NVIDIA GF100 & Friends speculation

[Arnold Beckenbauer]

Try installing the STREAM SDK? IIRC it's required for OpenCL atm regardless of driver version

Yepp. No ATi Stream SDK 2.01 - no OpenCL.
http://developer.amd.com/gpu/ATIStreamSDK/Pages/default.aspx#five

 

[CarstenS]

Yes I figured that out, I just want to hear Carsten's pov on it before I say much more.

And btw: The card lineup you see there is primarily what we did for our print magazine, serving the purpose of helping our readers with their purchase decisions. As you guys are surely aware, each piece of paper is not endless and we had to limit ourselves to a certain number of cards. Those were the most interesting ones.

For the only review, we're could eventually just add MGPU cards.

 

[Mintmaster]

Also, "overtaking in architecture" is irrelevant in regular consumers eyes. It's just like the debate about die sizes.

How the hell is it irrelevent? Did you not notice that NVidia had to chop $100 off the 9800's price and even more off the just released GT200 cards? Or that NVidia's margins went down the drain after that?

Someone back at [H] even suggested that the die size was very important for the consumer, to which I asked "do you go to a store, wanting to buy a card, and ask about the die size of the chip you are about to buy ?"...it's silly and has no relevance whatsoever...

They do look at the price, genius. You think ATI wanted to lose money in the R5xx days? Of course not. They priced their products as low as they could without going too far in the red, and if it wasn't for G7x they wouldn't have done so. NVidia went for record margins because they could, and if the low end RV530/RV515 parts were faster then they would have settled for lower profits.

As for the X2s, sure, but that's why I said "definitely took the performance crown" in regards to RV870, because with the RV670 X2 and RV770 X2, there was always a counter, with which NVIDIA took it back.

It still doesn't make sense to say ATI hadn't caught up until Cypress. They took the perf crown for a few months with the 3870x2, then at most price points after RV7xx was released, then took perf again for 4 months with 4870x2. The 4890 was usually within spitting distance of the GTX285.

That unequivocally meets the requirement of 'caught up'.

(Sontin, sorry, I misread the launch date of the 9800GX2)

 

[Broken Hope]

Try installing the STREAM SDK? IIRC it's required for OpenCL atm regardless of driver version

I'm sure he's aware how to get OpenCL support, but he specifically asked for "open standards support out of the box with certified, regular drivers" which AMD aren't providing since OpenCL support requires sign up at a developer website to download an SDK to get support, where's Nvidia have been providing support in the regular drivers for a while now.

 

[CarstenS]

Try installing the STREAM SDK? IIRC it's required for OpenCL atm regardless of driver version

Yepp. No ATi Stream SDK 2.01 - no OpenCL.
http://developer.amd.com/gpu/ATIStreamSDK/Pages/default.aspx#five

removed flame bait
Sure, and you don't even have to register as a developer any more. Nevertheless, I'd prefer a simple solution via the regular WHQL-drivers. And since that would allow many more people to try out some of the nice Open CL demos and programs, IMO that would also foster the acceptance and pervasiveness of OpenCL.

 

[willardjuice]

If you wish to make the case (for or against) the notion that the "type of fps" varies from card to card, create a different thread.

[edit] If you wish to make the case (for or against) the notion that Germans should be allowed to review video cards, create a different thread.

 

[digitalwanderer]

Back on topic, do we know the exact time the NDA expires? Is it going to be Friday afternoon or like midnight Thursday?

 

[trinibwoy]

Saw 7pm Friday somewhere.

 

[digitalwanderer]

That would be cool, I was expecting a late Thursday night to be honest.

 

[Mintmaster]

I was under the impressione that especially the last two would need bandwidth, too.

Of course. I explicity mentioned that. But if there was any task limited by ROP or texturing speed on a 4890 and used less than 60% of available bandwidth, it would double with the 5870. 8xMSAA with large polygons (and thus good compressibility) can be doubled with Cypress if it needs less than 9 bytes/pix, and magnified textures have very low BW requirements, too.

 

[psychoticdream]

And btw: The card lineup you see there is primarily what we did for our print magazine, serving the purpose of helping our readers with their purchase decisions. As you guys are surely aware, each piece of paper is not endless and we had to limit ourselves to a certain number of cards. Those were the most interesting ones.

For the only review, we're could eventually just add MGPU cards.

thats understandable

i think it is only fair to add the 5970 to the benchmark after all you guys did it with the release of the 5870 before by adding the nvidia gtx 295

http://www.pcgameshardware.de/screenshots/original/2009/09/HD5870-CoD-WaW-1280.png
http://www.pcgameshardware.de/screenshots/original/2009/09/HD5870-FC2-1280.png

not trying to be rude just would like to see it.

 

[Dave Baumann]

Hell, it's a graphics board. If I'd want more audio than my onboard solution from realtek, I'd buy an X-Fi

So you're advocating that adding value and more features is unncessary because you can pay more for other components that do it?? :smile:

 

[Psycho]

Haven't seen this one posted, looks legit:

If that is nvidia's look of it, performance is really close.. 470 is clearly not in 5870 territory on that picture.

 

[mapel110]

http://img522.imageshack.us/img522/6378/gtx4802.jpg

http://img10.imageshack.us/img10/4164/73529807.jpg

http://futuremark.yougamers.com/forum/showthread.php?t=125009

 

[aaronspink]

Why should nVidia have any interest in that? Their parts are basically billed as co-processors, and ones that can easily be upgraded to boot. There's been quite a lot of interest in building clusters with nVidia hardware among people I work with (this is for the Planck satellite, and we have a number of supercomputers dotted around Europe and the US for data crunching).

Data. Data Movement. Cost. Etc. When you can buy off the shelf hardware with comparable levels of performance with a shared memory model to the co-processing vs have to spend have the time moving data around...

 

[aaronspink]

That argument only holds water in low-end and mainstream...for high-end it's irrelevant, there perfomance is the "benchmark"

Even the high end cares about perf/$.

 

[I.S.T.]

Haven't seen this one posted, looks legit:

If that is nvidia's look of it, performance is really close.. 470 is clearly not in 5870 territory on that picture.

Are you REALLY basing performance projections off of a silly PR slide? Please, tell me you're joking.

 

[aaronspink]

Alternatively, they could work with other CPU-manufacturers. In the HTPC space, x86, though popular, is much less important.

No, x86 now only makes up like 95% by revenue and something like 99% by volume. So I guess that is less than the 100%/100% of the PC space, but to say its much less important is kinda funny.

 

[Topman]

Are you REALLY basing performance projections off of a silly PR slide? Please, tell me you're joking.

Check mapel110 post:

PR 'tests'?

http://img522.imageshack.us/img522/6378/gtx4802.jpg

http://img10.imageshack.us/img10/4164/73529807.jpg

Where's HD5970...?
"don't compare a sGPU card to a mGPU card"!!!?!!!

 

[aaronspink]

Could we extrapolate from here, that ATI, without changing their architecture, would be in the future on a less brightly spot? I mean, Cypress seems to have need for much much higher flop power than GF100. I wonder if a GF100 derivate at 22nm tech would really be a beast.

While ATI's perf/flop efficiency is pretty bad compared to nvidia, they more than make it up with their flop/area efficiency which is on the order of 3-4x better than nvidia.