A World Without Limits

Shoot me John but I'll start to worry about Doom3 when it's close to launch.

I am extremely curious though how it will fair in UT2003. In about a month or earlier we should know I guess.
 
Ailuros said:
Shoot me John but I'll start to worry about Doom3 when it's close to launch.

I am extremely curious though how it will fair in UT2003. In about a month or earlier we should know I guess.

Exactly. I was merely qualifying Carmack's quoted commentary.
 
John i see your point.. However I feel that other than the GF4's bandwidth usage enhancements most of the other cards really were not "that" big a jump in performance...

In this case however.. it is clear that the R300 is massive step beyond anything that has been released before. Its got 128bit Floating point color for crying out loud... ;)

But well see...
 
Hellbinder[CE said:
]I suppose that the R300 might get clocked at 250mhz but all the leaked information that comes out week after week keeps pointing to 350mhz. Everyone keeps saying its impossible... but we'll see....

All the "leaked" info comes from the same source; a year-old PPT presentation. The estimates of 350MHz are way off the mark and it hasn't helped that sites like The Inquirer have posted accurate information and then followed it with another rehash of 350MHz R300 bullshit. DX9 has no doubt grown in complexity since then, as has the R300 ASIC.

I wouldn't even be suprised if it was launched at 225MHz. After all, we are pretty sure now that the top RV250 GPU will be 300MHz at-speed, and I know for a fact that the part is capable of ~<325MHz. R300 is capable of ~<250MHz. At 225Mhz it'll still be an awesome performer in current and future titles. At-speed qualification was done at 250MHz, I'm almost certain, so that looks most likely.

Hopefully ATi have opened developers eyes enough with PS 1.4 (that was the whole idea, after all... ) so that we get some great looking shader-centric titles and demos in the future. R300 will no doubt be the showcase product for some spectacular advances in realtime rendering.

MuFu.
 
Remember the r8500 ran faster than a gf3 and IIRC had a higher transistor count but had a smaller HSF. The r300 could follow this trend.
 
king_iron_fist said:
Remember the r8500 ran faster than a gf3 and IIRC had a higher transistor count but had a smaller HSF. The r300 could follow this trend.

We are talking about an ASIC that consumes about 25W (originally ~30W) at-speed here. It won't have a dinky little fan. :)

MuFu.
 
Hellbinder[CE said:
]..However I feel that other than the GF4's bandwidth usage enhancements most of the other cards really were not "that" big a jump in performance...

Unreal Tournament 2003 1024*768 High Quality

GeForce 2 Ultra: 30.4 fps
GeForce 3 Ti 500: 55.6 fps
GeForce 4 Ti 4600: 94.5 fps

Close to doubled performance/generation.
I would call that a big jump.

Then add FSAA (which i guess isn't included in the high details) and the difference would probably be even larger.
 
Quake3 is pretty much a joke now as far as benchmarking is concerned. I won't be too concerned if the R300 loses to the GF4 in this while coming out on top in more demanding tests. UT2003 is one such example.

The radeon 8500 is soundly beaten by GF4 in high quality UT2003 but that's not likely with R300. All the hints (from anand, and others) point to a real sweet performer. A good indication of future performance, not in 3 year old games. Unreal2 is coming soon and should also benefit from r300.

Btw, my name implies that I'm nVidia user. True, true, I've lurked at Beyond3d ever since the demise of 3dfxgamers (I posted as NV20 there), and at 3DGPU for a while (hi John! :)). I'm also still using the NV20: aka Geforce3.

I never upgraded to GF4, thought I'd wait for the NV30, but the R300 is looking good right now, albeit based upon rumor and hearsay. Drivers have put me off ATi solutions in the past though that may change now, they seem to have got their act together. It would be nice to try a different brand too, I hate fanboy logic that says you have to be loyal to any given company.

Anyway, r300 is probably my next card. Only driver problems or lousy performance in current games, especially with aniso + fsaa, would make me reconsider.

edit: hehe, "fan-boy" isn't allowed? It was filtered from my post. I guess the word is overused and hated in equal measure ;)
 
NV25 said:
Quake3 is pretty much a joke now as far as benchmarking is concerned.

It may be an old and simple game graphics wise, but it is far from a joke. It still produces results that are consistant and scale predictably. And the Q3 engine is pretty popular. I suppose SOF2 or even RTCW scores would be more appropriat these days, as those games are pushing more polygons, more overdraw effects, and much higher texture usage.
 
I really *do* hope that ATi this time around have learn't from the driver lessons of old. The CATALYST drivers are a step in the right direction, but I hope for ATi's sake that all features are working off the bat for the R300.

Granted, I won't mind if the speed isn't great, i.e only 20% faster than a GF4 Ti4600 (well who knows), and with the promise of more speed to come. But what I would mind, is if features were disabled. So its buy now, enable later.

Currently im using an ATi Rad 8500, and its just sweet. I have no urge to upgrade. However for me with my P4 2.4, the NV30 is looking the better option. With Nvidia saying its the single biggest contribution they have ever made so far to the graphics industry. They have pegged my interest so far.
 
I hope ATI has their drivers better this time too, otherwise I'm not interested in buying an R300. I haven't been overly impressed with the Catalyst drivers. I can't believe fog is broken again... :rolleyes:

I have a GF4 now and there are no games that require me to upgrade until Doom 3 probably, as it looks like UT2 will run like a champ on a GF4.

I for one can wait till NV30 and then compare it to the R300.

I suspect that the R300 will debut with awesome hardware but crippled by less than stellar drivers as per ATI's norm. and that the NV30 will have great hardware too but with good drivers.

Whoever is better gets my $$$$. Either way its a win win situation.

If R300 is better then all I'm doing by waiting is giving the early buyers time to beta test the early drivers. :LOL:

If the Nv30 is better then I'll have it just in time for Doom 3. :D

Competition is sweet!!!! :D
 
It may be an old and simple game graphics wise, but it is far from a joke. It still produces results that are consistant and scale predictably.

True, but I meant a test to better represent future games - say, in 6 months to a year. It's quite common for a card to be behind in Q3 benches and yet outperform all the others when texture detail and polys are set high. I use HQ settings in just about everything - detail, geometry, textures - etc, and will even suffer a greatly reduced framerate to make sure games look their best.

I'm primarily a single player gamer so multiplay speed is not really a concern of mine. That's why the R300 looks pretty sweet right now if it lives up to expectations. Ut2003 is multiplay, but Unreal2 and many upcoming games are using the same engine - DeusEx 2, for example. Therefore, how the r300 performs in ut2003 is of greater interest to me than Quake3... (Doom3 would be even better, of course :))

In fact, I'm starting to avoid Q3 benchmarks in reviews. Once I see the top cards with over 150fps @ 1600x1200 I just smile and move to the next game...
 
Bjorn said:
Unreal Tournament 2003 1024*768 High Quality

GeForce 2 Ultra: 30.4 fps
GeForce 3 Ti 500: 55.6 fps
GeForce 4 Ti 4600: 94.5 fps

Close to doubled performance/generation.
I would call that a big jump.

Then add FSAA (which i guess isn't included in the high details) and the difference would probably be even larger.

Not really sure how relevant benchmarks are from an unfinished game. Other games and benchmarks don't show the gf4 with such a big lead
 
Well, the GF3 Ti500 is not really just a 'generation' away from the GF2 ultra. More like 1.5-2.

Otoh, a GF3 ti500 pretty much splits the performance crown with the R8500 in benchmarks, or at least did at the time of inception.

Since I give ATI the benefit of the doubt, and put them at parity or slightly ahead of Nvidia (since we're comparing a fundamentally new architecture with refreshes), I'd assume the R300 is a good half a generation to a generation ahead of the highend Gf4. So 1.5* times performance or so with chance to increase that lead through subsequent driver updates.

Something along those lines might be fair I think.

I;ve always been a big supporter of buying new parts, as opposed to refreshes. I feel they give better lifespans. EG, GF3, GF1 DDR, Voodoo3, etc as opposed to Gf4, Gf2, etc etc. Thus, the R300 holds my interest for a new purchase.
 
Andergum said:
I suspect that the R300 will debut with awesome hardware but crippled by less than stellar drivers as per ATI's norm. and that the NV30 will have great hardware too but with good drivers.

like the gf3 you mean :rolleyes:
 
ATi has consistantly put out chips that run cooler than their Nvidia counter-part. With the Radeon series they seem to have more transistors, run cooler and faster at usually lower price points.

That said, I'm not sure if the R300 will follow suite, this isn't designed by the same people and if folks remember the last convention, the R300 has a 10 layer PCB, aggressive cooling and mention was made about power draw being quite high, maybe this design team just doesn't put out such power conscious cards. *shrug*
 
Hehe, personally I often followed the following scenario:
A cople weeks after the the "true" next generation is out (e.g GF3), buy the refresh of the previous generation (GF2) - you usually get a great bang/buck ratio that way and the previous generation's features will be better exposed and most of the early problems solved. Of course that only works when you can be patient enough to wait like 12-18 months between card updates.
I only recently broke that tradition, as the Ti4200 delivers marvelous bang/buck for a newest generation product, but in the past it worked great for me... :)
 
Saem said:
ATi has consistantly put out chips that run cooler than their Nvidia counter-part. With the Radeon series they seem to have more transistors, run cooler and faster at usually lower price points.

Do you know where I could find an article about relative operating temperatures of Radeon 8500 compared to Geforces?
 
Saem said:
That said, I'm not sure if the R300 will follow suite, this isn't designed by the same people and if folks remember the last convention, the R300 has a 10 layer PCB, aggressive cooling and mention was made about power draw being quite high, maybe this design team just doesn't put out such power conscious cards. *shrug*

10-layer PCB? 942 (Computex R300 on VIA stand) used an 8-layer PCB, I'm pretty sure of that. R300 will only be 10-layer if board manufacturers can't guarantee QC with 8. It's not a certainty by any means. 10-layers will just be a workaround for poor manufacturing standards/design. I know Anandtech mentioned it, but that was from feedback from board manufacturers at Computex, not from ATi. I have no idea what the situation is now; we could well see 8 and 10 layer boards about. :-\

NV30 on 0.13u will most likely run cooler than R300 (at ~250MHz) on 0.15u, unless they feed it 2V at 500MHz.

MuFu.

P.S. For what it's worth Geeforcer, the two 8500's I've owned have run at ~45C (back of GPU measured) compared to ~65C (!) for an Inno3D GF4Ti4400 I had for a while and about 55C for my current card which is a Leadtek Ti4400. Doesn't really mean much at all...
 
0.22-micron
GeForce1: 15W

0.18-micron
GeForce2 GTS: 9W
GeForce2 MX: 7W

.018-micron
Radeon 64 Meg DDR:6W

0.15-micron
Radeon8500: 5W


Xbit Labs:
Despite a modest access time (3.6 ns against 3.5 ns of the GeForce3 Ti 500), the RADEON 8500 memory works successfully at a higher frequency, even without any heatsinks. The R200 and NV20 use different approaches for memory operation. While the NV20 prefers smaller blocks and an effective crossbar controller, the R200 uses larger blocks and an intensive caching

I written most of these down over the years from reiews, so I have no dirrect link for all of them.
 
Back
Top