The G92 Architecture Rumours & Speculation Thread

Status
Not open for further replies.
We shall have our own brand of 512MB ones which we should be able to do for around £120-£130+VAT.

Better than GTS 640MB performance for less than £150???

Wow! Let the good times role!

Combined with the insane CPU prices at the moment and cheap memory Its a hell of a time to buy a cheap high end PC.
 
why is that, pls explain.
Because you didn't want to take the effort to try to understand the technical aspects, and this is a technical forum. If you're not interested in anything else than 'zomg how many pipelines, how fast is this?' then I suspect you won't feel at home here.

As for your question - if the GeForce 8800GT is a 7 clusters part, then you can conclude that it probably (but not certainly) has 112 SPs and 28 TMUs. That does not affect the number of ROPs, on the other hand, which is still expected to be 16. Although there is more to that than just their number...

Megadrive1988 said:
shouldn't there be a seperate thread on G9x,
the new high-end refresh of G80 which is not G92 ?

be it G90, G9x ...
Assuming all the rumours are correct and there is indeed no high-end part announced on Oct 29th, then it would be a good idea to start a 'G9x High-End Speculation Thread' afterwards, yeah. For now, let's keep focusing on this though...
 
sorry Arun. i suggest you start your own forum and Nazi it all you want and only let people who are interested in technical aspects post. i saw no such requirement when i registered and I have no desire in the slightest to learn technical aspects and i definitely don't care about feeling at home here, i just want information, don't care where from.
 
sorry Arun. i suggest you start your own forum and Nazi it all you want and only let people who are interested in technical aspects post. i saw no such requirement when i registered and I have no desire in the slightest to learn technical aspects and i definitely don't care about feeling at home here, i just want information, don't care where from.
Errr, I suggest you might want to cool down a bit.

I'm not banning you, nor would I ever want to do so for such reasons alone. I am merely pointing out that that kind of behaviour tends to be frowned upon here, and goes against the entire raison-d'etre of the site.

If all you want is information, clearly this isn't the only forum in the universe that could provide you with it, and thus I suggested you might get more people willing to respond to you with that kind of behaviour elsewhere. If you don't care about that, be my guest, but don't bitch about it later and tell me we didn't warn you.

I have absolutely no problem with you asking such a question, but if you want to become a regular rather than just leave now that you've got your answer (albeit delayed, sorry about that - should have replied straight away along with my comment), then I thought it was important you realized you might not fit in perfectly here. That wasn't Nazi at all from my POV, and I'm sorry if I was so vague and direct that you felt that way.
---
And now back to the regular programme please...
 
sorry Arun. i suggest you start your own forum ...
LOL. He doesn't have to. He is one of the people who run the site. Frankly, I would have just deleted your post outright but clearly others have more patience than me.
 
As for your question - if the GeForce 8800GT is a 7 clusters part, then you can conclude that it probably (but not certainly) has 112 SPs and 28 TMUs. That does not affect the number of ROPs, on the other hand, which is still expected to be 16. Although there is more to that than just their number...
Except the fact that il should actually have 16 "free bilinear" TMUs and not 14.

GTX has 32 TMUs, GTS has 24, there's no reason fot them to put just 14 in there, that would leave "unused" bandwidth and increase routing complexity.

Now, I wait for the real product, as I suspect there's a lot of FUD around this GPU.
 
Except the fact that il should actually have 16 "free bilinear" TMUs and not 14.
My understanding of the rumours is that the 8800 GT has 28 'free trilinear' TMUs, just like the G80-based 8800 GTS has 24 and the GTX has 32. So that's 56 filtering units, vs 48 for the GTS and 64 for the GTX. As such, 14 and 16 don't have much to do with this.

Of course, if that wasn't the case, it would also explain why it's called the "8800 GT" rather than the 8850 GTS. But as you said yourself, there probably is a fair bit of FUD flying around, and NV's branding strategy lately has been far from logical.
 
sorry Arun. i suggest you start your own forum and Nazi it all you want and only let people who are interested in technical aspects post. i saw no such requirement when i registered and I have no desire in the slightest to learn technical aspects and i definitely don't care about feeling at home here, i just want information, don't care where from.

While I use this place for info on products aswell, it benifits me to read and learn from the articles and those here who have a far greater understanding of the products on the market and how they work/operate. I would suggest, as others have, that it would also benifit you to lread up on such things to give you a better idea as to the info you are asking about.
 
My understanding of the rumours is that the 8800 GT has 28 'free trilinear' TMUs, just like the G80-based 8800 GTS has 24 and the GTX has 32. So that's 56 filtering units, vs 48 for the GTS and 64 for the GTX. As such, 14 and 16 don't have much to do with this.

Of course, if that wasn't the case, it would also explain why it's called the "8800 GT" rather than the 8850 GTS. But as you said yourself, there probably is a fair bit of FUD flying around, and NV's branding strategy lately has been far from logical.

Why on earth is Foxcon saying that the 8800GT is G96?

excuse me if it has already been posted.

http://www.tomshardware.com/forum/245553-33-8800gt-specs-foxconn-site-dare-official

Also comfirms that it's not DX10.1.

This makes makes sence to me. Nvidia clearly stated g92 was going to pack a teraflop of power. Well obviously, 8800GT does not have a terraflop of computing power and with Foxcon claiming it's G96, that still leaves room for the G92 terraflop chip.
 
With the right clocks G92 could reach the TFLOP, where is the problem?

G96 is definitely not a 7C/112SP-chip, it is AFAIK G84@65/55nm.

And also, we see 8800GT without cooler -> G92-270.
 
AnarchX said:
With the right clocks G92 could reach the TFLOP, where is the problem?
2.51 GHz for 8 TCPs is a problem in my eyes...

or 2.02 GHz for 10 TCPs

or 1.68 GHz for 12 TCPs

Which seems most likely to you?


EDIT: Anyway, looking at the pic of the supposed G92 core it looks to easily have more than 800M transistors...
 
Last edited by a moderator:
With the right clocks G92 could reach the TFLOP, where is the problem?

Thats again assuming it's G92.

The 8800GT has less shader power than the GTX. The GT would need a serious clock to even come close to reaching 900gflops let alone 1 teraflop. So again, this fills a empty promise of Nvidia's terraflop statments.
 
Last edited by a moderator:
2.51 GHz for 8 TCPs is a problem in my eyes...

or 2.02 GHz for 10 TCPs

or 1.68 GHz for 12 TCPs

Which seems most likely to you?
Thats assuming G80 ALUs, G84 or anything else and you need to modify your numbers. Too many variables and Nvidia is doing a really good of confusing everyone, including their consumers. :???:
 
Thats assuming G80 ALUs, G84 or anything else and you need to modify your numbers. Too many variables and Nvidia is doing a really good of confusing everyone, including their consumers. :???:

Tell me.... How would Nvidia go about confusing ATi if they did not confuse the consumers as well? Are consumers somehow immune to NV smoke screens and ATi is not compatant enough understand?:p
 
Can't wait to get my dirty little hands on one of these GT babies. Time for some serious OC (i hope). Just rip that weak single slot cooler off first... :smile:

My boring 8800 GTS doesn't wanna play over 640 MHz. :???:
 
Tell me.... How would Nvidia go about confusing ATi if they did not confuse the consumers as well? Are consumers somehow immune to NV smoke screens and ATi is not compatant enough understand?:p
They could begin by not naming the thing 8800GT. Its got to be the retardest naming move in recent memory.
 
G92 is 8 TCPs / 128 SPs chip.
GT has 1 TCP disabled, 'new' GTS will have all 8 of them running (and from the way it all goes G92 may eventually replace G80 completely).
It doesn't support DX10.1 (and i don't see any reason to actually). But this may change in the future with simple driver update for all G8x/9x family (if those rumours were true).
As for the chip size -- how about DP added complexity? What if RV670 doesn't support DP and that's why it's relatively small compared to G92 which do support DP?
And where the hell is my new hi-end? :(
 
Status
Not open for further replies.
Back
Top