Could this mean anything? A little slip up by VR-Zone?
http://www.vr-zone.com/articles/H1_&...X2/5766-6.htmlAMD is coming up with their much-awaited Radeon HD 4870 X2 next-generation graphics card, while NVIDIA has responded with its GeForce GTX 260 and GTX 280 dual core GPUs.
Its in the trinity review, last page, final paragraph
Originally Posted by Hans de Vries
what about the pictures of the gt200?
the heatsink appears to be that of a single gpu card, and then there's a naked pcb shot leaked
Meaning to two GPU cores next to each other, like a dual core CPU. Not like the 3870X2 with two seperate GPUs.
To infinity, and beyond! (Azza was taken)
Will not happen.
Yeah seems a bit unlikely to me too since gpu's already consist of multiple shader processors?
making a dual core gpu-die doesn't make any sense to me..
Rig 1:
Intel E4300 @ 3Ghz - 2gb OCZ PC2-8500 - Asus P5N-e SLI - Club3d 9600gt @ 750/1950/1100Mhz - Vista 64
Rig 2:
Intel celeron L420 @ 2.6Ghz- 2gb OCZ PC2-6400 - Asus P5B - XFX 8800GS 384mb - XP 32
Laptop
Acer Aspire 3610, Pentium M725 OC @ 2.23Ghz - 2gb PC2-3200 - crappy Intel I915 gfx
Its amazing we have this...dualcore crap on the run. Who seriously makes these things up and still believe it. If you say enough times the moon is a big cheese people will believe it. At best 280 GTX would be a GX2 style card.
Crunching for Comrades and the Common good of the People.
which would be an utter failure... fsk dual cards, i dont need sli/cf...
Why Not?
Fast computers breed slow, lazy programmers
The price of reliability is the pursuit of the utmost simplicity. It is a price which the very rich find most hard to pay.
http://www.lighterra.com/papers/modernmicroprocessors/
Modern Ram, makes an old overclocker miss BH-5 and the fun it was
heat per square inch
Ryzen 9 3900X w/ NH-U14s on MSI X570 Unify
32 GB Patriot Viper Steel 3733 CL14 (1.51v)
RX 5700 XT w/ 2x 120mm fan mod (2 GHz)
Tons of NVMe & SATA SSDs
LG 27GL850 + Asus MG279Q
Meshify C white
"thermal dissipation density" lol sorry- same thing.
a "dual-core" GPU wouldnt work because the die size would be BIG and yields would suffer.
2 dies on one sub would work (like Intel quads,) or 2 discreet cores (like the 3870X2) or obviously theres the dual-PCB option
Then the thermal envelope becomes the limiting factor..... would only work on lower-end cores, then theres the question- are 2 lower-end dies gonna perform better than one high-end. The major case for that would be yields of small Vs big cores
Last edited by K404; 05-20-2008 at 11:16 AM.
______
Sometimes, it's not your time. Sometimes, you have to make it your time. Sometimes, it can ONLY be your time.
same argument was made about Intel's Smithfield (2 Prescott dies under one IHS)
Fast computers breed slow, lazy programmers
The price of reliability is the pursuit of the utmost simplicity. It is a price which the very rich find most hard to pay.
http://www.lighterra.com/papers/modernmicroprocessors/
Modern Ram, makes an old overclocker miss BH-5 and the fun it was
Plus 2x memory, plus bigger total diesize, plus CF/SLI scaling and other issues and so on"thermal dissipation density" lol sorry- same thing.
a "dual-core" GPU wouldnt work because the die size would be BIG and yields would suffer.
2 dies on one sub would work (like Intel quads,) or 2 discreet cores (like the 3870X2) or obviously theres the dual-PCB option
Then the thermal envelope becomes the limiting factor..... would only work on lower-end cores, then theres the question- are 2 lower-end dies gonna perform better than one high-end. The major case for that would be yields of small Vs big cores
CPUs MCM design makes sense since they have a low bandwidth interchange requirement, completely dfferent memory management and because scaling is vastly different. Specially due to scaling of the single die CPU. A 10 issue wide Core 2 wouldnt be any faster than the 4 issue wide we have. But a 10 wide GPU would be 2½x faster than a 4 wide. So why waste TDP and wafer on extra interconnects, MPU and so on. If an issue port on a CPU is bad. The entire CPU is bad. If its on a GPU. You just remove that and the rest still works. Same reason we have 112 and 128 shader G92 etc. You could call them 7 and 8 core GPUs.
Last edited by Shintai; 05-20-2008 at 11:33 AM.
Crunching for Comrades and the Common good of the People.
It is dual-core, but not like you guys seem to think...
GT-200 = one core, NVIO = one core. Put them together on the same card? That's right, dual core.
lol, that was smart
Are we there yet?
Priceless
Crunching for Comrades and the Common good of the People.
Nvidia seem behind on the race as far as dual core is concerned so doubt they will have a true dual core GPU out before ATI but stranger things have happened...
neither ati nor nv has a "dual core" gpu solution... what makes you think one is behind another when both of them have nothing on there hand?
If you refer to the 3870X2, thats no "dual core" gpu solution just crossfire on one board.
Last edited by Hornet331; 05-20-2008 at 05:13 PM.
Guys...DUAL CORE GPUS AREN'T LIKE DUAL CORE GPUS!
Every shader on a gpu IS a core!
??? He's right. In addition, VR-Zone has non-English based news writers that can easily miswrite dual gpu as dual core.lol, that was smart
Perkam
what i want to know is, if die stacking would be a possibility for gpus. Lets say rops+video processing are the base and the spus are stacked on the rops, but i dont know how much power the rops consume so it maybe would be a bad idea.
I haven't said/meant he's wrong, I agreed with the statement
Are we there yet?
Posting Permissions
Reply With Quote
Bookmarks