G71 Specifications

[Slitherythesnake]

Text

 

[Keysplayr]

Originally posted by: Slitherythesnake
Text

Dayum!!!!! again!

First, a X1800 clocked at 760 supposedly coming soon, then R580 supposedly in Jan/Feb.
Now a 32 pipe G71. Well, all I can say is, we.... shall.... see....

 

[Munky]

32 pipes? I'll believe it when it's official. The Inq also said the r520 will have 32 pipes... and then shortly before the launch revised it to 16

 

[SonicIce]

yea right

 

[RobertR1]

It really does make sense for both companies to push out their last DX9.0c asap. If they wait mid to late next year, enthusiast will hold off for DX10 cards. As for the specs both companies will push their tech to the limit since these flagship models will have to last for about 7-8months. Look for stupidly high MSRP and street prices!

 

[destrekor]

i could care less if its still a 24pipe, it it is indeed a 700mhz+ core and better than the gtx512 (and thus better than my new 7800gt),then i will be happy.
I especially hope they are right about the early february release
my evga step-up period ends at the end of february, so if its available i will be happy

 

[JackBurton]

Video card wars are fun.

 

[Genx87]

My theory about the original G70 being a 32pipe card with 1 quad disabled may have some legitimacy. Lets wait and see how close the transistor count is.

 

[Corporate Thug]

YES, coming from the inq., the info is as good as gold. just how they nailed the r520 pipes!

 

[fliguy84]

woah, if we have G71 with 32 pipes and R580 with 24 pipes that will be one hell of a performance war

 

[ElFenix]

it better have 1GB of ram.

 

[crazydingo]

Originally posted by: ElFenix
it better have 1GB of ram.

Probably 256, 512MB first and then milk the consumers again with 1GB.

Originally posted by: Inquirer
The memory is not the problem and even Samsung memory running at 1700 MHz+ is not in tight supply.

A 180 degree flip! But but ... the magical fairy nsists that Samsung is the culprit. :laugh:

 

[fliguy84]

:Q the new forceware better have the capability to use that amount of memory as virtual memory

 

[Gamer X]

I'll wait for G80 and ATI's R600,most probably they will be dual core architectures.

 

[Munky]

Originally posted by: fliguy84
woah, if we have G71 with 32 pipes and R580 with 24 pipes that will be one hell of a performance war

My theory is that the r580 is 16 pipes... with 48 pixel shaders.

 

[Cookie Monster]

Originally posted by: munky

Originally posted by: fliguy84
woah, if we have G71 with 32 pipes and R580 with 24 pipes that will be one hell of a performance war

My theory is that the r580 is 16 pipes... with 48 pixel shaders.

QFT. I highly believe the R580 is not a 24pipe card but rather a 16pipe card with 48 fragment shader units (shader processors whateva). I think its going to be clocked lower than the current X1800XT. Its defitnetly a 16-1-3-1 architecture, due to the fact its based on the R530 (X1600XT).

Are these cards still going to use GDDR3? or GDDR4? I heard the lowest GDDR4 starts at 2400mhz effective..

 

[Cooler]

Now if only the CPU industry was moving this fast like it did in late 90's This card will be cpu limited at just about any oc. It also looks like Rollo will be getting a new SLI soon.

 

[imported_Shaq]

It will have a very fast core, we still don?t know the number but it should be around 700MHz and it's obviously paired with super fast memory. The memory is not the problem and even Samsung memory running at 1700 MHz+ is not in tight supply. The chip will have 32 pipelines and it's obviously faster than just arrived and not so available Geforce 7800 GTX 512 cards. It's expected in early February and Nvidia is already playing with these cards.

It will probably be a dual slot card at those speeds. We may have seen the last high end single slot video card as Nvidia could get away with it since X1800XT hadn't arrived yet.

OT: I wonder if Nvidia is losing money on these 512MB GTX's and they are releasing enough to win the GPU wars, but not enough that they would lose too much money?? 512 MB of the fastest memory around, PCB redesign, GPU enhancements, nice dual slot HSF and the MSRP is only $50 higher than 256 MB GTX debut?

 

[fierydemise]

After what they inq did with R520 I'll wait for an official announcement before I believe it

 

[Jeff7181]

Originally posted by: Gamer X
I'll wait for G80 and ATI's R600,most probably they will be dual core architectures.

Aren't GPU's basically multi-core architectures already? I mean... all these pixel pipelines and vertex pipelines... it's already highly parallel. What would adding another core do that adding more pipelines won't do?

 

[Matthias99]

Originally posted by: Jeff7181

Originally posted by: Gamer X
I'll wait for G80 and ATI's R600,most probably they will be dual core architectures.

Aren't GPU's basically multi-core architectures already? I mean... all these pixel pipelines and vertex pipelines... it's already highly parallel. What would adding another core do that adding more pipelines won't do?

I think the idea is that, rather than making one GPU core with 300+ million transistors, you can make two or more with half that number (or less) and (theoretically) vastly improve yields. The transistor counts on today's high-end GPUs are *insane*; it's something like 8 times the size of a P4 core's logic (most of the transistors in the high-end P4s are the L1/L2/L3 cache).

 

[Jeff7181]

Originally posted by: Matthias99

Originally posted by: Jeff7181

Originally posted by: Gamer X
I'll wait for G80 and ATI's R600,most probably they will be dual core architectures.

Aren't GPU's basically multi-core architectures already? I mean... all these pixel pipelines and vertex pipelines... it's already highly parallel. What would adding another core do that adding more pipelines won't do?

I think the idea is that, rather than making one GPU core with 300+ million transistors, you can make two or more with half that number (or less) and (theoretically) vastly improve yields. The transistor counts on today's high-end GPUs are *insane*; it's something like 8 times the size of a P4 core's logic (most of the transistors in the high-end P4s are the L1/L2/L3 cache).

Ahh... so it would be more for a manufacturing cost standpoint. Cheaper to make a bunch of 32 pipeline GPU's, and put one on one board for the "value" line, and two on another board for the "high end" line. Makes sense.

 

[Slicedbread]

You know what they should do is just make a seperate circuitboard for graphics card and have it equal the size of an matx board then just pile 4 gigs of ram and a 2Ghz processor. 200,000 in 3dmark05!!!

 

[GTaudiophile]

Originally posted by: Matthias99

Originally posted by: Jeff7181

Originally posted by: Gamer X
I'll wait for G80 and ATI's R600,most probably they will be dual core architectures.

Aren't GPU's basically multi-core architectures already? I mean... all these pixel pipelines and vertex pipelines... it's already highly parallel. What would adding another core do that adding more pipelines won't do?

I think the idea is that, rather than making one GPU core with 300+ million transistors, you can make two or more with half that number (or less) and (theoretically) vastly improve yields. The transistor counts on today's high-end GPUs are *insane*; it's something like 8 times the size of a P4 core's logic (most of the transistors in the high-end P4s are the L1/L2/L3 cache).

Right. Think of it as on-chip SLI.

 

[compgeek89]

With my 7800 GTX512, I'll wait for the 8800 before I go up again.