NVIDIA get round to single board SLI with 55nm GT200
Sunday 21st December 2008, 01:21:00 PM, written by Rys
NVIDIA have pre-empted the release of GeForce GTX 295 in January at CES, with a set of sanctioned previews by some of the usual hardware-analysing suspects around the 'net.
The product implements single-slot SLI using two 55nm GT200s, NVIDIA using the shrink from 65 to get the physical characteristics needed to build a sensible product. Each GPU will have all ten clusters enabled, but one ROP partition and memory channel will be disabled.
That gives rise to headline numbers of 240 SPs, a 448-bit memory bus, 7 quad ROPs and 896MiB of connected GDDR3 DRAM memories, all per chip.
NVIDIA house the two-PCB assembly in their now-familiar cooler shroud, designed to direct airflow, with the end result not unsimilar to the old GeForce 9800 GX2.
NVIDIA claim it'll be faster and cheaper than the Radeon 4870 X2, with a price of $499 and availability as soon as the official announcement at CES happens.
Paired with recent drivers that enable multi-monitor with SLI, quad-SLI if you happen to buy two, and the GTX 295 is a reasonably attractive product for the well-heeled gamer.
It also makes a nice single-board package for doing dual-chip CUDA development, especially with the aforementioned multiple monitor support. I think I'll pick one up just for that.
Inter-chip communication is in a very similar vein to existing GX2 products.
The Tech Report have more specs and pictures, including clocks, and our forums rage on both sides about its expected performance, especially since the previews were a little canned in terms of what games were used.
ATI don't appear to have that much to worry about, and in general it's just nice to see competition at $499 with these type of products. Both have some instant appeal.
Now we wait for the GPU to show up in other products in the stack, to let consumers take advantage of the new power and thermal properties compared to the 65nm version powering other GeForce GTXs.
The product implements single-slot SLI using two 55nm GT200s, NVIDIA using the shrink from 65 to get the physical characteristics needed to build a sensible product. Each GPU will have all ten clusters enabled, but one ROP partition and memory channel will be disabled.
That gives rise to headline numbers of 240 SPs, a 448-bit memory bus, 7 quad ROPs and 896MiB of connected GDDR3 DRAM memories, all per chip.
NVIDIA house the two-PCB assembly in their now-familiar cooler shroud, designed to direct airflow, with the end result not unsimilar to the old GeForce 9800 GX2.
NVIDIA claim it'll be faster and cheaper than the Radeon 4870 X2, with a price of $499 and availability as soon as the official announcement at CES happens.
Paired with recent drivers that enable multi-monitor with SLI, quad-SLI if you happen to buy two, and the GTX 295 is a reasonably attractive product for the well-heeled gamer.
It also makes a nice single-board package for doing dual-chip CUDA development, especially with the aforementioned multiple monitor support. I think I'll pick one up just for that.
Inter-chip communication is in a very similar vein to existing GX2 products.
The Tech Report have more specs and pictures, including clocks, and our forums rage on both sides about its expected performance, especially since the previews were a little canned in terms of what games were used.
ATI don't appear to have that much to worry about, and in general it's just nice to see competition at $499 with these type of products. Both have some instant appeal.
Now we wait for the GPU to show up in other products in the stack, to let consumers take advantage of the new power and thermal properties compared to the 65nm version powering other GeForce GTXs.
Tagging
nvidia ± gtx, 295, sli, 55nm, gt200, 499, radeon, x2
Latest Thread Comments (135 total)
Add your comment in the forums
Posted by ChrisRay on Tuesday, 13-Jan-09 23:15:15 UTC 
Posted by Randell on Wednesday, 14-Jan-09 09:09:46 UTC
Posted by fellix Bump... on Wednesday, 25-Mar-09 10:58:22 UTC
Posted by Bludd on Wednesday, 25-Mar-09 12:57:40 UTC
Posted by fellix on Tuesday, 12-May-09 12:48:20 UTC
Posted by INKster on Thursday, 28-May-09 15:47:43 UTC
Posted by pjbliverpool on Thursday, 28-May-09 16:02:53 UTC
Posted by INKster on Thursday, 28-May-09 17:48:53 UTC
Posted by Psycho on Friday, 29-May-09 00:21:43 UTC
Posted by Sxotty on Friday, 29-May-09 20:07:06 UTC
Add your comment in the forums
Related nvidia News
NVIDIA GeForce GTX 275 at $250 to fight HD 4890
A look at NVIDIA's SLI Multi-OS and new Quadros
Ahead Nero gets CUDA support for video encoding
G92b renamed again, this time for notebooks
NVIDIA GeForce GTS 250 announced
New NVIDIA display driver for Windows 7 beta
NVIDIA Q4: Revenue as awful as expected, margins/income miss
55nm GeForce GTX 285 shows up too
NVIDIA launch 9-series GeForce IGP and get Macbook deals
NVIDIA quietly release 9 cluster GTX 260
A look at NVIDIA's SLI Multi-OS and new Quadros
Ahead Nero gets CUDA support for video encoding
G92b renamed again, this time for notebooks
NVIDIA GeForce GTS 250 announced
New NVIDIA display driver for Windows 7 beta
NVIDIA Q4: Revenue as awful as expected, margins/income miss
55nm GeForce GTX 285 shows up too
NVIDIA launch 9-series GeForce IGP and get Macbook deals
NVIDIA quietly release 9 cluster GTX 260
Well that certainly is quite nice scaling, IIRC at least over at AMD someone said that if they can get 3 times the performance with 4 chips, it's pretty much as perfect as possible, and while that's not yet 3x the performance, it's still well over 2x
now all I need is someone with access to 2 HD4870X2's etc do same test, which I doubt I can find sadly
this sorta behavior isnt exclusive to SLI though. It happens on Single GPUs. The problem seems to be the rendering bottlenecks themselves.