"Quad" SLI = Bad math?

Status
Not open for further replies.

Logan321

[H]ard|Gawd
Joined
Oct 9, 2003
Messages
1,900
Does anyone else notice that SLI means 2 GPUs, so Dual SLI would be 4 GPUs and Quad SLI would be 8 GPUs?

Dell's "Quad" SLI is 2 boards, and 4 GPUs. Someone's math doesn't add up.
 
there missing one word in that case, it would be core, "quad core" sli

thore
 
Logan321 said:
Does anyone else notice that SLI means 2 GPUs, so Dual SLI would be 4 GPUs and Quad SLI would be 8 GPUs?

Dell's "Quad" SLI is 2 boards, and 4 GPUs. Someone's math doesn't add up.

I dont think there are enough Dell super computer threads open, especially today. I think I will go start a new one.

Each card apparently has two gpu's but it looks to me as if its just an sli sandwich on top of an sli sandwich. So its 4 cards, 4 coolers, linked together in sli.

So its technically only 2 cards connected together. Just shitty mainboard integration on Nvidia's part. How Asus Managed to make a dual GPU card that only takes up one slot and have an external psu as opposed to this dell set up which takes up two slots is beyond me.

SLI doesnt mean 2gpu's SLI means Streamlined Link Interface. you can link as many gpu's as a motherboard and software can support.
 
No, you are wrong. I read up on the technology, and the cards are actually 2 cards, that use the 2 x16 lanes. But, somehow, the mobo actually gived EACH of the cards a x16 lane, not 4 x8 lanes, which is what you would think is what it would have. The only difference in this setup and 4 cards in 4 x16 connectors is that this setup just uses 2 x16 connectors to mount the 4 cards.
 
2 cards with 2 PCB's each = 4 PCBs
4 cores
2Gb of ram
each GPU has 16x lanes

though each pair uses a hub chip to translate that 2x 16 to a 16x lane to the mobo.

And whats cool is somehow the SLI connectors are alternating. the PCB's #1 & # 3 have an SLI connector and then 2 & 4 are connected with an SLI connector, also.
And thats weird, b/c pcb's 1 & 2 (and 3 & 4) are connected together with the "hub" chip which transfers the 2 x 16x lanes down to the 1 x 16 lane on that card (that has 2 pcbs).

so to retrace this:

pcb 1 & 2 are on same "card" and connected via new "hub" chip
pcb 3 & 4 are on same "card" and connected via new "hub" chip
then
pcb 1 & 3 are connected via an SLI bridge/connector
pcb 2 & 4 are connected via an SLI bridge/connector
 
chrisf6969 said:
2 cards with 2 PCB's each = 4 PCBs
4 cores
2Gb of ram
each GPU has 16x lanes

though each pair uses a hub chip to translate that 2x 16 to a 16x lane to the mobo.

And whats cool is somehow the SLI connectors are alternating. the PCB's #1 & # 3 have an SLI connector and then 2 & 4 are connected with an SLI connector, also.
And thats weird, b/c pcb's 1 & 2 (and 3 & 4) are connected together with the "hub" chip which transfers the 2 x 16x lanes down to the 1 x 16 lane on that card (that has 2 pcbs).

so to retrace this:

pcb 1 & 2 are on same "card" and connected via new "hub" chip
pcb 3 & 4 are on same "card" and connected via new "hub" chip
then
pcb 1 & 3 are connected via an SLI bridge/connector
pcb 2 & 4 are connected via an SLI bridge/connector

Thanks for elaboration on what I said, I couldn't think of how to put it, thanks!
 
Status
Not open for further replies.
Back
Top