[Dave Baumann]

Errr, I wouldn't pay too much attention to whats out there (or not now) until you see the reviews.

[Dave Baumann]

Originally Posted by no-X

edit: didn't Dave say, that ATi will not need XT-PE?

No, its already been pointed out before that I didn't say anything of the sort.

[no-X]

Thanks

[SugarCoat]

Originally Posted by trinibwoy

Hmmmm that doesnt gel with a lot of hints being dropped about higher-than-expected clocks. 470 would be extremely disappointing based on leaked info on higher voltage and tweaked process.

but eVGA has taken quite a stand in the AIB market. Look at the GTX for example, they allowed bios changes after the launch to increase stock clocks just to compete. I dont think their card would be any worse off then anyone elses, stock. I was expecting 550 stock ):

[trinibwoy]

Originally Posted by SugarCoat

but eVGA has taken quite a stand in the AIB market. Look at the GTX for example, they allowed bios changes after the launch to increase stock clocks just to compete. I dont think their card would be any worse off then anyone elses, stock. I was expecting 550 stock ):

I just want to know why there would be a need for that cooler on a 470Mhz GTX. How are they cooling the RAM on the back of the PCB anyway?

[Ailuros]

Originally Posted by SugarCoat

im wondering how far that 1:1 performance increase goes before it trickles down. And what effect the memory will have.

If it would trickle down NVIDIA would be in quite a lot of trouble in H1 2006, unless you believe they'll stop here and leave R580 unchallenged.

What will that translate into when real world performance is measured, is my curiosity. Will it be close to 40%? Or merely ~25%. How much improvment will the memory give in the most taxing of bandwidth situations in todays titles and benchmarks? Can the core make use of it all? Can games? Etc...

I guess soon we're all going to be wiser about it, but I don't see anything impossible in that paragraph either given the hypothetical increases in fill-rate and memory bandwidth. I estimated an average increase according to the fill-rate, based on my own experiments on the 7800GTX 256MB and it's behavious so far.

The highest clocked GTX appears to be using 460 cores and 1300MHz. Everything else that was higher, is discontinued. All the AIBs have tried to make a gap between the $450-500GTX and this new card. That was very intentional, now all that remains to be seen is if it was done for sales or to remove the threat of 2 competing products from the same company.

Witness, how the GeForce 7800 GTX with 486MHz engine speed and 1.35GHz memory manages to leave behind the RADEON X1800 XT in loads of benchmarks as we share our experience about ASUS Extreme N7800 GTX TOP graphics card.

http://www.xbitlabs.com/articles/vid...en7800gtx.html

That's a 13% fillrate and a 12.5% bandwidth increase.

BF2 as just one example, sees an increase on the higher clocked board of 12.5% in 1600/4xAA/8xAF.

http://www.xbitlabs.com/articles/vid...7800gtx_8.html

[Jawed]

It's interesting that the X1800XL at 1600x1200 4xAA/16xAF is the same as the OC'd 7800GTX in that test.

Somehow the 1GHz memory on the XL is matching the 1.35GHz memory on the GTX TOP.

It seems to me that BF2 is one of the games that ATI actually tweaked the drivers for X1k.

Is BF2 a shader heavy game? It looks pretty old-fashioned to me. Is it actually just vertex heavy?

Jawed

[Chalnoth]

Or maybe the XL is a bit better at handling occasional long texture read latencies?

[Jawed]

Without AA/AF the XL is slower than the GT, so I don't think it's texturing latency.

EDIT: Hmm, I suppose I'm forgetting that AF reads more texture data than non-AF - hmm. At least I presume that's the case. In which case texturing latency would start to cut hard.

Jawed

[Chalnoth]

Originally Posted by Jawed

EDIT: Hmm, I suppose I'm forgetting that AF reads more texture data than non-AF - hmm. At least I presume that's the case. In which case texturing latency would start to cut hard.

More than just that: an AF read can take up to 32 times as long on current hardware as a normal texture read just to do the filtering itself, assuming perfect texture cache usage (comparing 16-degree anisotropy with trilinear vs. no anisotropy on the base texture).

[Ailuros]

Originally Posted by Chalnoth

More than just that: an AF read can take up to 32 times as long on current hardware as a normal texture read just to do the filtering itself, assuming perfect texture cache usage (comparing 16-degree anisotropy with trilinear vs. no anisotropy on the base texture).

Feasable theory, especially if one notices the differences to the noAA/AF scores. By the way that review was written with the 78.01 drivers. Their latest 7800GT/SLi review uses the 81.85 set. That particular timedemo doesn't favour though the performance increases that came with the 8x.xx class of drivers for BF2. On average I'd say there is a 10-15% performance increase; some timedemos show even higher increases, but the former estimated percentage sounds more like it. Totally unrelated to the Radeon performance since it could be also higher in other instances/timedemos than the GFs.

[Jawed]

Sadly I can't find a B3D test of GTX and X1800XL/XT for just AF (no AA) that might illuminate this BF2 texture latency question.

The XT is only 20% faster than the XL at 1600x1200 4xAA/16xAF, despite having a core that's 25% faster and memory that's 50% faster. (NB: the frame-rate cap at 90fps may be distorting these results somewhat, across the board.)

So would that indicate that latency is limiting the XT? It does seem like it.

Jawed

[SugarCoat]

Originally Posted by trinibwoy

I just want to know why there would be a need for that cooler on a 470Mhz GTX. How are they cooling the RAM on the back of the PCB anyway?

I was under the impression its using the new ball package, and there should only be 512 ram on one side, same as the 1800XT.

[Mintmaster]

Originally Posted by SugarCoat

I was under the impression its using the new ball package, and there should only be 512 ram on one side, same as the 1800XT.

I never even realized that about the X1800XT. Neat.

[makaveli87]

http://www.guru3d.com/article/Videocards/278/

scroll down to the graph. looks like 560/1600

[trinibwoy]

Originally Posted by makaveli87

http://www.guru3d.com/article/Videocards/278/

scroll down to the graph. looks like 560/1600

Hmmmm what's that doing in there?

[Chalnoth]

Heh, that graph doesn't make any sense. It lists the memory speed as "(900) 1600"

[SugarCoat]

Originally Posted by Chalnoth

Heh, that graph doesn't make any sense. It lists the memory speed as "(900) 1600"

it does if you read it as (rated speed) Effective speed. I would be a little surprised to see every AIB at 1700 or 1800MHz, i still think it will be 1600MHz on the reference.


560gpu speed...thats not confusing


XFX GEFORCE 7800 GTX 512MB PCI EXPRESS DDR3 (#PVT70FYDF9) (Retail) (NO STOCK - Estimate arrival on 11/30/05).
The ETA is provided by the manufacturer as a guideline, but is not a guarantee.

http://www.mwave.com/mwave/viewspec....iteria=3000415

[Pete]

Originally Posted by Ailuros

Feasable theory, especially if one notices the differences to the noAA/AF scores. By the way that review was written with the 78.01 drivers. Their latest 7800GT/SLi review uses the 81.85 set. That particular timedemo doesn't favour though the performance increases that came with the 8x.xx class of drivers for BF2. On average I'd say there is a 10-15% performance increase; some timedemos show even higher increases, but the former estimated percentage sounds more like it. Totally unrelated to the Radeon performance since it could be also higher in other instances/timedemos than the GFs.

TR's 6800GS review includes FRAPS testing of BF2 with the 81.87 Forcewares. The GT is dead even with the XL at 16x12 with 4xAA and "High Filtering." The XT hands the GTX its ass on a platter, though. The interesting part is comparing this to the Xbit review, which shows the GT and GTX trailing the XL and XT by similar margins. In TR's review, the GT pulls even with the XL, but the GTX, while (slightly) ahead of the GT and XL, is way behind the XT.

Maybe this is just the extra RAM talking. We'll see in a week, when Scott updates that graph with the 512MB GTX.

[RobertR1]

Originally Posted by SugarCoat

it does if you read it as (rated speed) Effective speed. I would be a little surprised to see every AIB at 1700 or 1800MHz, i still think it will be 1600MHz on the reference.


560gpu speed...thats not confusing


XFX GEFORCE 7800 GTX 512MB PCI EXPRESS DDR3 (#PVT70FYDF9) (Retail) (NO STOCK - Estimate arrival on 11/30/05).
The ETA is provided by the manufacturer as a guideline, but is not a guarantee.

http://www.mwave.com/mwave/viewspec....iteria=3000415

$799.........At this point I think the e-retailers just want to one up each other on who can sell it at the highest price.

[SugarCoat]

Originally Posted by RobertR1

$799.........At this point I think the e-retailers just want to one up each other on who can sell it at the highest price.

if its at those bios speeds i can believe it. Im not to sure 550+ will be as easy for this core as everyone thought. The water blocked'd KO's on evga's forums appear to be shooting for 550MHz and 900MHz memory @ a $650 retail. That says something to me, as in the core isnt as flexible as some may think. Not to mention that memory has got to be costing quite a bit.

Heres the block:

[Chalnoth]

The memory is more likely to be holding the product back (as far as volume goes) than the core, considering that we know that the 7800 GTX already clocks close to the rumored speeds on less exotic cooling measures than are rumored to be used for this part.

[Junkstyle]

This is sounding more and more like Nvidia doesn't want to ship a real product here. Maybe this is just a stunt for benchmark coverage.

[Jawed]

Originally Posted by Pete

TR's 6800GS review includes FRAPS testing of BF2 with the 81.87 Forcewares.

Blimey, the "median low" framerate on the X1800XT is roughly the same as the average framerate on the GTX.

The XT's median low is around 50% higher than the XL's, which is in line with the memory bandwidth difference.

Interesting.

Jawed

[Ailuros]

Originally Posted by Pete

TR's 6800GS review includes FRAPS testing of BF2 with the 81.87 Forcewares. The GT is dead even with the XL at 16x12 with 4xAA and "High Filtering." The XT hands the GTX its ass on a platter, though. The interesting part is comparing this to the Xbit review, which shows the GT and GTX trailing the XL and XT by similar margins. In TR's review, the GT pulls even with the XL, but the GTX, while (slightly) ahead of the GT and XL, is way behind the XT.

Maybe this is just the extra RAM talking. We'll see in a week, when Scott updates that graph with the 512MB GTX.

Hmmmm I see they've used the latest patch for BF2. I never installed it to be honest; the former patch I installed made things even worse than out of the box for me. Wonder if I should dare

By the way it could easily also be the frequency talking; if the 512MB 7800 truly comes at =/>550MHz, any possible performance increase will be more due to the higher frequency IMO, than the larger framebuffer.