NVIDIA GF100 & Friends speculation

[rpg.314]

The Chinese name card is most likely GTX470,but I'm not sure.

Another rumor from Chinese forum:
http://we.pcinlife.com/thread-1366899-1-1.html

So loses to 5870 overall. Let's wait for more benches.

 

[Thraxis]

Yeah, it was against TTK's post which he referred to Xman's clock numbers which were astonishingly low. Also 448 was meant to be GTX470 but still works I guess.

 

[rpg.314]

Is it just me or do those numbers almost look like they may just possibly be real?

Also, the gap bw 5870 and 470 suggests that the Charlie 5% claim might actually be true. Bring on the leaks.

 

[Silus]

Is it just me or do those numbers almost look like they may just possibly be real?

No, it's just you!

 

[Silus]

So loses to 5870 overall. Let's wait for more benches.

So I assume you think these are real ? How come ?

 

[DegustatoR]

Thats a great achievement from NV .This could be used by Nvidia PR

That'd certainly be a great achivement if the performance delta stays the same or goes up this generation. And AMD's PR been using their die sizes comparisions for several years now, yeah.

 

[ShaidarHaran]

As XMAN26 wrote (too lazy to search for his post now), the remaining parts of what was once called the core, the ROPs and L2 cache, are probably clocked at 450-475 MHz. The GTX 470 has 8 ROPs disabled, so:
475*40=19000

The post to which you're referring was a guess on his part, an unlikely one at that. We haven't seen ROPs clocked @ only 475MHz since the days of NV40.

GTX 470, potentially a $500 part, will not be out-performed by GTX 280, a part launched 21 months ago. It is illogical.

 

[Mindfury]

So I assume you think these are real ? How come ?

Asuka is one of PCI's administrators,I don't think he faked up those benchmarks.

 

[GZ007]

That'd certainly be a great achivement if the performance delta stays the same or goes up this generation. And AMD's PR been using their die sizes comparisions for several years now, yeah.

I think u didnt understand me :smile:. I was thinking the oposite.

 

[Silus]

Asuka is one of PCI's administrators,I don't think he faked up those benchmarks.

LOL, I've heard that before during the R600 debacle, where a guy, that was said by many to be "hardly ever wrong" was saying the HD 2900 XT was faster, quieter and consumed less power than the 8800 GTX.

I think we need to wait for actual benchmarks from known sites/sources, with all the testing conditions known, so that we actually see how the new cards perform.

 

[rpg.314]

So I assume you think these are real ? How come ?

No, I am waiting for more benches.

So loses to 5870 overall. Let's wait for more benches.

 

[Silus]

No, I am waiting for more benches.

The numbers are very weird for Crysis WarHead. The GTX 470 scores show that it has almost no performance advantage over the GTX 285, with the same settings in-game (although different systems) as seen below for example:

http://techreport.com/articles.x/17652/8

I won't question the Dirt 2 scores, since those are probably close to reality, given that it's a AMD sponsored game.

 

[mczak]

Is it just me or do those numbers almost look like they may just possibly be real?

Wouldn't be so great for nvidia. Only wins in the benchmark using tesselation (and not by much neither). And by the looks of it only draws even in crysis warhead 8xAA because the H5870 is running out of memory.
Sounds plausible though, how much faster can the gtx480 compared to the gtx470 be? 20%?

 

[TKK]

The post to which you're referring was a guess on his part, an unlikely one at that. We haven't seen ROPs clocked @ only 475MHz since the days of NV40.

Ok, only remembered that number, not that it was just a guess on his part.

GTX 470, potentially a $500 part, will not be out-performed by GTX 280, a part launched 21 months ago. It is illogical.

I wasn't referring to absolute performance, just to the relative performance hit of 8xAA vs. no AA.

Anyway, if the numbers posted on PCI are anywhere near the truth, then my speculation was completely off, so yeah, I admit defeat

 

[rpg.314]

The numbers are very weird for Crysis WarHead. The GTX 470 scores show that it has almost no performance advantage over the GTX 285, with the same settings in-game (although different systems) as seen below for example:

http://techreport.com/articles.x/17652/8

Good find. These numbers sound dodgy now.

 

[mczak]

Good find. These numbers sound dodgy now.

I still see a 25% improvement for crysis warhead for gtx470 vs. gtx285 from these numbers. Why would that be unlikely?

 

[Sontin]

You can't compare numbers from different sites. Techreport uses a walkthrought and not the timedemo/benchmark.

 

[Jawed]

Memory bandwidth for 3200MHz/320-bit is only 128GB/s, which is the same as HD5850. So if those GTX470 scores are real then I'd say it's doing pretty well against HD5870's 153.6GB/s.

Jawed

 

[FrameBuffer]

you know better than use a single metric such as memory bandwidth to compare different products using different archs.. /slap

I mean look at the 5770s paultry 76.8GB/s and how it competes with cards almost 2x it's bandwidth (76.8 vs 128.8 of the GTX 260 216) or the 5870's 153.6 peak BW deals well against the GTX 295's 223.9 GB/s /double slap

 

[Psycho]

Considering the specs we know for 470 I see nothing dodgy in it being *sometimes* slower than gtx285.
Less total bandwidth, less raw/simple fillrate, at least in some cases probably less texture rate. With the architectual differences I would expect quite some variation compared to the gt200 cards.

I find the variation between 1x and 8x in the other crysis numbers much more dodgy, but funny enough mr.green didn't immediatly jump on that