Anand's retail Radeon 9500 Pro review - much faster!!!....

[Basic]

DT:
Change that to 50.6 fps for 9700pro and 31.2 fps for 9500pro.
(According to the test you refered to.)

 

[Joe DeFuria]

Especially in Doom3. I expect that game to be fillrate limited (not bandwidth limited) almost everywhere. (Except where it's CPU limited.)

I agree, except if you turn on AA. In that case, I'd wager that Doom3 tends to be bandwidth, rather than fill-rate limited.

If the speculation in this thread turns out to be true (regarding 'fillrate' and 'bandwidth' balance between the GFFX and R-300), this could make for a very interesting set of trade-offs for Doom3.

Namely:

1) NV30 should be able to handle it better at higher resolutions.
2) R-300 should be able to handle it better with higher AA

We may end up with a situation like NV30 "sweet-spot" being something like 1280x1024 with no AA, and Radeon 9700 being at 1024x768 with 4X AA.

Another side note if Doom3 is extremely fill-rate limited...that would mean that the Radeon 9500 Pro should be almost as fast as the 9700 at Doom3 when running in non AA situations.

 

[stevem]

For a couple of days now, Xbitlabs have been trying to insinuate something other than a standard 128-bit bus for the 9500Pro. I dunno if they've put 2+2=5, given the increased performance with the new driver or not... Their reasoning seems to be based of the memory chips used:

"PS. Although ATI declares 128-bit memory bus for the RADEON 9500 PRO, they utilise 8 memory chips with 4Mx32 organisation from Hynix."

Perhaps they think it's a dual 128-bit bus for 128MB boards as opposed to straight 128-bit for 64MB boards. I can't quite find the ref I saw the other day, & my Ruski isn't helping with their thoughts on ixbt...

 

[Doomtrooper]

DT:
Change that to 50.6 fps for 9700pro and 31.2 fps for 9500pro.
(According to the test you refered to.)

Yes my bad..

 

[Oats]

It's definetly a very nice card. But the price/performance point I think they need to target is that of the 4200, not the 4600. I am not sure I'd pay 50$ more for the 9500pro.

 

[martrox]

It's definetly a very nice card. But the price/performance point I think they need to target is that of the 4200, not the 4600. I am not sure I'd pay 50$ more for the 9500pro.

I'd gladly pay $50.00 more to have TI4600+ performance AND DX9 compatibility...... in a heartbeat!

 

[Fuz]

R9500 is an exellent product, there is simply no doubt about that!

Did hell freeze over?

 

[jvd]

Fuz it didn't freeze over , but it did get a few inches of snow...

 

[Mr.huang]

8 pipeline 3D Card is my next upgrade target, now i see today tons of review out and i am very impressive, maybe this timeframe i can buy a R9500 Pro or R9700 for my new year present , hopefullly i can prepare 300 bulks and go to the shop to choice R9500 Pro or R9700

 

[Dave Baumann]

Perhaps they think it's a dual 128-bit bus for 128MB boards as opposed to straight 128-bit for 64MB boards. I can't quite find the ref I saw the other day, & my Ruski isn't helping with their thoughts on ixbt...

Look at the arrangement of the chips. Its quite clear that the traces got to the top porotion of the chip, whereas the 256-bit bus versions go to the top and left sides of the chip.

 

[Nagorak]

It's definetly a very nice card. But the price/performance point I think they need to target is that of the 4200, not the 4600. I am not sure I'd pay 50$ more for the 9500pro.

Are you serious????? Ti4200 is old news now... It was last year's value card. Why the hell would they be aiming for Ti4200 performance at this point?

Outside of that all cards are introduced for a higher price when they are new. If you go back and look you'll see the Ti4200 MSRP'ing for around $200 upon release. You could use the same argument back then "it's not worth paying $50 more when I could just get a R8500". Bottomline: older cards are cheaper, it doesn't take a genius to understand why.

R9500 Pro is aiming for the mainstream segment. If it undercuts the Ti4600 so much the better for us consumers.

 

[Bjorn]

Are you serious????? Ti4200 is old news now... It was last year's value card. Why the hell would they be aiming for Ti4200 performance at this point?

I agree.

The R9500 is up to twice as fast as the Ti 4200 with aniso and FSAA. (Anand's UT2003 benchmarks f.e, 22 vs 44.9 fps). And then you get a DX9 card also to spice it up a bit more. I would gladly pay 50$ more for that if would buy a card right now.

 

[woolfe99]

It's definetly a very nice card. But the price/performance point I think they need to target is that of the 4200, not the 4600. I am not sure I'd pay 50$ more for the 9500pro.

FSAA/Aniso performance is the only thing that really matters anymore. Who doesn't turn on *some* level of FSAA/Aniso with the performance of the current generation of cards? The 9500 Pro is 75%-100% faster than a ti4200 in that regard. Indeed, it's 10%-50% faster than a ti4600. That and DX9 compatibility make the extra $50 kind of a no-brainer.

I can't remember being this excited about the price to performance/feature ratio of a new card in a long time.

 

[alexsok]

Did hell freeze over?

Not really m8, I just praise ATI where they deserve to be praised: their latest batch of cards simply rocks!

 

[PVR_Extremist]

It's definetly a very nice card. But the price/performance point I think they need to target is that of the 4200, not the 4600. I am not sure I'd pay 50$ more for the 9500pro.

FSAA/Aniso performance is the only thing that really matters anymore. Who doesn't turn on *some* level of FSAA/Aniso with the performance of the current generation of cards? The 9500 Pro is 75%-100% faster than a ti4200 in that regard. Indeed, it's 10%-50% faster than a ti4600. That and DX9 compatibility make the extra $50 kind of a no-brainer.

I can't remember being this excited about the price to performance/feature ratio of a new card in a long time.

Actually I don't turn on FSAA or AF for new games like UT2003. MY processor (Athlon 1Gig @ 1.43) simply aint meaty enough coupled with my 4200 for 1024 + graphics glamour. That'll change when the XP2000+ goes in a christmas mind

For older games (CS etc) bring on the effects.....

 

[Bjorn]

Actually I don't turn on FSAA or AF for new games like UT2003. MY processor (Athlon 1Gig @ 1.43) simply aint meaty enough coupled with my 4200 for 1024 + graphics glamour. That'll change when the XP2000+ goes in a christmas mind

For older games (CS etc) bring on the effects.....

AFAIK, aniso and FSAA doesn't cause any extra load on the CPU so i don't understand how a faster CPU would help you in this case.
Unless you're entirely CPU limited of course but then you should be able to enable aniso and FSAA without loosing any performance

 

[PVR_Extremist]

Isn't UT2003 a CPU dependent game anyhow?

Regardless. Here is what I have found:

Net effect = Turn (FSAA and/OR AF) on & 1024 is unplayable (FOR ME). Turn it off = playable.

Since benchmarking (done by anands for example) with an identical card at identical resolutions but with a faster processor yields higher frames I thought a faster CPU would allow things like FSAA to be playable all else being equal.

Someone educate me!

 

[mboeller]

The question is, is 9700 pro indeed fillrate-limited? Could someone do a quick test to determine that?

All you need to do is take something like UT2k3 Inferno, run the benchmark at high resolution at default clockspeed, then overclock the core by say, 5-10% and run the test again. If the card is indeed fillrate-limited, there should be a close correlation between percentage increases of core clock and performance. On the other hand, leaving core at the default frequency and overclocking memory alone should have less of an impact.

Using the controversal Digit-Life pics it seems the R300 is really fillrate-limited :

article : http://www.digit-life.com/articles2/radeon/r9700pro-oc.html

As can be seen in the picture above the OC of the core alone increases the fps from 95fps toward 113fps = 18%. So it seems the core is "slightly" fillrate-limited. Maybe an 8x2 system would be too much.

 

[arjan de lumens]

Isn't UT2003 a CPU dependent game anyhow?

Regardless. Here is what I have found:

Net effect = Turn (FSAA and/OR AF) on & 1024 is unplayable (FOR ME). Turn it off = playable.

Since benchmarking (done by anands for example) with an identical card at identical resolutions but with a faster processor yields higher frames I thought a faster CPU would allow things like FSAA to be playable all else being equal.

Someone educate me!

OK... AA and AF do not increase the load on the CPU at all, as all the extra work associated with these features is done in the renderer only - they increase the load on the GPU and its memory a lot. What presumably happens is that the game goes from being more or less CPU-limited to totally GPU-limited once you turn on AA/AF. So upgrading your CPU will likely increase the non-AA/AF performance nicely, while AA/AF performance remains just as bad as before.

 

[Bjorn]

Isn't UT2003 a CPU dependent game anyhow?

Regardless. Here is what I have found:

Net effect = Turn (FSAA and/OR AF) on & 1024 is unplayable (FOR ME). Turn it off = playable.

Since benchmarking (done by anands for example) with an identical card at identical resolutions but with a faster processor yields higher frames I thought a faster CPU would allow things like FSAA to be playable all else being equal.

Someone educate me!

As Arjan said, you might go from being sometimes CPU limited and sometimes GPU limited, perhaps depending on the level you're playing or what part of the level you're in.

So, what you might avoid are the dips in framrates caused by the CPU (= higher overall framerates in a benchmark) but you will still experiance the dips caused by the GPU (which you must have since you're saying that enabling FSAA and aniso makes the game unplayable).