Valve Half Life 2 Presentation and Benchmarks

please...

Oviously all of you have radeon cards, you poor people believe every thing you read. nobody can even get the REL. 50 for nvidia yet so please spare everyone your foolishness.

 

Re: please...

Quite the opposite - most people here do not believe what they read. And the Det 50's were distributed to websites by Nvidia a few days back, and subsequently leaked to the net. Nvidia asked websites to test with them, but then did the usual thing of denying all responsibility for the results - except where they made Nvidia faster in benchmarks like Aquamark 3 that was released at the same time.

 

nVidia was discussing fp16/32 officially since the nV30's announcement at Comdex in '02. Assuming nVidia had knowledge of nV30's design prior to announcing it, it also stands to reason they had prior knowledge of the upcoming DX9 specs in advance for a long period as well. They didn't need fp16 for compliance--they needed fp24 or fp32--which nV3x has always had. So obviously nVidia planned fp32 in the chip to get DX compliance and WHQL certification, but actually planned to use fp16/fx12 for 3d-game performance--thinking this would provide them with a performance advantage over competitors using fp24 (since all the IHVs knew about it being the standard for dx9 well in advance of last December.) Had fp16 always been a part of dx9 spec, nVidia could have left off fp32 completely--might as well have done so, considering its performance. That's my opinion, anyway. You seem to think that the IHVs were completely ignorant of the specs for DX9 prior to its runtime release in December...which is just not true, as the IHVs had direct input into the formulation of dx9. M$ doesn't just release the runtime as a surprise and say, "OK, guys, here it is! Hope you are in compliance!" Heh...

nVidia just made several bad judgement calls when designing nv3x--the most serious one, of course, was greatly underestimating its competition.

FX12 was exclusively nVidia's idea--as was fp16, IMO. Which is why FX12 isn't in the dx9 API...And of course there's the little matter of ATi, which did not go the fx12/fp16 route--which lends credence to the idea that this wasn't in the API specs at the time they were designing R3x0.

 

Do anybody know when exact ATI R420 will be introduce: October 2003, November 2003, Decmber 2003 or January 2004. Any clue???

 

If history is any guide, it should be a late february / early march launch. (Meaning april / may availability.)

 

Why so late; how come not this year for Christmas???

 

More time to sell R360's at top price. They can still get six months sales with the tuned up R360. Nvidia is no competition and won't have NV40 ready until about next Spring.

It also allows R420 to come out when the PCI-Express boards start to arrive.

 

ATI R420 GPU is expected to be released in Q4 [2003]

http://www.mikeshardware.co.uk

 

I can put up a web site too with "R420 expected release dates." Doesn't make it any more likely.

 

That date was from six months ago - it's been pushed back to next year along with PCI-Express. Gives them time to put R360 out.

 

Tenuous question:

Does cpu lag manifest itself differently as opposed to a graphics card being pushed beyond its limits; i.e. instead of 'spiking' it mantains a more consistent frame rate?

Thanks.

 

If Nvidia releases NV40 by Christmas then ATI has no choice: I heard that first ATI R420 would be with a bridge chip PCI-X to AGP converter.

 

If nVidia releases the NV40 by X-mas it ain't gonna be the same NV40 card they've been talking about.

 

The first ATI R420 would be with bridge chip PCI-Express to AGP converter: (NOT the Pure PCI-Express16X R420 Radeon)

[I think ATI said that.]

 

Then it is possible to see first version of ATI R420 by the end of this year.

 

NV40 has supposedly been pushed back and won't hit Christmas. Besides, Nvidia will want some space to sell NV38.

R420 will have a AGP-PCX bridge, but my understanding is that both models will be launched at about the same time, though I wouldn't be surprised if this changed. IIRC, they've already demoed their PCX-native interface with RV380, but not the R420.

 

IMHO, if Nvidia did have the NV40 ready for Christmas release, they wouldn't "want some space to sell NV38". IMHO, they want to put the whole NV3x line to rest ASAP. I mean, from what we "know" (supposed 500MHz core speed), the NV38 core is roughly 11% faster than the NV35 core. Barring unforeseen improvements to the architecture (and not just a new respin/cherry-picking), this means that in shader-limited situations it will still be totally crushed by the 9800 (not even mentioning a similar speed-binned XT without any architectural improvements). It will probably be faster in FSAA/AF situations (due to higher memory bandwidth), but FSAA quality probably still won't be there.

That, or the NV40 is relying a lot on the NV3x core (mixed precision, the quad-thing...), and they are not really looking forward to release it.

 

That's what I'm saying - they wouldn't be pushing NV38 if NV40 was ready for Christmas. Even if they intend to trickle it down to mid-range, the NV40 derivates would be better for that - unless NV40 suffers from the same "derived from the old design" problems too.

It's always possible that Nvidia really have no shame and intent to eventually market NV38 as a midrange card, even though it is inferior. Gotta get that return on investment, even if it means screwing the customer.

 

Nvidia as an company does not release revolutionary products and they do not develop revolutionary technology too. They only gradually alter their tech from generation to generation (NV1x - NV2x - NV3x ). They don’t start from scratch but tweak their known cores and add additional functionality.

You can clearly see that with their “revolutionary” NV30. It seems to be an NV25 with FP-units bolted on; well at least from the outside.

If they stick to this culture then the NV40 will only have an tweaked NV35-type of Pixel-Shader and so the performance of the NV40 will be sub par again.

Hopefully the R300 last year was an wakeup call for Nvidia and they altered the NV40 after the release of the R300 to be more competitive. According to an leaked PDF the development of the NV40 started around April 2002; so IMHO they could have altered the NV40 and the new release of the NV40 around April/Mai 2004 instead of autumn 2003 reflects this changes / new time scale.

 

MSI GeForce FX 5900 Ultra for sale !!!!!

Takers anyone ?