View Full Version : Speculation - NVIDIA to Phase Out 7800GTX 512MB in Feb '06?
Scunner
12-11-05, 07:58 PM
Get 'em while they're hot:
http://www.cooltechzone.com/index.php?option=content&task=view&id=2059
Recent conversations with insiders at NVIDIA have led us to believe that their top of the line card, the 7800GTX 512MB will be out of production sometime in Q1 next year, most likely somewhere around February ‘06.
Edit: this is pure speculation, as I haven't found any sources to back up CTZ's claims. I've heard on more than one occasion that CTZ isn't the most credible source of news, so take it for what it's worth. Should this be moved into the Hardware Rumour forum?
technoid
12-11-05, 08:34 PM
I would be very interested in hearing from the people that bought a GTX 512, do you wish you had waited and are you concerned? I almost ordered an EVGA GTX 512 friday night when TigerDirect had some but ended up talking myself out of it because of this news. This could end up being a very short lived card.
Considering how they are treating us and trying to get us to change our minds with other offers on the EVGA trade up program I believe it.
We are being told that odds are we will not see more cards available for us.
PSYCHODAD
12-12-05, 05:50 AM
I would be very interested in hearing from the people that bought a GTX 512, do you wish you had waited and are you concerned? I almost ordered an EVGA GTX 512 friday night when TigerDirect had some but ended up talking myself out of it because of this news. This could end up being a very short lived card.
Best card I've ever owned.
PeterJensen
12-12-05, 07:27 AM
Well my 7800GTX/512 is really nice, and i dont regret spending money on it.
They probably need that faster memory that is on the 512 and can't afford to pair it up with a (then) old GPU.
G70U's successor, G71, is rumoured to use faster (and different) Samsung GDDR3 chips than the GTX 512. The modules used on the 512 very likely suffer from design and production problems, which could explain the shortage of those chips.
This story seems totally believable. nVidia recently lowered the BOS on the GTX 256 - from $300 to $230, which would help them move any additional stockpiled units. nVidia would slow G70 production (ditto for G70U as well as older NVX series) to make room on partner production lines for G71 and G73 based parts.
I don't think there was much pre-planning involved with the GTX 512. Engineers may have refined G70 as a just-in-case scenario, and didn't take adequate measures to ensure that the VRAM they wanted to use would be available or simply didn't anticipate Samsung to have any problems. In any case, having a non-existent card on 'shelves' isn't really good business for nVidia, and killing off the GTX 512 early enough that G71 can begin full-swing production right now would be a Very Good Thing To Do.
It would be nice if we had some new-ish roadmaps so this sort of thing wouldn't all be speculatory.
FraGTastiK
12-12-05, 06:06 PM
512MB cards are very cool but I want my 1GiG card! Last time I checked UE3 full fledged needs that much according to Tim S. unless they had optimized it for better and lower mem consumption at its maximium full eye candy.
[EOCF] Tim
12-12-05, 07:39 PM
512MB cards are very cool but I want my 1GiG card! Last time I checked UE3 full fledged needs that much according to Tim S. unless they had optimized it for better and lower mem consumption at its maximium full eye candy.
last time I checked they ran a UT2007 demo flawlessly on a 7800GT,....512mb yes...1gig...no way :)
Yea, I haven't heard that 1GB would be necessary for UT2007... but you never know. Epic games are usually loaded with very high texture resolutions.
Yea, I haven't heard that 1GB would be necessary for UT2007... but you never know. Epic games are usually loaded with very high texture resolutions.we can only hope that nvidia finds a way to span the memory on two cards to look like 1GB of memory. i think this might happen one day with all the R&D that nvidia does with SLI...
Whenever NV or ATi introduce a new GPU/VPU etc the old ones always get phased out.....not totally thrown out of production ;)
superklye
12-12-05, 11:43 PM
we can only hope that nvidia finds a way to span the memory on two cards to look like 1GB of memory. i think this might happen one day with all the R&D that nvidia does with SLI...
That's what I've wanted SLI to do from the start. I understand that you can't combine the two GPUs to act as a single one (i.e. two 500MHz cores = 1GHz), but I don't understand why two 256MB cards don't have a frame buffer of 512MB that both cores can access at the same time.
It'd be sweet if nVIDIA could come up with something like the HTT but for the GPUs and on-card memory.
It'd be sweet if nVIDIA could come up with something like the HTT but for the GPUs and on-card memory.that really would be sweet. im willing to bet that nvidia is working on something like this right now for sli2/G80/dx10...ATi would have a hell of alot of catching up todo if they did..
You are assming ATi are not researching/designing that right now. I am sure NV haven't underestimated their rival...rivals
How would HTT work on GPU when everything is so parallel? ..just a thought... it's almost working like HTT now only each quad block does 4 instances of erm calculation in order?
I am sure NV haven't underestimated their rival...rivals
well im sure ati is trying todo the same but the diffrence here is that nvidia has has two years of retail exposure and exprience. ati has yet to release any crossfire boards yet to the public.
superklye
12-13-05, 12:50 AM
How would HTT work on GPU when everything is so parallel? ..just a thought... it's almost working like HTT now only each quad block does 4 instances of erm calculation in order?
Well, the current way things are done, it wouldn't work, I suppose.
A complete overhaul of how the GPU works would be in order, just as Creative did with the X-Fi, basing the SPU on their new ring architecture which allows whatever part of the card that needs processing cycles to access it dynamically, instead of having to run through the same linear pipeline for EVERYTHING.
I dunno, I'm just kinda speculating. I'm sure nVIDIA has thought of all these things and is working on them or something similar, as is ATI.
I dunno if u will ever need HTT in GPUs, you can change teh amount of PS calculations that can be done as Tertsi explained it but...HTT in GPus...? i duno
FraGTastiK
12-13-05, 02:22 AM
Tim']last time I checked they ran a UT2007 demo flawlessly on a 7800GT,....512mb yes...1gig...no way :)
yea I have the video of it running on GTX SLI with high details at hi-res w/AA .but for full blown textures you gonna need 1GiG cards unless Sweeny had come up with something like JC's megatextures by now.again it would need that much if you are gonna crank that last slider up to its maximum limit(wich is what I want to do) otherwise you'll be fine with 512 I guess.
BU: UT2004 required 5.5 gigs of hard drive space to install. This has got to be a strain a lot of people's systems. So you're talking 2048 x 2048 texture sets, what kind of system and memory is this next game going to take?
TS: Well, we are aiming at the kind of PC that we think will be mainstream in 2006. We will also be able to scale it down. Basically DirectX 9 cards will be minimum spec, so any DirectX 9 shipping today will be capable of running our game, but probably at reduced detail. If you only have a 256 meg video card you will be running the game one step down, whereas if you have a video card with a gig of memory then you'll be able to see the game at full detail.
Sweeney was actually anticipating 2GB cards by the time UT07 is released. But the nice thing about texture resolution is that it's easily scalable. You can ship with 4096x4096 textures and scale down to suit. No sweat.
You'll also have to remember the limitations of console hardware, as it's obvious a considerable portion of time has and will be spent on 360/PS3 optimization. In both cases, we're talking about a total of 512MB of RAM. You can bet that UT07 on PC and console platforms won't look wildly different.
EDITED BY MUYA : PLEASE DO NOT LEACH OTHER WEBSITES BANDWIDTH BY DIRECTLY LINKING THE FILES. OBTAIN PERMISSION FOR COPYRIGHTED MATERIALS BEFORE POSTING OR EVEN LINKING
FraGTastiK
12-13-05, 02:45 AM
Sweeney was actually anticipating 2GB cards by the time UT07 is released
I know thats why I'm gonna get my next card after the UT2k7 release.yea it can be scaled down but as I said for full blown textures...
and will be spent on 360/PS3 optimization. In both cases, we're talking about a total of 512MB of RAM. You can bet that UT07 on PC and console platforms won't look wildly different.
yes but the PC version is slated to have higher details etc. available for the user to TURN ON.so it wont be a console port.jAkUp's interview is the place to look for what he said about this.
Oh, and nVidia PR released some early benchmarks running UE3 when the GTX was released. Dig through the HardOCP 7800 GTX launch article and you'll find it
I think I already knew that:
yea I have the video of it running on GTX SLI with high details at hi-res w/AA .but for full blown textures...
Hmm, 5.5 gigs wouldn't be all that much bigger than 2k4. I would be a little surprised if it comes in that low.
we can only hope that nvidia finds a way to span the memory on two cards to look like 1GB of memory. i think this might happen one day with all the R&D that nvidia does with SLI...
The needed bus would be quite the opposite of the current 256bit highly parallel crossbar. And the increased latency... But why not, a pair of Opterons can do it. The performance wouldn't prolly ever be up to the dedicated model's. But a sweet middleground?
Wasn't JC's megatexture intended more for an outdoors renderer (quakewars)?. With indoors zooming "infinitively" into a texture wouldn't gain that much?
Anyhow, Nvidia sort of spoiled their "available from day 1" campaign with the GTX 512...
Does anyone else get the impression that the 7800GTX 512MB was just a Phantom edition card to steal X1800XTs thunder?(sofahide)
vBulletin® v3.7.1, Copyright ©2000-2012, Jelsoft Enterprises Ltd.