I know right now much isn't known about the nForce 4 chipset except for some Inq articles, but I'm wondering what to expect for built in 3D capabilities. I'm looking to build a new system later this year, and given that my work revolves around audio, I can wait to upgrade to a more capable seperate graphics card. Any ideas/rumors/informed guesses?
3D Capabilities of nForce 4?
- Thread starter drhicking
- Start date
Hopefully it will be possible to run integrated gfx alongside a pcie gfx card for dualmonitor support. Also, I DO wish they'd nuke the crappy PS/2 and serial/parallel ports. It's time for them to GO now.
GO THE F*** AWAY.
GO THE F*** AWAY.
Panajev2001a
Veteran
Guden, what a great idea: this way when XP locks because you have not installed all the USB drivers after the first install (you have not even received a chance ) you are really screwed .
) you are really screwed .Richteralan
Regular
PS/2 mouse refresh rate: adjustable to 200Hz
USB mouse refresh rate: locked @ 125Hz
I'd still prefer PS/2...
USB mouse refresh rate: locked @ 125Hz
I'd still prefer PS/2...
Well, since we never saw a refresh of the GF4MX integrated cores, I think it is pretty obvious that NVIDIA skipped the integrated GF FX product, and went straight to development of a NV4x based integrated video. With Intel already having hardware that supports SM 2.0, NVIDIA would be awfully silly to do anything less. The NV3x architecture is dead, and I hope they leave it that way!
Of course, this is speculation on my part, or perhaps wishful thinking?
Of course, this is speculation on my part, or perhaps wishful thinking?
Haha, define "way too big". If NVIDIA is now offering a 8 pipe/3vs unit in a $149 card, do you still think a 2 pipe/1vs unit integrated into a AMD Athlon 64 chipset is too big? Maybe for a single chip, but I doubt it. The last VIA southbridge was only comprised of about 300,000 gates, so southbridge functions do not take that many transistors overall.
Remember, they did say "top to bottom". The Athlon 64 platform has been around too long without a decent integrated video chipset (and the K8M800 is not exactly decent).
Remember, they did say "top to bottom". The Athlon 64 platform has been around too long without a decent integrated video chipset (and the K8M800 is not exactly decent).
I'm saying that it is possible that they use a 1/4 of that 8x3 design, and they can still make it on the 110 nm process, along with all the other nForce 4 features, and keep it under 35 million transistors. Plus, the integrated video on that chipset is not going to be running at 500 MHz (I would guess at that size it would run at either 250 MHz or 300 MHz). Note the cooler on the regular GeForce 6600, it is not terribly impressive, and since the chipset will not have nearly as many transistors as a full blown NV43, it will not require as much cooling.
There will probably be other aspects of the NV40 architecture that will not be included, like HCT support, possibly the video encoder, and who knows what else. I think they could easily, and economically, integrate a 2 pixel pipeline, 1 vertex shader video unit into a chipset for the Athlon 64 (remember, they don't have to have a dedicated memory controller for that part), keep it well under 50 million transistors, and keep it cool and cheap by using the 110 nm process.
That is what I mean.
There will probably be other aspects of the NV40 architecture that will not be included, like HCT support, possibly the video encoder, and who knows what else. I think they could easily, and economically, integrate a 2 pixel pipeline, 1 vertex shader video unit into a chipset for the Athlon 64 (remember, they don't have to have a dedicated memory controller for that part), keep it well under 50 million transistors, and keep it cool and cheap by using the 110 nm process.
That is what I mean.
thats just the thing . I don't think they can make that small enough.
a 5200 on the other hand I can easily see.
Also high clock speeds don't matter . Though they would need decent enough clock speeds for it to be a viable part . But it would still produce alot of heat the bigger it is .
Which is exactly why I think they will have an nv3x part 4x2 at around 30m transistors
This will be a much more viable part in my eyes than a very hacked up nv4x part
a 5200 on the other hand I can easily see.
Also high clock speeds don't matter . Though they would need decent enough clock speeds for it to be a viable part . But it would still produce alot of heat the bigger it is .
Which is exactly why I think they will have an nv3x part 4x2 at around 30m transistors
This will be a much more viable part in my eyes than a very hacked up nv4x part
Well, NV could cut it down even more by not putting in a VS unit. I can easily imagine a 2 pipeline NV4x based integrated product without all the other bells and whistles. From what I understand, each NV4x pipeline is smaller than a NV3x pipeline (4x NV35 = 125 million tran vs 16 x NV40 = 225 million tran- and this is really basic as it doesn't consider the extra transistors used for 6 VS units in the NV40 vs 3 VS units on the NV35, as well as the video encoder, and other odds and ends). When you come down to it, it would make a lot of sense to use a 2x1 NV40 product for integration, as it would be much smaller than a NV3x 4x2 as you had mentioned.
JoshMST said:
lol whats the point of a nv4x part with no vs units haha .
THe nv40 pipelines are most likely smaller because of the 1 texture unit.
If needed they could modify the nv3x to have only 1 texture unit per pipe.
Nvidia said they spent 60million transistors on sm3.0 .
I don't see where these 60million transistor added for sm3.0 are gonig to go .
Even with a shrink and vs out there is still going to be a sizable chunk added on.
I just don't see it happening.
Its the same reason they had geforce 2 with no tnl built into the nforce 2 boards . Instead of geforce 4 or fx parts .
Size , heat and cost problems
You can't do that. It wouldn't be able to texture.Sage said:
And by the way, jvd, a 2-pipeline NV4x part with other cost-savings should be roughly 1/8th the transistors of the 16-pipeline NV40, so that puts it at around 30 million transistors. Given that the pixel pipelines aren't all of the core, 35 million transistors for a full 2-pipeline NV4x shouldn't be hard to believe.
I really don't know why you keep saying the NV4x is too big for value products.
Pete said:I was under the impression that the nF1 used a full-featured GF2MX, and the nF2 a full-featured GF4MX.jvd said:
Besides, a "GeForce" without hardware TnL would be pretty close to false advertising, no?
I'm pretty sure it doesn't as my friend keeps getting an error with his . games saying it doesn't have hardware tnl.
haha how do u come to that
I agree with Chalnoth on those numbers, a stripped down NV4x derivative with 2 pipelines and 1 vertex shader could easily fit into 35 million transistors.
As for speed, the NV4x architecture is pretty efficient, and I think overall it would outperform the FX 5200, and definitely whoop it up when PS 2.0 rendering is involved. Perhaps in some older games the standalone 5200 would beat it, but mainly due to memory bandwidth available to the integrated graphics.
As for speed, the NV4x architecture is pretty efficient, and I think overall it would outperform the FX 5200, and definitely whoop it up when PS 2.0 rendering is involved. Perhaps in some older games the standalone 5200 would beat it, but mainly due to memory bandwidth available to the integrated graphics.
JoshMST said:
I don't get how your coming to that conclusion.
Even if a 2x1 nv4x is 1/8th the pipeline tranistors . There is still the sm3.0 transistors . Which nvidia claimed took 60million.
So even if we hack out all but 1 vertex shader i would still say it will be much more over 35 million transistors . I would have to say it be very close to 70m .
Lets assume you can even make that . What would the performance be ? 1 vertex shader ? . I would say it be slower than a nv3x part.
Similar threads
