Will ATi jump on the SLI bandwagon?

[digitalwanderer]

I have heard that there are actual working SLI rigs being tested, has anyone else heard that yet?

 

[DegustatoR]

It's my understanding that 3dmark05 will not be as GPU limited as 3dmark03 is. In that case if I'd want to waste enough money on a dual GPU setup, there would be theoretically little to nothing holding back from also going for a dual CPU config also.

If you plan to run 3DMark only then, well, yeah...

 

[Ailuros]

It's my understanding that 3dmark05 will not be as GPU limited as 3dmark03 is. In that case if I'd want to waste enough money on a dual GPU setup, there would be theoretically little to nothing holding back from also going for a dual CPU config also.

If you plan to run 3DMark only then, well, yeah...

I don't belong to the group of people that upgrade with one application in mind and especially not a benchmark. Want to tell me on the other hand how many out there upgrade or tweak/configure with the most insane sollutions available just in order to get a higher 3dmark score? I often have the feeling that if it'd be feasable to sit with a comp in a freezer, some would actually do that too....

Hell I probably spend around 1000$ average per year for the PC and that includes software also.

 

[duncan36]

I remember specifically the arguments Nvidia used to attack 3dfx's SLI. They seemed to be legitimate because as we all know Nvidia won and 3dfx lost.

You're wrong since you refer to NV's arguments on 3dfx's _multi-chip boards_, not _SLI_ itself. Current NV's SLI is more like Voodoo 2 SLI when you could _add_ another _card_ to the one you already own and get a nice speed bump from this. I do hope you remember how Voodoo 2 SLI demolished Riva TNT in every speed test back in 1998 or i'd say that your memory is somewhat fragmented

What are you talking about Nvidia specifically attacked 3dfx's dual board SLI as inefficent(true) and cumbersome, and basically said 'we'll give you better performance with one chip as they give you with 2 boards' and with the TnT2 they did.

 

[jvd]

I remember specifically the arguments Nvidia used to attack 3dfx's SLI. They seemed to be legitimate because as we all know Nvidia won and 3dfx lost.

You're wrong since you refer to NV's arguments on 3dfx's _multi-chip boards_, not _SLI_ itself. Current NV's SLI is more like Voodoo 2 SLI when you could _add_ another _card_ to the one you already own and get a nice speed bump from this. I do hope you remember how Voodoo 2 SLI demolished Riva TNT in every speed test back in 1998 or i'd say that your memory is somewhat fragmented

What are you talking about Nvidia specifically attacked 3dfx's dual board SLI as inefficent(true) and cumbersome, and basically said 'we'll give you better performance with one chip as they give you with 2 boards' and with the TnT2 they did.

they also were able to do that again against the voodoo 5 .

 

[DegustatoR]

What are you talking about Nvidia specifically attacked 3dfx's dual board SLI as inefficent(true) and cumbersome, and basically said 'we'll give you better performance with one chip as they give you with 2 boards' and with the TnT2 they did.

Your memory certainly works in mysterious ways

. Voodoo3 competed with TNT2 resonably well (though it's lack of features was already very apparent). It took NVIDIA one more year to make a card which was better all around (except for FSAA) than 3dfx's Voodoo5, which was _dual-chip board_, and not a card with SLI capabilities.

Basically dual-chip board is a way of improving performance when you're failed to make a competitive single chip. Look for Voodoo5, Volari and so on for examples. SLI is a way of offering your customers a great single-chip solution with an added feature which allows you... blah-blah-blah... i'm sure you know how to finish this phrase correctly.

So while there are NO dual-chip _boards_ from NV in the same prices as single-chip boards from ATI, there really aren't any reasons for what you're suggesting.

[edit] stupid spelling errors

 

[duncan36]

What are you talking about Nvidia specifically attacked 3dfx's dual board SLI as inefficent(true) and cumbersome, and basically said 'we'll give you better performance with one chip as they give you with 2 boards' and with the TnT2 they did.

Your memory certainly works in mysterious ways

. Voodoo3 competed with TNT2 resonably well (though it's lack of features was already very apparent). It took NVIDIA one more year to make a card which was better all around (except for FSAA) than 3dfx's Voodoo5, which was _dual-chip board_, and not a card with SLI capabilities.

Basically dual-chip board is a way of improving performance when you're failed to make a competitive single chip. Look for Voodoo5, Volari and so on for examples. SLI is a way of offering your customers a great single-chip solution with an added feature which allows you... blah-blah-blah... i'm sure you know how to finish this phrase correctly.

So while there are NO dual-chip _boards_ from NV in the same prices as single-chip boards from ATI, there really aren't any reasons for what you're suggesting.

[edit] stupid spelling errors

Dual chip designs are failure but SLI is genius. From an engineering standpoint thats simply not correct. There are more efficencies from a dual chip design, not even factoring in the physical impractacilities(heat, space, noise, power) of SLI.

 

[Fodder]

From an engineering standpoint thats simply not correct.

What about from a marketing/consumer standpoint?

 

[jvd]

From an engineering standpoint thats simply not correct.

What about from a marketing/consumer standpoint?

well with only a 1.24x improvement over a x800xt pe its not going to convince many consumers huh.

 

[digitalwanderer]

What about from a marketing/consumer standpoint?

SLI is a PR departments wet dream, I'm almost afraid at what nVidia's will do with it. :?

 

[DemoCoder]

There is an either-or fallacy going on in these threads. With respect to 3dfx, they went with SLI in lieu of using bleeding edge memory, process, and bus width.

I don't think anyone is arguing that 2 mid-range cards in SLI are more efficient than 1 high range card. But, given that NVidia and ATI are using the same memory, similar processes, similar bus widths, and similar # of pipelines, once these factors are maxed out, the only way to scale further is go SLI.

Once ATI and Nvidia hit the limits of the current memory available, and current transistor densities, how else do you scale performance? 512-bit bus? I don't think so. Tweak the IMR further? The IMR and shaders are already fairly efficient. There's no much left to do except for hax (drop filtering adaptively, etc)

If you're not satisifed with the highest end card each IHV offers, than SLI is the only alternative for more GPU performance.

 

[duncan36]

From an engineering standpoint thats simply not correct.

What about from a marketing/consumer standpoint?

For the 1-2% of users who have the hardware to run SLI and are capable/want to install 2 cards in their system its a consideration.

Of course the danger is you spend too much engineering time with SLI and your rival makes a Geforce2 while you're stuck with a Voodoo3. In that case SLI becomes more of a hindrance than an advantage.

 

[DegustatoR]

For the 1-2% of users who have the hardware to run SLI and are capable/want to install 2 cards in their system its a consideration.

That's for now. How about some forward-thinking?

Of course the danger is you spend too much engineering time with SLI and your rival makes a Geforce2 while you're stuck with a Voodoo3. In that case SLI becomes more of a hindrance than an advantage.

Show us any signs of someone doing a chip which would compare to 6800U as GF2GTS was compared to V3. Where? Who? When??!!

So all your talk is based on absolutely nothing for now. Then again you gotta consider todays NV's engeneering resources and compare them to 2000's 3dfx's resources. I doubt they'll even be in the same league... Not to mention, that NV had every 3dfx SLI-related patent and technology when it's decided to include SLI in PCIE NV4x cards line.

 

[DemoCoder]

I don't think there is much danger of that. Unlike the 3dfx solution, the dual-PEG solution doesn't require much HW engineering since the engineering has been effectively "outsourced" to VIA, Intel, and others who will produce dual PEG chipsets/Motherboards. Most of the work is in the software drivers, hardware wise, it's just the addition of a high-speed private bus between the cards for data transfer.

People attempting to do dual cores AGP style had to effectively hack-in a solution that would make two cores appear as one AGP device via a bridge chip, or just live with the fact that some OS's (e.g. NT) couldn't deal with dual AGP devices.

The situation is much different now because there's a different bus architecture (PCIE) and OS support to make handling these scenarios easier, and NV has much more resources available than 3dfx had, so I doubt they are "putting all their eggs in one basket" Hell, NV could just put all the old 3dfx engineers they hired on it.

 

[duncan36]

I don't think there is much danger of that. Unlike the 3dfx solution, the dual-PEG solution doesn't require much HW engineering since the engineering has been effectively "outsourced" to VIA, Intel, and others who will produce dual PEG chipsets/Motherboards. Most of the work is in the software drivers, hardware wise, it's just the addition of a high-speed private bus between the cards for data transfer.

People attempting to do dual cores AGP style had to effectively hack-in a solution that would make two cores appear as one AGP device via a bridge chip, or just live with the fact that some OS's (e.g. NT) couldn't deal with dual AGP devices.

The situation is much different now because there's a different bus architecture (PCIE) and OS support to make handling these scenarios easier, and NV has much more resources available than 3dfx had, so I doubt they are "putting all their eggs in one basket" Hell, NV could just put all the old 3dfx engineers they hired on it.

Amazing how Nvidia and their ilk whos mantra was single-chip, single-chip, single-chip back in the day are now shouting dual-board, dual-board, dual-board.
SLI is interesting but the real battle is fought in the single chip arena anyone who thinks any different is kidding themselves.

 

[trinibwoy]

Look it all boils down to who can deliver the best single chip solution, SLI is an interesting thing but the real war is fought in single chip designs. Whoever hits a home run with their single chip design wins the war.

Yes it does but how does that make SLI irrelevant. At least Nvidia is providing the option. Most of the anti-SLI rhetoric seems to come from people who don't realize that it is something 'extra'. It's an 'extra' option to those that we already have. If you don't have the $$$ or if it turns out to be bunk then we still have all the single card options that we've had before. The only loser if SLI doesn't work is Nvidia - and we are not in a position to know how big a loss that would be.

 

[duncan36]

Look it all boils down to who can deliver the best single chip solution, SLI is an interesting thing but the real war is fought in single chip designs. Whoever hits a home run with their single chip design wins the war.

Yes it does but how does that make SLI irrelevant. At least Nvidia is providing the option. Most of the anti-SLI rhetoric seems to come from people who don't realize that it is something 'extra'. It's an 'extra' option to those that we already have. If you don't have the $$$ or if it turns out to be bunk then we still have all the single card options that we've had before. The only loser if SLI doesn't work is Nvidia - and we are not in a position to know how big a loss that would be.

SLI is ugly and brutish, I personally would never have 2 videocards in my PC.

 

[trinibwoy]

SLI is ugly and brutish, I personally would never have 2 videocards in my PC.

Exactly!! You personally choose not to go SLI. Now does its mere existence affect you in any way? Nope. Is it good to have options for those of us who will consider them? Yep.

It's similar to arguing with someone who lives in Manhattan and decides to buy a Ferrari. Sure it's not a sound economic or practical move but it doesn't stop you from going out and getting a Kia does it?

 

[DemoCoder]

SLI is interesting but the real battle is fought in the single chip arena anyone who thinks any different is kidding themselves.

None of the proponents of SLI that I am aware are saying "don't build the best single chip, build an average single chip and combine two of them"

What SLI proponents are asking for is for NVidia and ATI to build the best chips possible, but support board vendor options for deploying those chips in dual or more scenarios. We've got a bus capable of supporting multiple vid-cards now, just like in the old days *before AGP*.

Why not add in a feature in the device drivers to allow one to use two cards for a) dual screen output b) higher resolution AA/super sampling and c) faster rendering?

The anti-SLI people are posing a false dichotomy (SLI or best single chip): that there is a parallel between Nvidia's SLI and 3dfx's SLI when it is simply not the case. NVidia is *NOT* using SLI to avoid going with faster memory or a wider bus. They are putting the same memory and bus width on their cards as ATI, and, they are shipping chips just as featured filled, if not more feature filled, than ATI's.

So the comparison with the Voodoo5 situation does not hold. It's not "SLI instead of more feature rich and faster single chip", it's "build the fastest and most feature filled single chip, *and* support different PCI-E motherboard configurations"

How anyone can be against users having more options and configurations for video cards is beyond me.

 

[PatrickL]

The presse release Nvidia issued yesterday was a shot in their foot. They perfectly demonstrated why it is not the future with a score so close from ATI best cards.
I know it is in 1024 768 but it is nvidia that provided theses numbers not me.