RV770 vs GT200 : hidden potential?

[chavvdarrr]

After arguing with DegustatoR at ixbt forums, I decided to make a poll and later see who was right.
So. Which of the above has more yet undiscovered or hidden from fresh drivers gaming potential ?

 

[AlphaWolf]

where's the 'both have good drivers already, minimal gains' option. I think we'll see the most in terms of crossfire performance, but not much otherwise.

 

[Mintmaster]

where's the 'both have good drivers already, minimal gains' option. I think we'll see the most in terms of crossfire performance, but not much otherwise.

Exactly what I think.

 

[chavvdarrr]

is it possible to change poll now?, I can't even fix "to BE played"

 

[Freak'n Big Panda]

where's the 'both have good drivers already, minimal gains' option. I think we'll see the most in terms of crossfire performance, but not much otherwise.

Yeah I agree. This is pretty much how it always is. I can't remember a card that upon release sucked and then was saved later down the road because of driver improvements.

 

[trinibwoy]

The performance of both cards relative to their predecessors are similiar to expectations based on architectural improvements so I don't think drivers will do much. However, I fully expect GT200 to pull away from G92 in more advanced titles.

 

[Mintmaster]

However, I fully expect GT200 to pull away from G92 in more advanced titles.

Probably, but I also expect RV770 to do the same. It also benefits relative to G92 when there is an increase in register usage, ALU:TEX ratio, and GS usage.

 

[Pressure]

At the current price points I see the RV770 rule.

I fully expect both GT200 and RV770 to gain from future drivers and it would be absurd to think otherwise. All graphic cards have gained performance in their life time

Yeah I agree. This is pretty much how it always is. I can't remember a card that upon release sucked and then was saved later down the road because of driver improvements.

ATI R200/RV200, Matrox G400 (one could argue that drivers kept sucking its entire life time).

They all gained a lot from driver updates.

 

[Mintmaster]

I fully expect both GT200 and RV770 to gain from future drivers and it would be absurd to think otherwise. All graphic cards have gained performance in their life time

Not when they're the third generation of the same basic architecture.

Yeah, RV770 has architectural enhancements, but they're all centered around efficiency and don't need the driver to do anything substantially different. Well, maybe some tweaks in memory controller parameters and data management, but RV770 seems to be pretty maxed out in those departments.

 

[Pressure]

Not when they're the third generation of the same basic architecture.

Yeah, RV770 has architectural enhancements, but they're all centered around efficiency and don't need the driver to do anything substantially different. Well, maybe some tweaks in memory controller parameters and data management, but RV770 seems to be pretty maxed out in those departments.

That "maybe som tweaks to the memory controller parameters" is a huge one. I wouldn't downplay it. A lot of refinement have gone into the RV770. I am sure the drivers will have to follow. Time will tell though.

Considering drivers were not exactly fit for fight a month prior to launch, I wouldn't be surprised to see some gains.

 

[Mintmaster]

That "maybe som tweaks to the memory controller parameters" is a huge one. I wouldn't downplay it.

Not really, because all indications are that RV770's memory controllers are extremely efficient right now. There just isn't any room to improve.

The only test I've seen perform below expectations is Z-only rendering without AA, and that's not really that useful except for shadow maps (which are generally setup limited). In terms of games, I think Lost Planet is the only one performing below expectations. Overall, I just don't see much room for improvement.

 

[igg]

Although I planned to buy a GTX280 card I might buy a HD4870 instead. Why?

• Performance: Because the GTX280 doesn't do Crysis, VH, 1920 in 30+FPS I don't see any use for it; almost every other game plays well in 1920x1200, even on a HD4870
• Price
• Features: AMD likely will support PhysX (according to latest news) and already supports DX10.1

 

[CarstenS]

Although I planned to buy a GTX280 card I might buy a HD4870 instead. Why?

• Performance: Because the GTX280 doesn't do Crysis, VH, 1920 in 30+FPS I don't see any use for it; almost every other game plays well in 1920x1200, even on a HD4870
• Price
• Features: AMD likely will support PhysX (according to latest news) and already supports DX10.1

Apart from the points you mentioned, I also finde the HD48x0 in it's final form very attractive and consider buying a 4850 myself - but with 1 GiB of RAM and a nice dual-slot cooler venting the air outside.

 

[neliz]

Considering that G200 being the third itteration of nV's DX10 architecture, there could be some improvements (game wise) but no big across-the-board improvements. 200b is again, more of the same thing, no Miracle Drivers there..

I think RV770 is brute-forcing it's way through games right now and this could only be improved further by driver intelligence

 

[Pantagruel's Friend]

Considering that G200 being the third itteration of nV's DX10 architecture, there could be some improvements (game wise) but no big across-the-board improvements. 200b is again, more of the same thing, no Miracle Drivers there..

I think RV770 is brute-forcing it's way through games right now and this could only be improved further by driver intelligence

the only significant improvement (apart from raw numbers) is that nVidia seems to have fixed the geometry shader anomaly of the G8x/G9x GPUs, the new card will perform much better in eg. the Global Illumination demo - but it really hasn't much to do with drivers.

 

[Slyne]

Some synthetic tests have shown RV770 to perform below expectations, so I expect improvements there. Not sure that'll translate into games tho.

The other thing is RV770 have the texture units attached to the multiprocessors, while R(V)6x0 have them decoupled. I'm pretty sure the drivers have to optimize differently, and there might be more gains there.

 

[Subtlesnake]

I notice Dave Baumann voted for the third option!

 

[ShaidarHaran]

After arguing with DegustatoR

LOL, say no more

I voted both will receive similar perf. gains via new drivers.

 

[swaaye]

The question is: did R600 and G80 actually benefit tangibly from newer drivers? My 8800GTX feels pretty much the same as it did with the 97.xx series drivers, honestly. Sure there have been bug fixes (and new bugs), but it performs pretty much the same. HD 2900 hasn't really moved out of its sad position either. If drivers are going to improve a graphics card, the first generation of the new architecture is where it would happen I'd think.

 

[neliz]

HD 2900 hasn't really moved out of its sad position either. If drivers are going to improve a graphics card, the first generation of the new architecture is where it would happen I'd think.

Lol, I remember a catalyst performance review and it said "nearly matches the 8800GTs"

which is ofcourse a sad thing, but it wasn't uncommon for a 2900 to jump something like 15 to 20% in minimum performance going from one cat. to another.

To be honest. I haven't turned on my pc with the 2900 in it since december so I can't honestly say if I saw any improvements since then.

but by all means, check some of the performance analysis.
http://www.tweaktown.com/articles/1123/5/ati_catalyst_7_6_performance_analysis/index.html
http://www.pcper.com/article.php?aid=463&type=expert&pid=3