View Full Version : NV40 has 3 times the power of GeForce FX 5950!
cy_a253
02-10-04, 05:39 PM
http://www.theinquirer.net/?article=14058
It's even more interesting that Nvidia claims that its card will end up four times faster in Doom 3 and seven times faster in Half Life 2.
Also, details about the NV45. (which is basically the NV40 with PCI express)
http://www.theinquirer.net/?article=14063
Uhh, your quote doesn't quite match up to your topic title...
Lezmaka
02-10-04, 05:56 PM
And I bet the quote won't quite match up with what happens in real life.
Malfunction
02-10-04, 06:06 PM
Eazy there killa, there is a forum for this type of stuff called "The Hardware Rumor Mill." I'm just as excited about the NV40, though I take anything from "The Inquirer" with a grain of salt like so many here as well.
Beyond that, Welcome to the Forum!:D
Peace,
:afro:
StealthHawk
02-10-04, 07:07 PM
Double post: http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=24678
NV40 has 3 times the power of GeForce FX 5950!
Now there's a headline for the front page guys. Come on, post it! :D
Or not.. :)
Skinner
02-11-04, 05:12 AM
Well, 3 of 7 times is a little relative but I certainly believe in some shaders the nv40 will be 5-7 times faster, they have to!.
Originally posted by DMA
NV40 has 3 times the power of GeForce FX 5950!
Now there's a headline for the front page guys. Come on, post it! :D
Or not.. :)
thats absolutely ridiculous and you know that.
Hellbinder
02-11-04, 10:42 PM
Actually i would not be supprised if their FP32 performance is between 2.5 to 3x faster than the Nv38. That would put it about 50% faster than their Current FP16 speed and right in the ballpark for expectations for a next gen card.
SurfMonkey
02-12-04, 12:10 AM
But FP16 isn't inherently faster than FP32 currently. It still all boils down to dodgy register allocation. So if it's 2.5x to 3x faster in FP32 then it follows that FP16 will also be 2.5x to 3x faster. If they've increased the number of temp registers and decreased latency between state changes then it may well be a little quicker than that.
Either way it's going to be interesting to see.
Hellbinder
02-12-04, 07:30 PM
Originally posted by SurfMonkey
But FP16 isn't inherently faster than FP32 currently. It still all boils down to dodgy register allocation. So if it's 2.5x to 3x faster in FP32 then it follows that FP16 will also be 2.5x to 3x faster. If they've increased the number of temp registers and decreased latency between state changes then it may well be a little quicker than that.
Either way it's going to be interesting to see.
Not if they Focused on FP32 Support.
Who knows till we see it though.
MatiasZ[OC]
02-14-04, 05:12 PM
Originally posted by Hellbinder
Actually i would not be supprised if their FP32 performance is between 2.5 to 3x faster than the Nv38. That would put it about 50% faster than their Current FP16 speed and right in the ballpark for expectations for a next gen card.
would that imply that kind of increase in every shading application or only certain places? i mean... there's not much FP32 used right now right?
BTW nice to see posts from you not bashing nVIDIA for the sake of it :)
lets speculate about those numbers for a second..
if ATi and Nvidia claims are more or less acurrate.(not really sure about if those where Nvidia claims or only a rumor of the inquirer")
lets say that both companies focus is PS2.0 FP32 performance.
so ATi target (best case) in the R420 is 2x times the Ps2.0 speed of the R300.
NVidia Nv40 (best case) is (3x times the speed) of the 5950 in ps2.0 Fp32.
then whe have the following.. (using carmack coments about Fp32 and 3dmark)
R300 is ~ 2xtimes the performance of the Nv30 (when both are at their fullest precision... ex..doom3)
R420 is ~ 2x times the performance of the R300 and ~ 4 times the performance of the Nv30.
Nv40 ~ 3x times the performance of the NV30 (for simplicity i use Nv30 ,so maybe its a just a way more againts the Nv30 like 3.5xtimes)
so we have the R420 at full precision is faster than the Nv40.
but i find dificult to believe that the less faster card there will be anything less than good in Dx9 anymore and at full precision. since it should be a way faster than the faster card of ATi today,the ones that valve recomends for Hl2 to play it at its fullest quality. also the diferences should be much less dramatic than NV30 vs r300. because ATi new card increase is 2x and Nvidia new card is 3x. simply math :) however this should be the case of theorical speed in syntheticDx9 benchmarks at full precision . in games is more dificult because there are -other- things that can make the diferences between both cards greater or smaller.things like Memory bandwidht / VS speed and Ultrashadow or similar ATi shadows tech and compression techniques,singletexturing/multitexturing , AF/AA speed that can turn the balance to one side or the other. So its looks like Nv40 advantages will be its Huge bandwidht ~52Gb/s ?and maybe the ultrashadow ,the R420 advantage can be its very powefull Ps/VS units.. Nv40 looks like will have the advantage in the flexibility in their pipes configurations 6x2/8x2/16x1? .. both cards most likely will have similar core clock speed.
if i were to predict the performance about where each video card stand ,using all the Rumors posted here /speculations or facts. then i think that the diferences in speed between the Nv40 and R420 will be similar in general to the fight of an R300 vs the
NV35. the later winning all the Dx7/Dx8 and equivalent OpenGl games benchmarks and the other winning all the Dx9 Ps2.0/vs2.0 benchamrks and games when both forced at least to Full precision.
i think we are going to see this time a much more interesting battle.. because of the huge improvements that each COmpany seems that have made in their next generation hardware. so i expect very nice surprises from both companies. :) so you cant be wrong choosing anyone .if both cards performance is good ,in any game ,(which is what most likely will happen)then the final choice will not be about speed but about FEatures ,IQ and included games bundle . just imagine the R420 shipping with Hl2 and the Nv40 with Doom3.. :) how about Stalker or Farcry? very hard to resist any of those if the diferences in performance are not that much. however if Hl2 is delayed until late this year and doom3 ships in april with the Nv40. then ATi is out of luck. i think at the end it will not be the faster card the one that will get the most sales ,but the one that ships first with one of those games.. :)
btw.. i find suspicious the Nvidia praise to Doom3 in the conference call.. either one..
they are very happy with their Doom3 performance and /or they are suggesting they have some plans to do something with that game or engine?..
"Oh my goodness!" when commenting on Doom3's impact on graphics - was compared to Quake, which created the GPU market - Doom3 will be more than a game (the game engine will become a franchise)
jimbob0i0
02-15-04, 08:14 AM
NV40 I agree with you to a lareg extent there.... the only thing that will twist it for me is *if* NV actually meet all the DX9 *current* techniues... ie. if they finally support floating point render targets and HDR properly. If they don't then in *proper* DX9 games ATI will still have the IQ/feature edge and NV will be left behind again - remember DX9 isn't just about shader type and speed as it was with DX8/8.1!
this is my take on the whole thing. if both cards are very close in performace then nvidia will win.... all cheating aside... ati has had a very bad history with drivers for new cards (but they have gotten much better) and nvidia driver team is second to none .... i think just the problems alone for ati right out of the gate with a new card will persuade many people that stand on the fence to go nvidia oh well what ever
rets:afro2:
Originally posted by jimbob0i0
NV40 I agree with you to a lareg extent there.... the only thing that will twist it for me is *if* NV actually meet all the DX9 *current* techniues... ie. if they finally support floating point render targets and HDR properly. If they don't then in *proper* DX9 games ATI will still have the IQ/feature edge and NV will be left behind again - remember DX9 isn't just about shader type and speed as it was with DX8/8.1!
well yes.. i suppose that Nvidia had enough time to Work and Fix anything concerning old Dx9 in the Nv40.even time to support more features.. like the Ps3.0. (when claiming "Dx9.1") . but someone correct me here.. i thought that FPRT where subnets of Dx9.(not required by M$ to have full compliance DX9 hardware) ,in the other hand HDR is currently in its infacy in -today- DX9 cards (not good enough) according to the latest comments of future technology by Valve ,something that will be corrected in the incoming hardware. the AA issue in Hl2 should be now done in a more propertly way.(without hacks). however all this is merely a guess about RUmored performance/features that ATI and NVidia next cards will have.in the end both cards can end being much faster in the case of ATI or much slower in the case of Nvidia, or vice-versa,than what is rumored . with technology you never know ,there could be problems ,with Heat/power issues/Stuttering/etc... and broken features that can delay by months any card .
jimbob0i0
02-15-04, 07:16 PM
Originally posted by Nv40
with technology you never know ,there could be problems ,with Heat/power issues/Stuttering/etc... and broken features that can delay by months any card .
Yup till CeBit all we have to go on for now are rumours. Anything could happen between now and then and it's impossible to ay which card will be the better buy until drivers and hardware are both avaliable for testing in a few months.
bkswaney
02-24-04, 05:39 PM
I just hope Nvidia's jump to 11nm does not hold up
future cards the way the NV30 was held up with 13nm.
I'm guessing not though. We already know the NV40
is built in 13nm. So my guess is we could see a the NV45
built on 11 but I think the NV50 will be the first.
Originally posted by retsam
this is my take on the whole thing. if both cards are very close in performace then nvidia will win.... all cheating aside... ati has had a very bad history with drivers for new cards (but they have gotten much better) and nvidia driver team is second to none .... i think just the problems alone for ati right out of the gate with a new card will persuade many people that stand on the fence to go nvidia oh well what ever
rets:afro2:
The R300 drivers were very good right from the start. I am willing to bet (short of an actual bug tally) that R300 and its drivers had fewer bugs in the initial release than NV30 did. (all cheats and hacks aside) The track record over the last 18-20 months on the release of new products for NV is far from another jewel in their tarnished crown. Thats something to consider when considering the NV40 for early adoption.
Spotch I think you are incorrect, but if you put all the cheats that NV said were bugs into the bug category then Nv has more bugs...
This is my reasoning ATI released a bajillion drivers and each one had new bugs so they went thru a whole lot more bugs than normal... i.e. that they fixed 4 and made 2 or fixed 2 and made 4 either way it is 6.
Originally posted by sxotty
Spotch I think you are incorrect, but if you put all the cheats that NV said were bugs into the bug category then Nv has more bugs...
This is my reasoning ATI released a bajillion drivers and each one had new bugs so they went thru a whole lot more bugs than normal... i.e. that they fixed 4 and made 2 or fixed 2 and made 4 either way it is 6.
I am not sure what you are saying here... Do you mean to say that when NV doesnt perform AA or Aniso as it is requested by the settings its a cheat? When ATIs Aniso doesn't work in one game or if AA doesn't work in another, its a bug?
I think that many of NV so called "cheats" were in fact bugs. Its just at the time of NV30-35's released, NV's reputation was under quite a bit of scrutiny. So at least some of these "bugs" may have been a bit overhyped as "cheats" by the ATI FanBoys and some hardware sites.
Malfunction
02-24-04, 09:18 PM
Originally posted by Spotch
I am not sure what you are saying here... Do you mean to say that when NV doesnt perform AA or Aniso as it is requested by the settings its a cheat? When ATIs Aniso doesn't work in one game or if AA doesn't work in another, its a bug?
I think that many of NV so called "cheats" were in fact bugs. Its just at the time of NV30-35's released, NV's reputation was under quite a bit of scrutiny. So at least some of these "bugs" may have been a bit overhyped as "cheats" by the ATI FanBoys and some hardware sites.
Thank you, it needed to be said and I am now glad it has been. :thumbsup: Heck, if you don't believe that the misinterputation of a fanboy is possible... you should check out the processor forum...lol. :D Shockingly so many unbias people end up being... well, bias. :D
Peace,
:afro:
Originally posted by Malfunction
Thank you, it needed to be said and I am now glad it has been. :thumbsup: Heck, if you don't believe that the misinterputation of a fanboy is possible... you should check out the processor forum...lol. :D Shockingly so many unbias people end up being... well, bias. :D
Peace,
:afro:
Dont get me wrong though... those lousey farging bastages tried to swindle us all with thier dirly lowdown lie, cheat, steal, Enron style crap...
(bur)
Well I guess thats enough.
StealthHawk
02-25-04, 02:44 AM
Originally posted by Spotch
I think that many of NV so called "cheats" were in fact bugs. Its just at the time of NV30-35's released, NV's reputation was under quite a bit of scrutiny. So at least some of these "bugs" may have been a bit overhyped as "cheats" by the ATI FanBoys and some hardware sites.
Considering that on several occassions they(NVIDIA) admitted to "overly aggressive optimizations," I don't think so.
Although I will agree that some things might have been bugs. re: Halo lighting exposed on Tom's Hardware. Splinter Cell shaded water was another bug. Certainly not everything was a bug.
vBulletin® v3.7.1, Copyright ©2000-2012, Jelsoft Enterprises Ltd.