The right price for the GTX200 family...

[ChrisRay]

I'm curious, then. Why do you think ATI chose to ignore VRAM increase and bus width and instead focus solely on bandwidth by tossing GDDR5 on the PCB?

I dont know? How is tossing GDDR5 with a 256 bus focusing on bandwith anymore than GDDR3 with a 448 Bit bus? The bandwith is basically identical give or take. The GTX 260 has about 114 gigs of bandwith while the GTX 280 has about 140. I'm not going to get into a cherry picked benchmark war with you. Your completely missing the point. You dont need to justify your 512 meg cards to me.

Its completely irrelevant to what I am saying. If you run out of memory. No amount of bandwith is going to save your performance because texture/AA data is now being passed through the system bus . Where your bandwith is limited by the PCIE clocks. The bandwith provided by the GPUS doesnt help[ in this case because data is being "waited" on by the system bus. This is not something new or different. You either have the memory or you dont. The entire point of the PCIE bus accessing system memory ((And AGP for that matter)) was to allow for framebuffer overflow to utilize system memory. But the cost is high.

Memory utilization in games is going up. Not down.

[Xion X2]

I dont know? How is tossing GDDR5 with a 256 bus focusing on bandwith anymore than GDDR3 with a 448 Bit bus?

It's not. My perspective was that GDDR5 is more expensive, and ATI could've spent that money to increase VRAM and going with a higher bus, instead.

Originally Posted by ChrisRay

I'm not going to get into a cherry picked benchmark war with you. Your completely missing the point. You dont need to justify your 512 meg cards to me.

"Cherry picked?" You need to drop the gloves a little. I could pick practically any benchmark on that site (or many others) to show how well Crossfire scales at 2560x with or without AA.

I told you before that I understand how VRAM works, yet you felt compelled to go off on another tangential graphics lesson. All I am saying is that, at this point in the game, this "512 meg" thing seems to be overstated--at least when it comes to the 4800 series. We have a ton of games on the market running at the highest possible resolution with high amounts of AA, and it's not even phasing these cards. Whether or not it will in the near future, we will have to wait and see.

And I'm not trying to justify my "512 meg" cards to you. Like I care? If I really cared what you thought about me then I would've bought a pair of Nvidia cards, wouldn't I?

[ChrisRay]

Then perhaps your question about "Why did ATI choose a narrow bus and GDDR5" would be better suited for ATI and not me.

I told you before that I understand how VRAM works, yet you felt compelled to go off on another tangential graphics lesson. All I am saying is that, at this point in the game, this "512 meg" thing seems to be overstated--at least when it comes to the 4800 series. We have a ton of games on the market running at the highest possible resolution with high amounts of AA, and it's not even phasing these cards. Whether or not it will in the near future, we will have to wait and see.

I see that as pretty short sighted. It doesnt take alot of memory seepage for performance to nose dive. What may scale well may not scale well in the future due to a completely artificial memory bottleneck. After using a significant amount of SLI configurations with 512 Megs, 768, 896 and 1 gigabyte. The higher the memory. The more consistent the experience. Due to never forcing texture data across the system bus. I've seen 9800GX2 Quad SLI cards deliver absymmal performance for the resolutions they are supposed to be augmenting. Once you start passing 1600x1200 the 512 cards start to slowly lose ground regardless of how well they scale. Yet they could be 80% faster than a 768 meg card at specific settings. The higher the memory. The less compromise.

I felt the "need" to point it out because any amount of texture data being constantly sent through the PCIE bus is a cause of what many people refer to as "Stuttering". You also are saying things like "Bandwith can make up for lower memory amounts". And that makes zero sense because its not true. So you either need to clarify or I'm going too object to such a statement. Because its completely untrue.

[Xion X2]

Originally Posted by ChrisRay

Then perhaps your question about "Why did ATI choose a narrow bus and GDDR5" would be better suited for ATI and not me.

Perhaps so. I just thought we might could have a civil, friendly discussion about it. Guess not.

After using a significant amount of SLI configurations with 512 Megs, 768, 896 and 1 gigabyte. The higher the memory. The more consistent the experience. Due to never forcing texture data across the system bus.

I don't argue that's the case most of the time, as it makes sense. I'm just seeing different things when it comes to the 4800 series. Like the following, for another example.

Here we have Oblivion running at 2560x this time with 8xAA and HDR enabled. Yet the 4870 still flies through it like a bat out of hell--scoring 4fps higher than the GTX280.



High resolution, high AA, and HDR all combined. The very things that eat up VRAM, yet the 4870's not even phased by it.

Now I'm not saying it won't be VRAM limited at some point in the future, but the ceiling on this card appears to be a bit higher than many think it is. If current games run just fine at 2560x and 8xAA with HDR, then next-gen games, barring they don't change too much (which most of the time they don't except for the occasional Crysis), should run pretty well if you just lower the settings a little.

[CaptNKILL]

Originally Posted by Xion X2

Perhaps so. I just thought we might could have a civil, friendly discussion about it. Guess not.

Is this a joke?

[Xion X2]

Originally Posted by CaptNKILL

Is this a joke?

Not a joke at all--why you say? Because he owns Nvidia and I own ATI?

Guess I'm one of those Rodney King "can't we all just get along" types...

In my opinion, most guys take this sh*t too personal. Some will fight you to the death over a graphics card as if it was their firstborn child, or something. None of that insecure BS concerns me; my only goal is to chat with fellow enthusiasts about something that I enjoy.

[ChrisRay]

Originally Posted by cvearl

Thank you Chris. I will run the exact same test at home. I love comparisons regardless of the outcome. I like truth and facts.

I'm such a friggin nerd.

So on my 4870 it's Edge detect 4xAA right?

Set everything to Very High in Crysis? That's it? Did you want 16xAF enabled as well or is that moot as you say.

C.

If you want the complete preview. You can see it here. Its a little sloppy and rushed compared to some of my work. I'll be fixing/correcting errors over time.

[CaptNKILL]

Originally Posted by Xion X2

Not a joke at all--why you say? Because he owns Nvidia and I own ATI?

Guess I'm one of those Rodney King "can't we all just get along" types...

In my opinion, most guys take this sh*t too personal. Some will fight you to the death over a graphics card as if it was their firstborn child, or something.

No I was actually taking a cheap jab at you because you've been rather testy yourself in this discussion.

All in fun of course.

[cvearl]

Originally Posted by ChrisRay

A single GPU might not be able too. But a multi GPU can. Once you start looking at multi GPU. The situation becomes even more problematic because you hamstring hardware potential by the limits of its memory.

These are benchmarks I took today as I am preparing my 9800GTX + preview. This graph is actually part of that preview. But its interesting because it specifically illustrates the problem.



P.S ignore the 16xAF. It doesnt work with very high. Its just forced through the control panel.

OK I just ran Crysis tests at what I think is the same setting??

Play Time: 103.16s, Average FPS: 19.39
Min FPS: 11.64 at frame 144, Max FPS: 24.42 at frame 872
Average Tri/Sec: -8215555, Tri/Frame: -423771
Recorded/Played Tris ratio: -2.16
TimeDemo Play Ended, (3 Runs Performed)
================================================== ============

Completed All Tests

<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>

10/07/2008 6:29:08 PM - Vista

Run #3- DX10 1600x1200 AA=16x, 32 bit test, Quality: VeryHigh ~~ Overall Average FPS: 19.39

[Xion X2]

Originally Posted by CaptNKILL

No I was actually taking a cheap jab at you because you've been rather testy yourself in this discussion.

I'd say I've been pretty tame after being repeatedly misquoted and told I'm trying to "justify" my "512 meg" purchases. You're barking up the wrong tree here, I think.

[ChrisRay]

I didnt use the time demo btw. This was taken from in game. I really dunno how the cards perform in a time demo cvearl

Chris

[Medion]

ATI and NV seem to handle memory differently. Comparing their hardware on just memory, when there are other variables, just isn't going to work. When the 1GB 4870 comes out, compare that to the 512MB version, and then we'll see if there's a 512MB limitation.

After using a significant amount of SLI configurations with 512 Megs, 768, 896 and 1 gigabyte. The higher the memory. The more consistent the experience.

I'm sorry Chris, but I have to disagree with this statement. If it were the same hardware and different configs, then yes, I could see this being a valid test. But comparing a 512MB 9800GTX to a 1GB GTX280, and saying the framerate differences are ONLY due to memory, is complete bull. There are numerous other hardware differences that affect it. Take two cards that are the same hardware, and compare them using two different memory configurations. Right now, the only high end single GPU we can do this with is the HD4870, well, once the 1GB version is released.