Will ATi jump on the SLI bandwagon?

dksuiko said:
jvd said:
dksuiko said:
Why would you factor in the R500 and NV50 cards when the dual PCI-E boards come out? Those cards would obviously not be out yet at that time, so it's obviously just going to be a tech demo right? In fact, why discuss R500 and NV50 now - they're certainly nothing to talk about now nor when the dual PCI-E boards are available.

Why are we talking about sli right now , i can't buy an sli system so we should lock the thread.

The r520 cards are not all that far away , some are saying perhaps first half of next year , which is only about 8 or 9 months away ,if dual boards peg boards don't come out till around x mass , its just going to make the sli window on those cards smaller.

Dual PCI-E boards aren't that quite far away, either. Less than 8 to 9 months, that's for sure. Yet, you used the "it's not available now, so don't discuss it" argument when arguing against it. But then when it's time to discuss R520, which is even further away, it's all flip-flopped for you. I just thought it'd be funny to point out your hypocrisy and how you contradict yourself. LOL!

Dual PCIe boards are already out (Intel Tumwater). Both nVidia & Via have said they are releasing their dual boards in Nov-Dec. Nvidia has already announced & released the 6600. SLI is not far in the future but just around the corner if not here already. As Brent from Hard OCP has said, "it is not the SLI system or the video cards that are hard to come by, it is the connectors between the cards, which are harder to find than hen's teeth". nVidia is apparently guarding them very closely until they release their Nforce4 boards.
 
JVD is suggesting that instead of buying one NV40 and then upgrading to NV50, a cutsomer will buy a NV40, and then when NV50 arrives, will buy (a much cheaper by then) second NV40 for SLI.

So instead of a customer buying a full price NV40 and then a full price NV50, they will buy a full price NV40, then another NV40 which is a lot cheaper. Overall, Nvidia will lose out if customers upgrade down the line to a second NV40 in SLI rather than upgrade to NV50.

Not if Nvidia keeps prices for NV40s high enough.
I think what they will do in the future is roll out some SLI disabled NV40s. Those NV40s will drop in price like high end cards do. SLI NV40s will remain stable just very little below an NV50. It just depends on how fast their next poduct will be.
Of course the new product will offer SLI too.
I don't see a big problem here.

Antoher option besides that would be to phase out NV40s soon enough when the next products arrive.
No big deal - just business as usual.
 
can you setup an sli system where you dont have identical cards? Maybe have the third frame being rendered by the slower card? I guess it depends on how you divide the work. Just curious.

epic
 
epicstruggle said:
can you setup an sli system where you dont have identical cards? Maybe have the third frame being rendered by the slower card? I guess it depends on how you divide the work. Just curious.

epic
You can't, according to NVidia.
 
jvd said:
anyway i fully believe based off the jump from the r300 - r420 that the r520 will be almost 3 times the speed of the x800xt pe , not to mention that i will most likely have a more advanced feature set . So yes i can see it being faster than a dual 6800ultra set up , because that be only at best case 2x the performance of a x800xt pe .
I think neither R520 nor NV50 will be near 3 times X800XTPE performance. Memory bandwidth just isn't keeping up with that.
 
Can you SLI two of the same cards if they're different brands? Like a Gainward 6800 GT with a Chaintech 6800 GT? Or do they have to be identical cards?
 
jvd said:
anyway i fully believe based off the jump from the r300 - r420 that the r520 will be almost 3 times the speed of the x800xt pe , not to mention that i will most likely have a more advanced feature set . So yes i can see it being faster than a dual 6800ultra set up , because that be only at best case 2x the performance of a x800xt pe .

The next generation boards will not have a similar jump in performance to what we have seen with both the GeForce 6800Ultra and the Radeon X800 XT Platinum Edition this time around.
 
digitalwanderer said:
Can you SLI two of the same cards if they're different brands? Like a Gainward 6800 GT with a Chaintech 6800 GT? Or do they have to be identical cards?

That's one of the potential biggest caveats of nVidia's SLI, and I don't think anyone truly knows the answer - How will different brand of boards, BIOS revisions etc interact?

It'll be interesting to see just how well the solution works in real-world circumstances such as that, that's for sure.
 
bigz said:
The next generation boards will not have a similar jump in performance to what we have seen with both the GeForce 6800Ultra and the Radeon X800 XT Platinum Edition this time around.
That's my thinking too, we're looking at more of a 9700 to 9800 or 5800 to 5900 bump next.

One of the reasons I got an X800 pro when my 9700 pro died, I figured I'd be skipping the next 'generation' then since it'll just basically be a refresh....and I just read too many stories of people who never did regret buying their 9700 pros just when they came out since they had such a long life. (Hell, I'd still be on my 9700 pro if it hadn't died! Doom3 and FarCry were the only real games to challenge it and it could still run them pretty good, but I must admit the next gen cards are muchly better. ;) )
 
digitalwanderer said:
That's my thinking too, we're looking at more of a 9700 to 9800 or 5800 to 5900 bump next.

One of the reasons I got an X800 pro when my 9700 pro died, I figured I'd be skipping the next 'generation' then since it'll just basically be a refresh....and I just read too many stories of people who never did regret buying their 9700 pros just when they came out since they had such a long life. (Hell, I'd still be on my 9700 pro if it hadn't died! Doom3 and FarCry were the only real games to challenge it and it could still run them pretty good, but I must admit the next gen cards are muchly better. ;) )

Yeah, see this is kinda why I started this thread. What I want is to play games @ 1900*1200 with high AA & AF. With the X800XT/6800U now, you can play with 6AA & 16AF at lower resolutions because you get CPU limited, but you can't really play at the highest resolution (HDTV/big screen) with AA & AF. I am looking to get a PCIe system with AMD, but I'm trying to decide if I have to wait for the dual-PCIe boards to get that high performance or if I should just get a one slot PCIe board & stick in a R520.

I'll probably get the dual board just in case anyway (early next year), then a few months later get the R520. No, wait, cause the second I get a PCIe board I have to get a PCIe video card. I could wait until the R520 is released then upgrade everything at once, but my system is long in the tooth now............. :| sigh nm
 
trinibwoy said:
What new features do we expect from R520/NV50 that will make them more attractive than 2xNV40 ?

Image Quality as far as AntiAliasing over 4x is concerned. Image Quality as far as texture filtering is concerned, if what I read about horrible texture-shimmering on NV4x based cards isn't fixable.
 
trinibwoy said:
What new features do we expect from R520/NV50 that will make them more attractive than 2xNV40 ?

Mordenkainen said:
SM 3.a, SM 3.b ?

;)

:?

flick556 said:
Price, dxnext, motherboard/power requirements

Price and power requirements aren't really features you know ;) And I thought the next gen would be sticking to DX9 ?

BRiT said:
Image Quality as far as AntiAliasing over 4x is concerned. Image Quality as far as texture filtering is concerned, if what I read about horrible texture-shimmering on NV4x based cards isn't fixable.

Are we going to see something similar to 6x on the NV50? Or do you mean that 8x will provide higher performance than the current FSAA/MSAA mix? I haven't read anything on the root cause of the texture-shimmering but what's stopping them from fixing it in the refresh parts?
 
jvd said:
anyway i fully believe based off the jump from the r300 - r420 that the r520 will be almost 3 times the speed of the x800xt pe , not to mention that i will most likely have a more advanced feature set . So yes i can see it being faster than a dual 6800ultra set up , because that be only at best case 2x the performance of a x800xt pe .

R O F L, someone put this in the hall of fame for most ridiculous comments ever
 
Xmas said:
I think neither R520 nor NV50 will be near 3 times X800XTPE performance. Memory bandwidth just isn't keeping up with that.

People should listen more often to you; me including :oops:

---------------------------------------

On the topic: if ATI should truly have a multi-board config in it's plans after all and/or releases a R4xx based refresh late enough, it could only indicate that R520 shouldn't be expected as early as many believe or would like to believe.

I may be wrong but I'd expect to see R520 on low-k 90nm, unless something has changed at foundries lately and 110 is being made available for high performance chips too. If we're truly talking about low-k 90nm, then a great deal depends on the maturity of the manufacturing process.
 
Evans & Sutherland had a dual R200 card. they also have dual R300 and quad R300 cards.

SGI has highend visualization systems (Onyx 4 Ultimate Vision) with upto 32 ATI VPUs.

the R300 introduced the capability of having upto 256 VPUs working together. SGI is probably the best example of using some of that capability with the Ultimate Vision line.

It should be no problem for ATI to introduce PC gaming-grade cards with 2, or perhaps even as many as 4 VPUs, on board.

whether it's SLI or some other implementation (like what E&S and SGI have) doesn't really matter. as long as we can get a worthwhile boost in performance, both geometry-wise and pixelfill-wise, it should be great, provided it can be used practically, in games.
 
On the topic: if ATI should truly have a multi-board config in it's plans after all and/or releases a R4xx based refresh late enough, it could only indicate that R520 shouldn't be expected as early as many believe or would like to believe.

I may be wrong but I'd expect to see R520 on low-k 90nm, unless something has changed at foundries lately and 110 is being made available for high performance chips too. If we're truly talking about low-k 90nm, then a great deal depends on the maturity of the manufacturing process.

Since ATi is just barely stepping into the 110nm process with their X700, I really don't think they will use low-k 90nm for their next high end chip. It'll just be regular 110nm. At that size they can still put 300 million transistors on a same size chip as a R420, from what I've heard.

But, hear is my question: which would be better, in your opinion, a single R520 with 24 pipes, or 2 6800's in SLI? Or is that a rhetorical question until we get more info?
 
Back
Top