQ2]

Because Solidworks flow simulations don't take advantage of any of that kind of thing, as far as I can tell. What kind of computing card are you thinking of? Tesla?

Well, ok - so that would be a good benchmark perhaps. I don't know the software. But I do know that a lot of the benchmarketing benchmarks are seriously biased due to spintel compilers, instruction sets - indeed there is ongoing controversy with the instrux sets right now - see the FTC complaints. Perhaps your software is another spintel compiled thing, designed to kill the competition - the impairment is not insignificant either. It would be interesting to know if recompiling would give opposite results.

Perhaps, your software has particular demands - I really have no clue - but I doubt that the zillion gamer freaks here will really care about it.

I have heard that kind of "it takes days" talk re movie special fx rendering also. I guess those folks used to use all AMD, then couple years ago switched to BRAND X. Now I hear some are switching back to AMD.

Basically you are presenting a niche market - perhaps of some size - I don't know.

My complaints are mainly about lying cheating benchmarketing tactics using compilers that cripple the competition, and guarantee spintel wins. It's all part of the dirt that's ongoingly being revealed in the antitrust arena. I speak of known issues, known grievances; and much of it is old news. But Verndewd's idiot masses can be pretty slow to wake up and smell the coffee.

But, I think, as yannif (I think) is saying, the gpgpu scene is the way of the future - we really don't need spintel except for their precious licence. and the sooner they learn that their dirt is not acceptable to the human race, the better. We could be waiting for a cold snap in July. Beware the dark side of the force.

 

[LoneWolf_53]

I'm beginning to suspect that global warming may well have something to do with this thread. 😀


As a distributed computing cruncher participant for over ten years now I can assure anyone that I don't pay too much attention to benchmarks and such but simply look at the results my equipment puts out whilst crunching.

Now when you factor in the reality that many of the really avid crunching addicts who always run the latest and greatest hardware are still sitting in the Intel camp it should indicate something don't you think?

Our machines are running 24/7/365 under full CPU (and in many cases now GPU) load so imagine running your equipment with Prime 95 on all the time.

Fact is so far the Intel's are holding up and producing and the numbers prove it.

When the tide changes and probably it will then the power users in the crunching world will shift back to using AMD as their preferred processor.

 

[Chad Boga]

This compiler nonsense is just the latest excuse the loonies from AMDZone are hanging their hats on to explain away AMD's lesser performance.

 

[BadTrip]

I really have no clue.

I couldn't agree more

 

[sighQ2]

Hmmm....
That's interesting, since I know a friend with an AMD Phenom II 955, and my computer is a LOT faster for solidworks flow simulations (which I run all the time) than his. I would guess ~30%. This is on simulations that can take several days too, so it's pretty significant (I've got one running right now that's been going for 42 hours, and has another 16 or so to go). I'd say that either I'm hallucinating, or my inferior spintel crap is actually quite a bit faster in real world computational tasks.

Of course, AMD is pretty close for gaming. That's because games don't use as much multithreaded performance, and in general don't take advantage of faster CPUs as much as they do faster GPUs.

Well, I really don't know, as I said.

According to your sig, you run i7 965 at 4.2 ghz.
So a PhII 955 runs stock at 3.2 (?)
do the math...
33% overclock on 3.2 Ghz = 4266

So, assuming same clockspeeds, the 2 cpu's are identical performance; give or take a few minutes on your software.

Is you friend's Phenom II oclox'd at 4.2 Ghz??? = that would be a good test = can you try that? It might take a C3-stepping 965 on good cooling - but if what you say is true, you paid too much for that i7.

You could go twice as fast with 2 965 systems for about the same cost. (yeh, I know, that's pretty rough figures).

And the GPGPU factor, well, that's AMD's thing too isn't it.?

Cost of watercooling still probably comes in cheaper.

Sorry I don't know details of your rig. But I think you paid too much.

It would be nice if the idiots would shutup long enough for there to be an intelligent discussion of this concept - hey, why not?

Oh yeh,

do the math on a 965 @3400 Ghz x 23% oclox = 4182 Ghz. LOTS of people have done this easily on watercooling.

So yes, perhaps a 965 would do it.

of course, if you believe in hyperthread being that great..... BTW, do you
- run Hyperthread on/off?
- run Turbo on/off?

It would be a great contest, dontcha think? Kinda makes me wonder.

The IPC is the big factor.

Now the Real Test would be to ReCompile first.
hahaha = you could report results to the FTC.

This is idea is all way over my head; it's just an idea based on what I have learned generally.

I don't trust anything spintel says. Can you tell? How much of an impairment is involved in the compiler? If the IPC really that different on a level playing field?

Can someone actually do this test?

 

[Chad Boga]

This is idea is all way over my head

Obviously

 

[JAYDEEJOHN]

You know many people said the same thing I say: Intels IGPs suck for gaming but still work for casual gaming.

And I am sorry but I have little faith in Fusion doing anything beyond removing the IGP from the mobo to the CPU. Even with Fusion or whatever Intel gets out to compete I gurantee that a discrete low end GPU will still be the better performer.

So youre saying that todays IGPs with lower clocks wont be bested by ones using 2 process changes, HKMG, higher clocks and 320 shaders?
Are you just slamming ATI here, or AMD?
You must think very little of ATIs abilities if you really believe this.
Or, is it more to the slam AMD at all costs thing?
You should sell your crap ATI card then, and dont look forwards to anything better, necause AMD bought them up, if thats the excuse you use.
As for casual gaming, what the people saying things like that dont understand is, this isnt good enough, and leaves the gap in gaming too wide for dev down the road, which is what Ive been saying all along, and have welcomed this "not gonna happen, not needed, they sell the most of both so why bother?" IGP thats coming, despite such successes, so again, why bother, especially, as you say, fusion wont make a difference anyways?

 

[Upendra09]

Ugh Upendra there has been something just itching at me for a long time now about your member config. It just says 'HP 2.0 GHz'... what processor is that supposed to be???

it says intel now

Hmmm....
That's interesting, since I know a friend with an AMD Phenom II 955, and my computer is a LOT faster for solidworks flow simulations (which I run all the time) than his. I would guess ~30%. This is on simulations that can take several days too, so it's pretty significant (I've got one running right now that's been going for 42 hours, and has another 16 or so to go). I'd say that either I'm hallucinating, or my inferior spintel crap is actually quite a bit faster in real world computational tasks.

Of course, AMD is pretty close for gaming. That's because games don't use as much multithreaded performance, and in general don't take advantage of faster CPUs as much as they do faster GPUs.

Wherem do you guys work? i have never seen so many different apps, and non that use more than two hours of processing, i mean 3 days?!?!

I couldn't agree more

Why do u always take people's words out of context? If you're going to insult someone, atleast do it sensibly and properly don't take cheap shots at them

 

[amdfangirl]

In other words: Worship Intel or DIE 😀

Another favourable outlook.

 

[sighQ2]

Upendra
I doubt he comprehends "out of context". It's probably within his context entirely. If I didn't have something to say, he wouldn't have anything to say. We could be saving his life pending a realization on his part that he's a complete dolt. I doubt he knows what that means either. He next comment will be an indicator. plz stand by.

 

[sighQ2]

There are serious commercial niches where cost is irrelevent cos time is money. These are serious professional people who don't make decisions lightly based on brand loyalty or politics.

As was explained to me, in Hollywood special FX rendering, a few seconds per frame adds up to days or hours if one is processing thousands of frames. Deadline due date rules - it's get it done on time. Speed is everything.

It's also possible they have done the type of test I mentioned, on their own proprietary software, using their own compilers - testing 2 systems side by side - before they bought their particular solution.

Don't expect these people to reveal who they work for, or even what software they use; it's probably confidential info. (NDA)

I would just like to hear general details about how they decided on what they bought; and if they considered all possibilities.

 

[Upendra09]

one thingAMD did good on is the fact that they add HD to their gfx card names, so people think it will display really good HD vids and pics

 

[cjl]

Well, I really don't know, as I said.

According to your sig, you run i7 965 at 4.2 ghz.
So a PhII 955 runs stock at 3.2 (?)
do the math...
33% overclock on 3.2 Ghz = 4266

So, assuming same clockspeeds, the 2 cpu's are identical performance; give or take a few minutes on your software.

Is you friend's Phenom II oclox'd at 4.2 Ghz??? = that would be a good test = can you try that? It might take a C3-stepping 965 on good cooling - but if what you say is true, you paid too much for that i7.

You could go twice as fast with 2 965 systems for about the same cost. (yeh, I know, that's pretty rough figures).

And the GPGPU factor, well, that's AMD's thing too isn't it.?

Cost of watercooling still probably comes in cheaper.

Sorry I don't know details of your rig. But I think you paid too much.

It would be nice if the idiots would shutup long enough for there to be an intelligent discussion of this concept - hey, why not?

Oh yeh,

do the math on a 965 @3400 Ghz x 23% oclox = 4182 Ghz. LOTS of people have done this easily on watercooling.

So yes, perhaps a 965 would do it.

of course, if you believe in hyperthread being that great..... BTW, do you
- run Hyperthread on/off?
- run Turbo on/off?

It would be a great contest, dontcha think? Kinda makes me wonder.

The IPC is the big factor.

Now the Real Test would be to ReCompile first.
hahaha = you could report results to the FTC.

This is idea is all way over my head; it's just an idea based on what I have learned generally.

I don't trust anything spintel says. Can you tell? How much of an impairment is involved in the compiler? If the IPC really that different on a level playing field?

Can someone actually do this test?

That comparison was done stock vs stock, turbo and HT on (the flow simulations will happily use all 8 threads). So, it was (realistically) a 3.33GHz i7 vs 3.2GHz PhII. The Phenom certainly didn't to badly - it was still faster than the core 2 workstations that we have in the school labs, but it can't compete with an i7 for CFD at least, which was a big factor in my purchasing decision. As for 2 965 systems? It depends. Sometimes, I don't care about running several simulations, I just want to get one done as quickly as possible. I seriously considered a dual socket system when building, but I wanted something to handle both gaming and serious work, and dual sockets aren't really suited for gaming.

 

[JAYDEEJOHN]

Heres an interesting outlook on Intels abilities concerning gpus, LRB and netbooks:

GPU manufacturers, such as Nvidia and AMD, through its affiliate, ATI, are developing General Purpose GPUs and programming interfaces that
4
threaten Intel’s control over the computing platform. This General Purpose GPU computing (“GP GPU”) platform has the potential to marginalize Intel’s long-standing CPU-centric, x86-based strategy. Currently, both high-performance computing and mainstream applications and operating systems are beginning to adopt GP GPU computing functionality.
17.
GPUs also could facilitate new entry or expansion in the relevant CPU markets by other firms, such as Nvidia, AMD, or Via. The need for high-end microprocessors may be reduced as more computing tasks are handled by the GPU. Some OEMs could get equivalent performance at a cheaper cost by using a lower-end CPU with a GPU microprocessor.
18.
As it did in the CPU markets, Intel recognized the threat posed by GPUs and GP GPU computing and its technological inferiority in these markets and has taken a number of anticompetitive measures to combat it. These tactics include, among others, deception relating to competitors’ efforts to enable their GPUs to interoperate with Intel’s newest CPUs; adopting a new policy of denying interoperability for certain competitive GPUs; establishing various barriers to interoperability; degrading certain connections between GPUs and CPUs; making misleading statements to industry participants about the readiness of Intel’s GPUs; and unlawful bundling or tying of Intel’s GPUs with its CPUs resulting in below-cost pricing of relevant products.
http://www.ftc.gov/os/adjpro/d9341/091216intelcmpt.pdf

Of course, LRB is almost here, coming soon, IGPs have no effect on the markets, and lastly, Intel should charge more for the Atom chip seperately, compared to a SoC setup. Oh and Vista? We dont need no stinking Vista for our IGPs

Now, dont come after me for staing this, as this is the FTCs position regarding Intel

 

[cjl]

Wherem do you guys work? i have never seen so many different apps, and non that use more than two hours of processing, i mean 3 days?!?!

Well, I'm doing school stuff (I'm a junior in Aerospace engineering), and the stuff that takes a LONG time is the computational fluid dynamics. Subsonic simulations take 2-6 hours typically, and supersonic simulations take 12-60 hours.

 

[notty22]

Its getting redundant to keep reading misinformed people dredge on about compilers.
http://www.eetimes.com/showArticle.jhtml?articleID=222002499&pgno=2
Quoted:
"And any and all of the competition had the option to write compilers that make use of any quirks of their implementations of the x86 architecture — number of cycles to access registers, the order in which registers are polled, number of pipeline stages, byte ordering and so on."
But if benchmark organizations measure your products as performing well, would you choose to suppress the information?

What the FTC is presenting as unfair is Intel using supposed third party benchmark results, in their ADVERTISING. Since AMD does not make a compiler and some of those products were compiled with Intel code, should Intel have disclosed this in advertising,
come on, people.

 

[JAYDEEJOHN]

To me, the complier issue is exactly like the Physx issue, where it looks first, acts second, and you get what its meant to do, or, if not Intel, shut it off, or if not nVidia, shut it off, same thing.
I do believe Intel is alot like nVidia in alot of ways

Now you ask, why should Intel be required to do this? Because they already do it anyways, for their newer cpus:

In other words, they claim that they are optimizing for specific processor models rather than for specific instruction sets. If true, this gives Intel an argument for not supporting AMD processors properly. But it also means that all software developers who use an Intel compiler have to recompile their code and distribute new versions to their customers every time a new Intel processor appears on the market. Now, this was three years ago. What happens if I try to run a program compiled with an old version of Intel's compiler on the newest Intel processors? You guessed it: It still runs the optimal code path. But the reason is more difficult to guess: Intel have manipulated the CPUID family numbers on new processors in such a way that they appear as known models to older Intel software. I have described the technical details elsewhere.
http://www.agner.org/optimize/blog/read.php?i=49

 

[sighQ2]

That comparison was done stock vs stock, turbo and HT on (the flow simulations will happily use all 8 threads). So, it was (realistically) a 3.33GHz i7 vs 3.2GHz PhII. The Phenom certainly didn't to badly - it was still faster than the core 2 workstations that we have in the school labs, but it can't compete with an i7 for CFD at least, which was a big factor in my purchasing decision. As for 2 965 systems? It depends. Sometimes, I don't care about running several simulations, I just want to get one done as quickly as possible. I seriously considered a dual socket system when building, but I wanted something to handle both gaming and serious work, and dual sockets aren't really suited for gaming.

HEY
Thank you for this info. That's a hefty commitment for schooling.

Was any (RE)compiling done as part of your tests or in normal useage?

Thanks again.

 

[JAYDEEJOHN]

This would all be solved by AMD telling Intel these models support this instruction set, so adding ID numbers is the extent of Intels resposibility here, along with AMDs and VIAs cooperation, which only benefits them, and would happen, and Im thinking this is what we will end up with, per the FTCs ruling.
So, maybe that AMD cpu isnt quite as bad as it looks, but again, who knows, it may effect a few apps and such, or all, or even none, but the law is the law

 

[cjl]

HEY
Thank you for this info. That's a hefty commitment for schooling.

Was any (RE)compiling done as part of your tests or in normal useage?

Thanks again.

The code isn't open source, and I don't have access to it, so no recompiling was done. I don't know if any Intel-specific optimizations were included or not.

As for it being a hefty commitment for schooling? That's true, but since I'm paying ~25k per year anyways (counting books, housing, etc), a $3500 system (counting monitor, peripherals, etc) suddenly doesn't seem as extravagant. I do use its full capability somewhat regularly too, so it definitely wasn't a waste.

Depending on benchmarks, I may switch out my 965 for a Gulftown when it comes out too - it'll all depend on how much performance it gains and how it does in power and thermal characteristics.

 

[randomizer]

HEY
Thank you for this info. That's a hefty commitment for schooling.

Was any (RE)compiling done as part of your tests or in normal useage?

Thanks again.

Recompiling would require access to the source code. No software company gives away its most valuable asset for free 😉

 

[cjl]

Here's the current flow sim I'm doing - it's a rocket at mach 2.8 (note the CPU time readout):

And the CPU usage: