8800GT review at techreport.com

[Denithor]

See their review here.

Close to matching 8800GTS in most benchmarks, slips a bit in some (notably Crysis demo, looks like it just doesn't have enough RAM even at 1280x1024).

Low power draw and quiet cooler!

 

[Pocatello]

Originally posted by: Denithor
See their review here.

Close to matching 8800GTS in most benchmarks, slips a bit in some (notably Crysis demo, looks like it just doesn't have enough RAM even at 1280x1024).

Low power draw and quiet cooler!

Hmmm, interesting.

 

[BAMAVOO]

But at my native resolution of 1680x1050 this thing should kick my x1900xtx all over the place. The Crysis demo actually made me want to buy a new card

I had to turn it down to medium, which made the game look better and play better. I actually didn't like the graphics at high settings, maybe it was the card and not the game though. Now to talk the wife into a new christmas gift.

 

[techgamer]

The best 8800gt review I've seen so far is on the anandtech main page. I would make my opinions based off that.

 

[Denithor]

I agree but AT neglected to address noise levels and did not test in the Crysis demo, both of which were covered by techreport.

Overall looks like a great card, especially at $200 for stock and $250 for OC models as mentioned in AT's coverage.

All I can say is this looks to be the best price/performance card since maybe the Radeon 9700pro days.

 

[taltamir]

I think the reason the GT gets such an impressive improvement on Anandtech's test compared to some others is because anand is actually testing it with 4GB of ram... So it is not a limiting factor

 

[DarthV]

My question is why would they use over clocked 320/640mb GTS cards in their benchmarks and not label it that was in their graphs? They do have a blurb about it on the testing methods page, but not having it labeled on the graphs is misleading at best.

 

[HigherGround]

Originally posted by: DarthV
My question is why would they use over clocked 320/640mb GTS cards in their benchmarks and not label it that was in their graphs? They do have a blurb about it on the testing methods page, but not having it labeled on the graphs is misleading at best.

I'm pretty sure they wanted, for whatever reason, GTS to at least match GT.

 

[thilanliyan]

Something disturbing I noticed (from other reviews as well) was that once you go to 1920 res and add some AA, the GT tanks and loses to the GTS 640mb....in some games anyway.

Damn, and I just sold my GTS. I would have SLIed but GTSs still cost a lot. Oh well, hopefully the RV670 will do a bit better...at least the XT version.

 

[DarthV]

Originally posted by: thilan29
Something disturbing I noticed (from other reviews as well) was that once you go to 1920 res and add some AA, the GT tanks and loses to the GTS 640mb....in some games anyway.

Damn, and I just sold my GTS. I would have SLIed but GTSs still cost a lot. Oh well, hopefully the RV670 will do a bit better...at least the XT version.

3 FPS difference isn't that big of a deal. Even the 24 FPS from the 640 OC'd GTS isn't really high enough to be considered playable anyways. On the other review sites, was the GTS playable @ 1920x1200 with AA? If not, it's a moot point. You'll probably need to go with a GTX or Ultra to be able to have high enough oomph to make it feasible to add AA past 1600x1200.

And remember, at TR, the GT is a stock reference card and the 640mb GTS is not. They really should have either used different GTS cards or at least put the effort into getting a hold of an eVGA superclocked 8800GT. Not sure what other reviews you are talking about, but did they do the same thing as TR did?

 

[CP5670]

The biggest difference seems to be that huge minimum framerate drop in Crysis, which is surprising as the memory differences aren't that large. It could be that the GTS just manages to get over the line at that resolution (which isn't exactly high) while the GT runs short.

 

[Denithor]

Originally posted by: CP5670
The biggest difference seems to be that huge minimum framerate drop in Crysis, which is surprising as the memory differences aren't that large. It could be that the GTS just manages to get over the line at that resolution (which isn't exactly high) while the GT runs short.

I think you're exactly right, the lower VRAM cards have exceeded their onboard memory capacity and overflowed into system memory. Note that the GTS320 performed appallingly in this test while the GTS640 was borderline acceptable. The 512MB cards were obviously hit by the limited VRAM available but not to the same extent as the 320MB card.

It will be very interesting to see more reviews of these cards on the Crysis platform to see if these early results hold up.

 

[alcoholbob]

Someone already posted a memory usage meter when playing Crysis on high 1280x1024 2xAA/16xAF, the memory usage was at 505MB.

Also remember the 8800GT has less ROPs than the 8800GTS meaning it has less AA power. I imagine up to 1900x1200 the GT should have no problems having a performance advantage over the GTS.

 

[thilanliyan]

Originally posted by: DarthV
3 FPS difference isn't that big of a deal. Even the 24 FPS from the 640 OC'd GTS isn't really high enough to be considered playable anyways. On the other review sites, was the GTS playable @ 1920x1200 with AA? If not, it's a moot point. You'll probably need to go with a GTX or Ultra to be able to have high enough oomph to make it feasible to add AA past 1600x1200.

I was also referring to other reviews such as the one from Guru3D. I guess I'm just disappointed that the GT isn't faster across the board than the GTS 640mb.

 

[1ManArmY]

I hope some info leaks out about the rest of the G92 batch of cards due in mid November GTS on 65 supposedly a blast playing Crysis.