R300/RV250 Information

[up]

Just saw at Ace's Forum This exciting news
Amazing!
Via
http://www.rage3d.com/#1023836126
and 'full story' you will find a chineese scan with summary!
So if the drivers are OK get me one!

 

[Doomtrooper]

Not bad for a Beta board, 20% is not accurate and its also not showing what happens when all the eye Candy is cranked up

 

[Joe DeFuria]

Well, 20% could be accurate...the problem is that number by itself means nothing unless we have context (as you alluded to.)

20% at what benchmark, what resolution, what detail settings....

I've seen a lot of debate on several boards over whether 20% is "good or bad." There's no point in really having that debate, because we have no idea in what context that 20% figure, even if accurate, applies to.

 

[Randell]

amen

 

[ChrisK]

The scan...

The scan here: http://www.3dgpu.com/misc_images/pcmarket-r300.jpg says they tested it under 1024x768 without any form of AA. So isn't this rather meaningless? I don't think that R300 would be fillrate or bandwidth limited under these circumstances.

 

[Hellbinder]

Well, considering the *rumors* that R300 is 256bit delivering slightly over 20GB bandwidth....

Id have to say that is a little low.

I do think that they where refering to 3dmark, being that R300 is Dx9 you may not see a giant increase., until 3dmark 2002. There is this wierdness though. i just cant see how a 4 Vertex shader/8pipeline card could only get that minimal of a gain. even in a beta form. Drivers etc will play into it of course.

I guess if the R300 is only clocked at 250mhz or less (in its current form) that score is not to bad. A final board clocked at 300 to 350 would be about 50%+ faster. That is more like what people are expecting i think.

 

[Doomtrooper]

It's still a Beta board and from what I've heard ATI wants 400 mhz Ram...so I think there will be different flavors of the R300

 

[BROGS]

its so hard to get a feeling one way or the other on the r300 with only bits and pieces of info.

 

[Bambers]

3d mark is fairly cpu intensive in many of the tests. At only 1024x768x32 its quite possible that the test is getting cpu limited so a 20% increase is reasonable.

 

[Doomtrooper]

I really don't think we will be dissapointed by the hardware, I just hope ATI executes well and drivers are solid from the start.

 

[BROGS]

i have a feeling that after the early 8500 drivers weren t so good and all that mess, ati has learned its important to start out with decent drivers.i think they ve come a long way in the driver area ,so they don t want take a step backwards right?

 

[BROGS]

i just have a feeling ati is going to keep up good driver support .i could be wrong,but i hope i m not

 

[LittlePenny]

Would 3dMark even take advantage of the 8 pixel pipes and 24 textures per pass? I would imagine new cards like the R300 and Parhelia are not going to be all that much faster in older games until you have every feature enabled.

Let me put it this way, the GF4 is like a modern sports car with all the speed, the R300 is like the classic 454 with the speed and torque. Both are fast, but the 454 could tow a trailer and still kick butt.

 

[Entropy]

3d mark is fairly cpu intensive in many of the tests. At only 1024x768x32 its quite possible that the test is getting cpu limited so a 20% increase is reasonable.

Amen.
You can test this yourselves by reducing the test resolution to 640x480. On my system this made my 3DMark score go from 9200 to 10600. That's less than 20%, despite almost a factor of three reduction in number of rendered pixels.

For an R300 at default settings on a typical test system of today, the 3DMark2001 bottlenecks probably lie elsewhere in the system.

Entropy

 

[LittlePenny]

I had another thought. What were the score differences between the GeForce 3 and GeForce 2 in 3DMark2000?

 

[Mintmaster]

3d mark is fairly cpu intensive in many of the tests. At only 1024x768x32 its quite possible that the test is getting cpu limited so a 20% increase is reasonable.

You're right. This is what I posted at Rage3D:

20% increase of R300 over T4600 makes perfect sense, and is very good.

Remember how CPU limited 3DMark2001 is with the newest cards. The Ti4600 is over 30% faster than the Ti4200 in most games. However, it's not even 15% faster in 3DMark2001.

If that rumour is true, 20% increase in 3DMark will equate to some 50% increase or more in games, as with the R300 it will be even more CPU limited. And that's with preliminary silicon that can't run at full speed, has bugs, and with immature drivers. Noting that the Ti4600 is already about 35% faster than the 8500, that'll make something like 2x performance from the 8500 to R300 (1.5 x 1.35 = 2.025).

Now we don't know if this rumour is true, but what I'm telling you is it makes perfect sense. Stop expecting so much from the video card manufacturers, as they can't work miracles.

 

[Tatchan]

That's running from a very high-end system....in fact, my colleague is the one who wrote this article..... -_-

 

[Doomtrooper]

That's running from a very high-end system....in fact, my colleague is the one who wrote this article..... -_-

And I assume that was a final board

 

[Tatchan]

I never though of my article being scanned
Anyway, I don't think it's a finished board and the driver is really simple, no TAB appears under properities....

 

[Doomtrooper]