NV35 - already working, over twice as fast as NV30?

Vorläufige Spezifikationen des NV35:
130 Millionen Transistoren (NV30: 125 Mio.)
Herstellungsprozess: 0,13 Mikrometer (wie NV30)
Speicher-Schnittstelle: 256 Bit (NV30: 128 Bit)
Chiptakt: geplant sind 500 MHz (wie NV30)
Speichertakt: 500 MHz DDR-I (effektiv 1 GHz, NV30: 500 MHz DDR-II)

Hmmm, nice to see more confirmation that it is using a 256 bit wide memory bus. Notice that they are also saying DDR-I instead of DDR-II, which is a bit surprising.
 
In a QÃ￾ bench mark with maximum attitudes -: demo001, dissolution 1,600 x 1,200, 4xAA, 8xAF - the NV35 in the CeBIT Hinterzimmer struck not completely unexpectedly the present flagship NV30. that the projection/lead however with 110 to 48 pictures per second more than around the double precipitates, surprised so some fence guest.

So in Quake III @ 1600x1200 with 4xFSAA and 8xAF it lead the NV30 by up to 110fps??

Or however: The NV30 had not activated all functions yet at all because of its heat problems.

So not everything in the NV30 is actually functional at the moment because of heat problems. I wonder what they might be?
 
Well, it'd be the usual NV PR hype if the article writter was quoting what NV told him, but It almost sounded like he saw the benchmark running? Also was it me or did it sound like he saw twice the performance (it almost sounded like Quake was used ) with the NV35 running at only 250 mhz? Here's a crappy translation of the article.


Munich (daf/db) - as from informed circles became, already points Nvidia admits the successor of the GeForce FX 5800 Ultra of the CeBIT to selected customer in action. The NV35-Chip runs already in the Hinterzimmer - however still in the prototype stage with reduced chip clock rate.

Provisional specifications of the NV35:

130 million transistors (NV30: 125 millions.)

Production process: 0.13 micrometers (like NV30)

Memory interface: 256 bits (NV30: 128 bits)

Chip clock: 500 MHz are planned (like NV30)

Storing act: 500 MHz GDR i (effectively 1 GHz, NV30: 500 MHz GDR ii)

First NV35-Benchmarks
To likewise see a first sample of the abilities of the GeForce FX successor gab's in the Hinterzimmer. There the NV35-Prototyp still with only half strength (250 MHz chip clock) runs at present, Nvidia a NV30-System with likewise 250 MHz to clocked diagram chip against it placed. On both systems Windows XP pro SP1 and the Detonator version 42,74 (a classical "Performance" driver) ran.

In a QÃ￾ bench mark with maximum attitudes -: demo001, dissolution 1,600 x 1,200, 4xAA, 8xAF - the NV35 in the CeBIT Hinterzimmer struck not completely unexpectedly the present flagship NV30. that the projection/lead however with 110 to 48 pictures per second more than around the double precipitates, surprised so some fence guest.

The NV35-Karte did not need extra cooling despite full storing act of 500 MHz for video RAM; the NV35 is to get along according to Nvidias specifications also in its later standard clocking without aufwaendiges refrigerator.

CHIP on-line one means:
The largest progress is naturally the duplication the memory bus width of 128 to 256 bits thereby alone a bench mark value in otherwise same clocking however hardly more than doubling itself there can, must into again which were added the 5 million transistors still further optimizations be. Or however: The NV30 had not activated all functions yet at all because of its heat problems.

The fact that "secret" information penetrates like first performance data of the NV35 outward could be intended by Nvidia also - finally it does not harm at all, if on the world largest computer fair a few rumors circulates, the Nvidia again in the poles position in the 3D-Rennen sees...
 
<cough>

There the NV35-Prototyp still with only half strength (250 MHz chip clock) runs at present, Nvidia a NV30-System with likewise 250 MHz to clocked diagram chip against it placed.
 
So they're basically saying that they compared it to a downclocked NV30 during the benchmark?
 
ah so they were both clocked at 250... but they are still targetting 500mhz for the Nv35. I guess that was NV showing that clock for clock it's faster.
 
Not too surprising that at the same clock speed a chip with double the bandwidth would be faster on a bandwidth-limited benchmark. Perhaps the NV35 is also a true 8 pipe architecture.
 
:eek:

Am I the only one that finds that to be a tad too much, even with twice the memory bandwith and a true 8 pipeline design?

110FPS vs. 47FPS, thats a large difference. But is this the usual nVIDIA PR again? :oops:
 
Radea said:
:eek:

Am I the only one that finds that to be a tad too much, even with twice the memory bandwith and a true 8 pipeline design?

110FPS vs. 47FPS, thats a large difference. But is this the usual nVIDIA PR again? :oops:

Yea at first glance the numbers look too good for a part only running at 250 Mhz with DDR1 at same speed for 16x12 x4AA x8 AF? What are the fill rates and mb requirements? I know on my 9700pro on an AMD x2100 I get about 95 fps in those same conditions.
 
Humm...What's not clear from that translation is what speed the memory was clocked at. It says that the core clock was 250MHz

It also states this -

The NV35-Karte did not need extra cooling despite full storing act of 500 MHz for video RAM

So were both cards clocked at 250/500 a the demo?
 
McElvis said:
Humm...What's not clear from that translation is what speed the memory was clocked at. It says that the core clock was 250MHz

It also states this -

The NV35-Karte did not need extra cooling despite full storing act of 500 MHz for video RAM

So were both cards clocked at 250/500 a the demo?

You just beat me to it. I was looking at the same thing.
 
<cough in>

Ask yourself this: If lack of bandwidth was the main reason what NV30 Ultra didn't take the market by storm, then why did they go to such lengths to run a high core speed. (Hint: It wasn't the main reason).

<cough out>
 
Back
Top