Will there be a comparable Nvidia multi GPU card to the 4870x2?

Mojo · 21 Jun 2008 at 01:25

Both companies bring out dual GPU cards to fight for the performance crown, but with the imminent release of the 4870x2, I begin to wonder, will Nvidia be able to compete? Granted they have the most powerful single GPU card out there, but just look at the power requirements. Several websites didn't even bother trying dual GPU in conventional SLI as the power requirements were too high for most if not all commercially available consumer PSUs. Will this prevent them from producing a comparable dual GPU card?

So, I ask the forum, do you think Nvidia have shot themselves in the foot with such a monster as the GT200? Technology is slowly moving in the direction of multi GPU systems, has the crippling power requirements of the GTX280 prevented any possibility of a competitor to the 4870?

Lightnix · 21 Jun 2008 at 01:40

Will this prevent them from producing a comparable dual GPU card?

Short answer: Yes.

Long answer:

Well it is long.

JediFragger · 21 Jun 2008 at 01:54

ePeen f.t.w.

Tom|Nbk · 21 Jun 2008 at 01:56

The cores are huge would need a shink before they get two on one PCB or a GX2 jobbie with more heat than the sun.

helmutcheese · 21 Jun 2008 at 01:57

Single Core, Single Cards is still where its at today.

I (and read others here same) do not want issues/lack of full support with SLI/Crossfire etc.

Someday it will be Dual Cores but more like a C2D 2 Dies on the Card not like now with 2 CPU like objects on the Cards (cant think of correct terms), tired and bedtime, bit early for me but had and early and busy day.

Zzzzzzzzzzzz.

Lightnix · 21 Jun 2008 at 02:02

helmutcheese said:
Single Core, Single Cards is still where its at today.

I (and read others here same) do not want issues/lack of full support with SLI/Crossfire etc.

Someday it will be Dual Cores but more like a C2D 2 Dies on the Card not like now.

See, the problem with the 'single die dual core' approach to GPU design is that GPUs are already EXTREMELY multi-core, connected by one big bus. If you were to stick two of those cores on one die you'd have to add another bus for the chips to communicate. You might as well just stick it all on one core and be done with it.

The reason we get these X2/GX2 dealies is because it's cheaper than manufacturing an entire new GPU just to serve one segment of the market (and a small one at that), when they can get two mid-range products to serve as the high-end.

lordfletch · 21 Jun 2008 at 02:15

helmutcheese said:
Single Core, Single Cards is still where its at today.

I (and read others here same) do not want issues/lack of full support with SLI/Crossfire etc.

Someday it will be Dual Cores but more like a C2D 2 Dies on the Card not like now with 2 CPU like objects on the Cards (cant think of correct terms), tired and bedtime, bit early for me but had and early and busy day.

Zzzzzzzzzzzz.

times they are a changing.

Tom|Nbk · 21 Jun 2008 at 02:17

I think there is a big push for Multi GPU especially with ATi's new cards. Single chip is shortly going out of fashion. People now want cheap multigpu soltion that works well and offers good performance for their cash.

Monolithic GPU's are a thing of the past IMO.

thegoonden · 21 Jun 2008 at 02:59

Multi GPU's will always be a peen-only thing until drivers are released which make them work as one.
I don't want to buy a card to find that most of my games can only use half of it.

Defiance_co · 21 Jun 2008 at 03:31

Well i want a card that only uses half of it

That way it might last instead of having to get card after card with 3% performance increase increments

My 2p worth is in the future one chip multi core will be made, Crippled as needed for low end mid-range and high-end users etc

So it would be 1 chipper, Nobbled at birth to suit it's market, Makes sense to me but i aint in the know it's just my futuristic opinion

Ciao

Def

Bobly · 21 Jun 2008 at 03:42

If they drop to 55nm fabrication process then we should see a drop in consumption and heat which would make a x2 feasible which would make it a direct competitor to the 4870x2.

Phatzy · 21 Jun 2008 at 03:47

wtf will that fit in any large case?

Defiance_co · 21 Jun 2008 at 03:50

Bobly, Compared to the size now (pretty large) how much smaller would the 55nm make it ?

Because looking at the cards the lengths now are at about max really for mobo's to take, So unless somewhat smaller surely they are going to struggle because of the size (obvious)

Ohh and why do we not see cards x2 with chip/pcb/chip eg: like a burger ?? wouldn't that make x2 etc easier even with large chips ? (some kind of funky pipe cooling etc but hope you get my drift)

Ciao

Def

Bobly · 21 Jun 2008 at 12:02

Defiance_co said:
Bobly, Compared to the size now (pretty large) how much smaller would the 55nm make it ?

Because looking at the cards the lengths now are at about max really for mobo's to take, So unless somewhat smaller surely they are going to struggle because of the size (obvious)

Ohh and why do we not see cards x2 with chip/pcb/chip eg: like a burger ?? wouldn't that make x2 etc easier even with large chips ? (some kind of funky pipe cooling etc but hope you get my drift)

Ciao

Def

I think the reason for this is that PCBs make use of both the front and back of the board, with components on the front and loads of connections on the back. If you want to sandwich two GPUs on either side of a same PCB you'd have to re-route everything and it would be hell.

They could always have two PCBs back to back or barely touching but that bring in a new heat issue. While the fans would be able to cool the bottom and/or top card, the space in between them would get unbearably hot with low ventilation.

If they're spaced far enough apart to fit a heatsink links to the cooling system then I think you will have lost any benefit of having them sandwiched, which is why they're both facing the same way

Regarding the actual size, the board wouldn't have to be any longer, however I do see a problem in it's width (as in how many PCI lanes it takes). While the GX2 took up 2 lanes and was only able to use a small amount of the back of the computer chassis to expulse air, most of it getting blown into the case, make the GTX 280 an X2 would also lead to problem in that the board would either have to span 3+ lanes or cut back on the huge vents at the back.... Could the days of a 3 line GPU be soon? ^^

Durzel · 21 Jun 2008 at 12:19

If you think about it - the cards cant get any bigger physically, the top end ones are already the width of motherboards (and then some in some cases).

Multiple-card configs are always going to be around but they aren't an option for a significant percentage of the consumer base - your average Joe Punter just wants a "very fast graphics card" - they don't want to have to deal with more than one card and the necessary ancilliary requirements that go with it (Xfire/nForce mobos, etc).

Look at the direction CPUs have gone - wasn't so long ago that it was single-core, with punters wanting more processing power having to fork out for motherboards with more than one socket. Nowadays dual & quad core are the standard even on the budget chips.

The future is die shrinkage and multiple GPU core per PCB imo.

BubbySoup · 21 Jun 2008 at 12:26

helmutcheese said:
Single Core, Single Cards is still where its at today.

I (and read others here same) do not want issues/lack of full support with SLI/Crossfire etc.

Someday it will be Dual Cores but more like a C2D 2 Dies on the Card not like now with 2 CPU like objects on the Cards (cant think of correct terms), tired and bedtime, bit early for me but had and early and busy day.

Zzzzzzzzzzzz.

Dinosaur

Cheaper and simpler to manufacture lots of smaller cores and place them on a card rather than a single, massive core. Can you even imagine a chip, say 1.5 or 2x the physical size of the 280 :eek:

Personally, I'd never go back to a single core GPU setup now - been multi-GPU for far too long

Skyboat · 21 Jun 2008 at 14:49

Well if the 65nm->55nm die shrunk of the 9800GTX is anything to go by, the die area only shrunk by 1mm sq. That's still not enough to make the GT200GX2 viable.

helmutcheese · 21 Jun 2008 at 15:06

BubbySoup said:
Dinosaur

Cheaper and simpler to manufacture lots of smaller cores and place them on a card rather than a single, massive core. Can you even imagine a chip, say 1.5 or 2x the physical size of the 280

Personally, I'd never go back to a single core GPU setup now - been multi-GPU for far too long

Thats cool but you can read weekly in here that peeps want a Single card that acts as a Single card right now, not SLI or other.

Obv there is the other half they do want it.

I will get it when its supported much better and its really needed.

mrk · 21 Jun 2008 at 15:20

Phatzy said:
wtf will that fit in any large case?

lawls...

Duff-Man · 21 Jun 2008 at 16:49

BubbySoup said:
Dinosaur

Cheaper and simpler to manufacture lots of smaller cores and place them on a card rather than a single, massive core. Can you even imagine a chip, say 1.5 or 2x the physical size of the 280

Personally, I'd never go back to a single core GPU setup now - been multi-GPU for far too long

There are fundamental practical issues with multi-GPU setups - requiring an equal set of memory for each GPU, for example.

While a dual-GPU setup is a nice way to increase performance (ignoring the more subtle performance issues for now), greater parallelism is just not practical. Having an 8-GPU card with 8Gb of memory, which can load only 1Gb of textures, is not cost effective.

Large 'monolithic' GPU setups are here to stay - at least until inter-GPU bandwidth and latencies can be improved by several orders of magnitude, allowing all the parallel GPUs to read from the same memory stack - and that is way into the future.