Well, Futuremark has definitely done some things right. For one, they have a great online score comparison service. I just think that they could do a whole lot better in the game tests. That is, they could partner with a few game companies and demo actual games (or future games, rather). Then it might be a worthwhile benchmark.
there is no NV47, but G70 is on the way
- Thread starter DOGMA1138
- Start date
for last few years i've been using that rthdribl to get a estimation on performance, probably more accurate that 3dmark. i say 7800gtx should get about 100-120fps in this thing in default res.
and what about 3dmark2000, it was by far the best 1, nice demo with good music. still fun running it now with 8xs
and what about 3dmark2000, it was by far the best 1, nice demo with good music. still fun running it now with 8xs
Chalnoth said:
Some more IQ comparison tools would be nice too. Back in the day, later versions of WinBench would show you what the reference image should be for various scenarios/features, and what your card was putting out. I always liked that (tho of course the tech is much more complex now!)
Yeah. What I'd really like to see is them put out a software renderer that has truly superb quality (e.g. angle-independent 64x anisotropy, 64-sample sparse supersampling, etc.), and use that as sort of dream image for IHV's to strive for (there is, after all, no simple reference image any longer, not with anisotropic filtering and antialiasing available).
And the IQ tools should be in the free version, so that there would be millions more eyes out there looking at this stuff. . .and some thousands of those would be reasonably competent types (I certainly don't include myself) who would effectively be acting as "bird dogs" for the Baumanns/Anand/Brent/Rys etc of the world. . . If it is just the paid version, there aren't enough out there, and too many are in the hands of types who aren't going to talk about it publicly for one reason or another.
DegustatoR said:Next NVs hi-end GPU should be WGF2-capable imho. Whether this will be in Spring'06 (optimistic), Summer'06 (realistic) or Autumn'06 (pessimisticAiluros said:) is hard to guess atm.
Doesn't exactly fall in line with what Michael Hara stated in March 2005:
http://www.beyond3d.com/forum/viewtopic.php?t=20937
I think they'll use the same approch as NVIDIA with NV47 oops G70
Probably but there's still something not adding up in mind yet and that for both IHVs roadmaps.
Ailuros said:
Well, Wavey has (so I perceive at least) shifted to saying both IHV's will have WGF2.0 parts "significantly" before WGF2.0. . .does that help rationalize it for you?
Edit: Oh. The Hara piece clearly points at a "new architecture" Spring 2006. Yet that might turn out to be a year before WGF2.0, given they are rumored to be separating it from Longhorn for later release. So very tweenerish. Personally, I still puzzle over the performance implications of one architecture for both WGF2.0 and non-WGF2.0, but I take Wavey's comments to mean both IHV's are feeling comfortable about it.
Expecting a year's disparity between WGF 2.0 and the first hardware that can use it is a bit much. Six months would be an absolute maximum.
Furthermore, it seems more than a bit odd that ATI would release their first WGF 2.0 part immediately on the heels of their first SM3 part. No, I expect to see the WGF 2.0 parts from both companies to be available no sooner than the middle of next year.
Furthermore, it seems more than a bit odd that ATI would release their first WGF 2.0 part immediately on the heels of their first SM3 part. No, I expect to see the WGF 2.0 parts from both companies to be available no sooner than the middle of next year.
geo said:
No because I can't possibly know for which timeframe exactly the WGF2.0 release is being projected.
Where cause I'm obviously blind?
If I am to trust the above it sounds more to me like the "spring part" turned out to be a "summer part" (G70) and the "fall part" to turn out to be a "holiday part" (G80?). I still don't see anything beyond SM3.0 up to that point. In that case something like late summer 2006 for a WGF2.0 part could actually make sense.
ATI's timeline sounds somewhat "tight" though if I am to think R520/R580/R600 all within less than a year?
I'm actually thinking next fall (or late summer). . .but that does leave a gehenna big hole between now and then. ATI is in a relative position for a minor-ish tweak, since their process/generation is relatively new. Otoh, we had some rumors about R580 that seemed more ambitious than that.
NV has a bigger mountain to climb I would think. . .and hence you get into the problem of resolving that with Hara's pointing at Spring 2006 for a new generation vs expected WGF2.0 availabilty --which would seem to point at being smack dab tweener for NV. There's a certain tension between the two ideas. Hara is at least 6 months off on his statement, or NV has to "hop" generations in a year (or less) instead of two this time. If I was them, just shrinking to 90nm and clock bumping (with maybe a minor tweak or two) would seem to make sense, and be least risky, aimed early 2006; and then the big one (with 90nm experience already under your belt) in the Fall. But that's not what the Hara said.
NV has a bigger mountain to climb I would think. . .and hence you get into the problem of resolving that with Hara's pointing at Spring 2006 for a new generation vs expected WGF2.0 availabilty --which would seem to point at being smack dab tweener for NV. There's a certain tension between the two ideas. Hara is at least 6 months off on his statement, or NV has to "hop" generations in a year (or less) instead of two this time. If I was them, just shrinking to 90nm and clock bumping (with maybe a minor tweak or two) would seem to make sense, and be least risky, aimed early 2006; and then the big one (with 90nm experience already under your belt) in the Fall. But that's not what the Hara said.
Ailuros said:
Let me change the bolding. . .
NV40 is April 14, 2004. Next "revolutionary architecture" is therefore Spring 2006, give or take a touch (maybe one quarter?).
Unless you want to argue he wasn't pointing at NV40 as the most recent baseline --but I think that will get us in deeper water rather than shallower. So far as I can tell, they took a pass on a Spring refresh for 2005, and 7800 is their Fall refresh moved up a touch. I'd also argue that SLI took the place of their Fall 2004 refresh.
Ailuros said:
We argue (in the good sense) within the limits of the data. . .and six months is a long time in this business.
Tho I suppose I could argue that 7800 is their Spring refresh *pushed back* a quarter (real availability on the announce date supports this theory, as do various smug references that G70 is NV47). . .slide an ultra in a quarter later (they certainly seem headed there!). . . the Fall refresh becomes the Spring 2006 refresh and now you have your extra six months.
Come to think of it, going further back might help. If you consider NV30 "on the roadmap" (compared to when it was actually released) as Fall 2002 and NV40 as the manhattan project they weren't expecting, then you really get two generations from NV30 as Fall 2006. . .
Bah. Not enuf data. I think it will be a lot clearer by September or so.
digitalwanderer said:Actually 3dm2k5 pretty much did lay 'em to rest for me, it's definately one I put almost no stock whatsoever by.martrox said:Can't do anything but hope so........it's about time to lay Futuremark to rest.....
I still like 3dm2k1, and I never really did like 3dm2k3...it's definately looking that FM's useful life-span is rapidly reaching it's end.
At least 03 has a DX7 game which is cpu limited, two DX8 games of varying number of polygons and one DX 9 game, so it has a wide spread.
You are right though, it looks like 05 dropped the ball somewhat.
Just spoke with someone benchmarking the g70's (and a lot of you will read his review tomorrow) that plainly said that g80 is nv50 just as g70 is nv47.
He also mentioned "three times shader performance" and that a "512" bit memory bus for g70ultra may not be far from the truth...
He also mentioned "three times shader performance" and that a "512" bit memory bus for g70ultra may not be far from the truth...
The first as a theoretical maximum I can believe, for the latter I have severe doubts. Did he get the latter from benchmarking G70? (ok ok bad tasted joke )
True. I was about to post something similar.
Cancel =! postpone. Unless Fuad is considered a reliable....damn I have a hard time even saying it...source
Similar threads
