NVIDIA Announces 112 SP 8800 GTS Works w/ Origional 8800 GTS In SLI

[NamesDontMatter]

NVIDIA told their AICs not to market the new Geforce 8800 GTS 640MB as a card that has 112 SPs on board (upgraded from previous 96 SPs) but Partners can label it as a special edition, or OC edition. Also the new 112 SP 8800 GTS will still work with the original 96 SP variants in SLI configurations. Further NVIDIA stressed that they will not offer an upgraded 112 SP 8800 GTS in 320MB configurations.

Source: VR-Zone

[wickerman]

Seems rather poorly thought out to re-release a card under the same name but with different specs, especially as significant as G80 vs G92. Wouldn't it have been smarter to call it an 8850GTS? The upgrade from 96sp to 112sp clock for clock is pretty significant, and with the 65nm core expected to scale much better than the 90nm G80, speeds in excess of 600mhz shouldn't be much trouble. Hell calling it an 8850GTS would probably increase sales in an of itself because sometimes people don't look too closely at the specs, the model numbers can often dictate whats gunna end up in someones system.

If the new 8800GTS 640 remains on a 320bit memory bus, it will be faster clock for clock and what would be the point of pairing with a slower card for SLI, its nice that you can at least...
Either way, its gunna be a nice card if the current G80 performance numbers are anything to judge by. I doubt I'll trade in my new hd2900pro until the new high end cards launch, but it will be nice to see how these new cards compete with one another.

[Tatty_One]

Originally Posted by wickerman

Seems rather poorly thought out to re-release a card under the same name but with different specs, especially as significant as G80 vs G92. Wouldn't it have been smarter to call it an 8850GTS? The upgrade from 96sp to 112sp clock for clock is pretty significant, and with the 65nm core expected to scale much better than the 90nm G80, speeds in excess of 600mhz shouldn't be much trouble. Hell calling it an 8850GTS would probably increase sales in an of itself because sometimes people don't look too closely at the specs, the model numbers can often dictate whats gunna end up in someones system.

If the new 8800GTS 640 remains on a 320bit memory bus, it will be faster clock for clock and what would be the point of pairing with a slower card for SLI, its nice that you can at least...
Either way, its gunna be a nice card if the current G80 performance numbers are anything to judge by. I doubt I'll trade in my new hd2900pro until the new high end cards launch, but it will be nice to see how these new cards compete with one another.

I was under the impression that this card is not G92 but G80, it is not the lower fabrication process rather an upgrade of the exisiting 8800GTS which allegedly already has the extra SP's but they are disabled/laser cut??? Would not stake my life on that but am fairly sure that this is nothing but a slightly faster 8800GTS G80......might be wrong tho.

[[I.R.A]_FBi]

bwoy .. tehm nuh have nobady with sense @ nvidia?

[devguy]

Originally Posted by Tatty_One

I was under the impression that this card is not G92 but G80, it is not the lower fabrication process rather an upgrade of the exisiting 8800GTS which allegedly already has the extra SP's but they are disabled/laser cut??? Would not stake my life on that but am fairly sure that this is nothing but a slightly faster 8800GTS G80......might be wrong tho.

Yeah, the g92 is labeled the 8800GT. This is the same 640mb GTS we've all seen before, but rebranded with a few more sps.

[DaMulta]

They don't want the old ones setting on the shelf. That's my idea on this.

[DanTheBanjoman]

Originally Posted by [I.R.A]_FBi

bwoy .. tehm nuh have nobady with sense @ nvidia?

You're not exactly giving a great example of sense with language.

I don't really see the issue, it's basically the same card but the latest model is a bit faster, I'd thank them for that. ie if it's the same card you might get a new one when you RMA the old one.
Besides, manufacturers are free to name the card differently. Either way, as long as they're drop-ins for the old cards I'm fine with it, they're not limiting you in any way.

[[I.R.A]_FBi]

exactly ...

[wickerman]

Originally Posted by Tatty_One

I was under the impression that this card is not G92 but G80, it is not the lower fabrication process rather an upgrade of the exisiting 8800GTS which allegedly already has the extra SP's but they are disabled/laser cut??? Would not stake my life on that but am fairly sure that this is nothing but a slightly faster 8800GTS G80......might be wrong tho.

Well it quite easily could go either way, some rumors say they are taking a 128 shader G80 and cutting it down to 112 is certainly possible but if you have the means to produce a card for a lower price, usually you do and apparently G92 is completely capable of doing just that. It does make sense that they take a core thats cheaper to produce and put it in a card that can increase profit. I guess theres no way to know for sure till the product launches, but I wouldnt be surprised to see G92 instead of G80 under the heatsink. It just doesnt make sense to do it any other way unless yields for the 65nm G92 are sub par?

[Jizzler]

Originally Posted by DanTheBanjoman

You're not exactly giving a great example of sense with language.

I don't really see the issue, it's basically the same card but the latest model is a bit faster, I'd thank them for that. ie if it's the same card you might get a new one when you RMA the old one.
Besides, manufacturers are free to name the card differently. Either way, as long as they're drop-ins for the old cards I'm fine with it, they're not limiting you in any way.

Yeah, it's not going to keep me up at night, but I do like the 8850GTS idea that was brought up earlier. If the extra sp's aren't enough to earn a half-step moniker, bump up the stock core/mem speed a tad.

I'm just remembering the days of selling computers. I could be explaining the differences of sp's, or I could just say that's an 8800GTS and that's a 8850GTS

[newtekie1]

The problem with naming/calling it an 8850GTS is that you instantly make it sound better than the 8800GTX/Ultra. The same idiot consumers that won't know the difference between an 8800GTS with 96 SPs and 112 SPs will also assume anything with a larger number is instantly better. Therfor they will usually think that an 8850GTS is better than an 8800GTX simply because 8850 > 8800. Personally I would have just used a different suffix. 8800GTP or 8800GTS+ would have been better, IMO.

Originally Posted by wickerman

Well it quite easily could go either way, some rumors say they are taking a 128 shader G80 and cutting it down to 112 is certainly possible but if you have the means to produce a card for a lower price, usually you do and apparently G92 is completely capable of doing just that. It does make sense that they take a core thats cheaper to produce and put it in a card that can increase profit. I guess theres no way to know for sure till the product launches, but I wouldnt be surprised to see G92 instead of G80 under the heatsink. It just doesnt make sense to do it any other way unless yields for the 65nm G92 are sub par?

Or the G92 doesn't have 112 SPs to begin with. I haven't seen any hard specs on G92 yet, but last I heard, the rumors were saying it only had 64 SPs.

[Jizzler]

Hmmmm... 8800GTS+ FTW!

I, and I assume the previous poster was thinking like what they did with the 7 series and the 7900GTX, 7950GT.

[newtekie1]

Yeah, I actually got a few customers that were pissed off at me for trying to sell them the more expensive 7900GTX over the 7950GT, which is just a rebadged 7900GT. Which is why I think the 8850GTS name isn't that great of an idea either. I think 8800GTS+ is really the way to go. Only because, now that I think about it some more, if they were to use 8800GTP or something like that, then people would get confused with SLI. 8800GTS+ conforms to nVidia's requirements that it still be called an 8800GTS, and it give the customer an idea that it isn't like the other 8800GTS varients.

[pbmaster]

I like the fact tha SLI is becoming a little less picky on what cards you have in there. True you will not get as big a boost as with 2 of the exact same card, but the compatibility is there, doesn't matter how you look at it.

[KennyT772]

Model numbers mean nothing, these are simply filling the gap in true gts chips. There simply isn't enough defective cores to warrent a full cut to 96sps. I believe that these are unused ultra cores, cut to 112, and sold to AIB's as "overclocked" chips. This way nVidia gets more for the chips, AIB's get more for the cards, and both get a better opinion of the public because these overclocked cards are suddenly quite faster than the normal gts's. My guess is that they will clock rather well on top of it.

[wickerman]

well its not uncommon to launch a new upper middle range card that ends up faster than the high end cards that launched before it. Nvidia did it once before when they launched the 7900GT (that replaced the 7800GT price point) which was faster than the 7800GTX.

Some sources are saying G92 will be a direct shrink of G80, and is capable of the full 128 shaders, dropping it to 96 would yield great performance still, but 64 is just rather low. It would easily outclass the 8600GTS, but lets face it..people had much higher expectations for the 8600GTS (as well s the 7600/8600GT/HD2600/x1600). With high demand games already out and a hell of a lot more to come, why not ensure your product remains the fastest? Im not saying the 8800GT will be faster than the 8800GTS, but there is no reason for it not to be. The 8800s have been out since late 2006, should it remain "high end" for much longer? The new midrange (or upper middle range if you prefer) should be as fast as the previous high end.
If they manufactured the G92 to specifically sit only on the 8800GT it is a nice move, but they could stand to make more money by refreshing the entire 8800 line up with 65nm versions as ATI may not have a tried and true 8800 killer till later in 08. Making a 65nm version of G80 for the high end would result in cheaper production, higher clock speeds, and probably lower power consumption and heat output. They wouldnt really have to do a big multi million dollar ad campaign for it anything, and it would mean Nvidia has a stronger high end for the upcoming game titles (like crysis) to compete against ATI. Going from the 110nm G70 to the 90nm G71 was a worthwhile step, why not take your current high end from 90nm to 65nm.
Usually you releases a high end on a tried and true manufacturing process and bring in the new technology on the midrange and low end (where cutting manufacturing costs is well worth it) but it still..

But then again its easy for me to sit here and speculate, its not my company and I dont have to answer to the share holders

[Tatty_One]

And of course the 1950Pro actually bested the 256MB 1950XT in some games an certain resolutions......there are always some hazy area's out there but I am all for a more simplified approach!

[yogurt_21]

lol sli with regular gts's whats the point of that? why not just sell the old gts and buy another new one and sli them. then you'd have two 112shader 8800's. which should run faster. (or just sli you old gts with another old gts and be happy)

[Tatty_One]

Originally Posted by yogurt_21

lol sli with regular gts's whats the point of that? why not just sell the old gts and buy another new one and sli them. then you'd have two 112shader 8800's. which should run faster. (or just sli you old gts with another old gts and be happy)

Yes.....but cost you more.

[WarEagleAU]

Wait, I didnt know the 8800 GTS used 96 Stream processors.....wow, this will definitely improve performance some.