R520 benchmarks - Hardware Analysis

Status
Not open for further replies.
NV was hot to tell everyone they took out game profiles from their compiler for G70. But they have those new ALUs (which caused the more cynical to say _of course_ they took out the compiler profiles. . they wouldn't work any more). So surely there should be an opportunity for some decent increases as they play with their new toy. Having said that, ATI should have an even bigger opportunity along those lines over time as it seems pretty clear that R520 is a greater departure from R420 than G70 is from NV40. . .
 
neliz said:
The api isn't there for SM4.0, what would ati have done if it release the r600 the way it was originally planned? as the r400 on the r420 launch schedule..
It would look ridiculous because there simply isn't anyone or anything using it.
It would thus be slower because it would have to recode PS/VS data to unified shader data and you'd have a very expensive card that just wouldn't be able to hang in there with the competition.. and when SM4.0 finally would arrive.. there'd hardly be any game actually worth running it on .. in competative benchmarks..

You do know about SM-3.0 on Nvidia 6 boards were released way back then, when it was not even required, and now they've released a card with more Pixel Pipelines with SM-3.0, and they can still easily release a latest card based on SM-3.0 with more pipelines.

I do get your point, but somehow I think its better to be a one step ahead, rather than one step behind, especially if a competition is so hight.
 
The original comment from www.hardwareoc.at regarding the errors is:

Update 81.26 beta: Der Treiber ist sehr schnell, optimal für neue Benchmark Rekorde. Aufgefallen sind uns Fehler bei den Benchmarks, aber nur bei den CPU Tests komischerweise. Bei den 3D Messungen konnten bislang keine Grafikfehler gefunden werden. Natürlich werden wir es euch wissen lassen, falls sich das ändert.

Translation:
The drivers are very fast, optimal for new benchmark records. We have noticed errors with the benchmarks, but strangely only with cpu tests. There were no errors during the 3D measurements so far. Of we will let you know if that changes.
 
Supreet Virdi said:
You do know about SM-3.0 on Nvidia 6 boards were released way back then, when it was not even required, and now they've released a card with more Pixel Pipelines with SM-3.0, and they can still easily release a latest card based on SM-3.0 with more pipelines.

I do get your point, but somehow I think its better to be a one step ahead, rather than one step behind, especially if a competition is so hight.
Pipelines don't mean jack. Look at fillrate, not pipelines. And talking about SM4.0? It's not even finalized as far as I know, so implementing it now would be equivalent to the original Radeon's PS support. Maybe you should check that out and see what happened when someone tried to do pretty much exactly what you're suggesting instead of blaming ATI for not supporting a spec that doesn't exist.

Considering they're using a new process, too, I'd much rather they optimize it correctly through respins than decide to NV30ize it and jam up the voltage and strap on a box fan.

And why would NVIDIA be scared? They've got a month before the ATI cards are available in volume. They've had no competitors in the high-end since G70's release. They've sold plenty of SLI setups. They could, in theory, make a 90nm G70 and jack up the clocks. NVIDIA has nothing to be scared of from R520's performance. Features, though, are another story.
 
Last edited by a moderator:
The Baron said:
what you're suggesting instead of blaming ATI for not supporting a spec that doesn't exist.
fan.

I know about Pipelines, Core, Memory and fill rate, my point is they could've tried to implement SM-4.0, so that it would've become easier for ATI to release new products with more pipelines... that's it. ( noob people would judge a performance through pipelines, not me) However, 24 Parallel Pixel Pipelines with 750Mhz on Core doesn't hurt...does it ;)

And In case you don't know about me.... I dont think any ATI fanyboy is bigger than me.

I am 101% Pure ATI Whoop ASS ! ;)
 
I guess this goes to show both the strength and the weakness of the internet. Anyone can set up a website and publish to millions of people, making their one voice disproportionalty loud.

In this case Sander forgot some basic tenants of journalism - check your facts and stay objective. He could have got away with it if he presented things as an editorial or opinion piece, but instead he presented his "review" as facts, and called his own impartiality into question by revealing he has an axe to grind.

Whether Sander's numbers are right or wrong is pretty irrelevant, as they cannot be trusted from someone who has blown his own credibility out of the water so completely.
 
Druga Runda said:
Reality leaves a lot to the imagination ;)

From my point of view it looks like the e-mail leak was taken out of original context. "The Threat" was revealed in order to prove a malicious intent on the behalf of the person posting the alleged "review". IMO a fair and square PR department wouldn't go that far, as they know the real results, to be shown in few weeks, will vindicate their product and this is all just the usual pre-launch hype. (they surely have better things to do with "real" world product launch preparation etc... besides of the point: Can an internet publication really "threaten" IHV with some questionable review ) In reality :smile: even "fake" bad scores had no chance of hurting the product, as the real numbers will speak for themselves when they are finally released.

That I why I said "someone will get a slap" as it distracts from the product launch and it sends the wrong message to the other industry contacts: "All your correspondance with us is fair game when you cross our path" which might take some effort to repair despite the statements to the contrary. Besides that I am sure the PR team are more than full-time busy to set everything up for the major event of the year. A typical and simple "this does not represent the results of a real product blah blah blah" to the community would have been more than enough.

However if anything the leak of this "threat" gave us the drama factor and much better publicity to new product launch. If this was premeditated it was a shrewd move.
icon14.gif
Perhaps without that leak there would be no drama, the hype machine wouldn't go into stratoshpere, so maybe this will be a part of college marketing 101 textbooks in 10 years as a case study. "How to best utilize the fanbase in order to maximize the product visibility at launch. " :p

Dear Official Sander Sassen Apologist,

Out of context? Sander was asked 4 or 5 times what context it was taken out of and he coudn't answer. It was a threat, plain and simple - there are people, plenty of people out there that would have believed this moron if the ATI PR dept hadn't stepped in. If you don't want threats published, don't threaten people.

Even Sander's own mother wouldn't defend him on this point. You're probably one of those bleeding hearts who think that people who rob banks and are injured in the process should have the right to sue!!

Mayhem
 
Supreet Virdi said:
I know about Pipelines, Core, Memory and fill rate, my point is they could've tried to implement SM-4.0, so that it would've become easier for ATI to release new products with more pipelines... that's it. ( noob people would judge a performance through pipelines, not me) However, 24 Parallel Pixel Pipelines with 750Mhz on Core doesn't hurt...does it ;)

And In case you don't know about me.... I dont think any ATI fanyboy is bigger than me.

I am 101% Pure ATI Whoop ASS ! ;)
And releasing an "SM4.0" part WHEN THERE IS NO SM4.0 TO BEGIN WITH makes how much sense?

24 pipelines at 750Mhz is completely impossible on 90nm, I'd bet, unless you're using stupid cooling or somehow equating unified shaders to pipelines. Once you get unified shaders, they're not even really pipelines per se anymore, but SM3.0 and below are probably not conducive to unified shaders and would require Really Huge Investments in compilers to make them not suck. The only reason Xenos can do it is because there's one chip for the plattform--that's what you're targetting. If you have 1 chip that uses unified shaders and every other chip that doesn't, what are you going to target? If, in the next gen, you know everything is going to target chips with unified shaders, why wouldn't you wait?

PS--yes, Sander could have solved all of this by not being an idiot about how he published the numbers, or by publishing every email between him and ATI from the past year without any commentary, or by publishing his timedemos, but he won't. Wonder why. Oh wait, I don't have to.
 
The Baron said:
And releasing an "SM4.0" part WHEN THERE IS NO SM4.0 TO BEGIN WITH makes how much sense??

Nobody put the same question to Nvidia when they released Geforce 6, and at that time I had hardly seen a game which is fully compliant with DX9C. I know what you are saying, all I am saying its better to be one step ahead rather than one step behind.
 
neliz said:
Yes, Like Ratchet's 7800GTX review, you will write a b*ll licking X1800XT review ;) .. unless.. in andrzej's view, you're not the biggest nv fansite :D

btw, maybe sanders source was the same as rage3d's sherman, claiming that his friend over at asus told him how bad the r520 was and all the "render errors" it had..
Err... hrm, was I just insulted?
 
Supreet Virdi said:
Nobody put the same question to Nvidia when they released Geforce 6, and at that time I had hardly seen a game which is fully compliant with DX9C. I know what you are saying, all I am saying its better to be one step ahead rather than one step behind.
Here's the difference. The spec for SM3.0 existed when GF6 came out. It does not exist for SM4.0 yet, and probably won't for another six months or so. Your statements about "being one step ahead" make absolutely no sense.
 
Oh, so we're including cascade cooling in this now, too? :rolleyes: Shipping a 24 pipeline part at 750Mhz and overclocking a part to 750Mhz with exotic cooling have, I don't know, nothing in common.
 
The Baron said:
Here's the difference. The spec for SM3.0 existed when GF6 came out. It does not exist for SM4.0 yet, and probably won't for another six months or so. Your statements about "being one step ahead" make absolutely no sense.

Does it hurts?
 
Supreet, give it up - You're talking nonsense and only serving to dilute the purpose of the thread. If you really want to have a 'ATI NEEDZ SM 7.1 NOW OMGWTFLOLZ?!?!' discussion it would be best to start a new thread.
 
Status
Not open for further replies.
Back
Top