some strange reason i have a post to make-re NV30

[ben6]

I've been corresponding with NVIDIA the last few days on the Siggraph 2002 presentations. Here's what they had to say:

1. It will be marketed as the first Cinematic Shading GPU
2. Pixel shaders can be 1024 instructions in a PASS . Of course you can multipass, but the program instruction length is considered a pass
ALU functions for the Pixel Shader: ADD,DP3, DP4, LRP, MOV, MAD, SUB, X2D (add and multiply) ,TEX, TXD, TXP (texturing), COS, EX2, FLR, FRC, LG2, POW< RCP, RSQ, SIN (math instructions), SEQ, SLF, SGR, SGT, SLE, SNE, STR (set "on" instructions), DST, LIT, RFL (graphics), MIN, MAX (minimum , maximum) PK2H , PK2US, PK4B, PK4UB, PK4UBG (pack) UP2H, UP2US, UP4B, UP4UB, UP4UBG(unpack) , Kill (kill)
3. Vertex Shaders programs can be 65,536 instructions in length with the loops branches etc , it's also considered one pass .
Vertex Shader Instructions: ARL (4 component A0 and A1, ARR rounding instruction instead of truncating like ARL, BRA , CAL, RET (branchching instructions) COS , SIN (high precision trigonometric functions), FLR, FRC (floor and fractions of floating point values), EX2 , LG2 (high precision exponentiation and logarithym function), ARAm SEQ, SFL, SGT, SLE, SNE, STR, SSG
4. .13 micron
5. NVIDIA said fall part in June , and I've heard nothing from them that changes that outlook . despite the rumors.
6. NVIDIA will be using a form of DDR2 memory . This is CONFIRMED information. What clockspeeds , what bit interface, or memory bandwidth I
cannot confirm at this time
7. 12bit fixed point (sorry for confusion), 16bit 32bit float formats supported
8.16 textures per pixel ala DX9.
9. 16 texture units I cannot at this time confirm the configuration of those 16 units but they are there
10. John Carmack quote in it's entirety: Nvidia is the first of the consumer graphics companies to firmly understand what is going to be happening with the convergence of consumer realtime and professional offline rendering. The architectural decision in the NV30 to allow full floating point precision all the way to the framebuffer and texture fetch, instead of just in internal paths, is a good example of far sighted planning. It has been obvious to me for some time how things are going to come together, but Nvidia has made moves on both the technical and company strategic fronts that are going to accelerate my timetable over my original estimations

My current work on Doom is designed around what was possible on the original Geforce, and reaches an optimal impliementation on the NV30. My next generation of work is designed around what is made possible on the NV30
11. NV30 and DX9 schedules are now aligned. This may change if Microsoft delays DX9 , but not the other way around.
12. DX9 shift from bandwidth to computation. Pixel Shading is not pixel filling

Small correction on name

 

[KimB]

1. It will be marketed as the first Cinematic GPU

Hehe, ATI and their supporters will be having fun with this one

6. NVIDIA will be using a form of DDR2 memory . This is CONFIRMED information. What clockspeeds , what bit interface, or memory bandwidth I cannot confirm at this time

Interesting. I'm willing to bet that nVidia will use a 128-bit memory interface. This does give the R300 somewhere around a 20% theoretical memory bandwidth advantage. I'm very interested to see how this develops...

7. 12bit, 16bit 32bit float formats supported

Are all three formats floating-point? Or just the 16bit and 32bit formats? Personally, I don't see much reason for 12-bit float...

11. NV30 and DX9 schedules are now aligned. This may change if Microsoft delays DX9 , but not the other way around.

This is very interesting, and runs very, very contrary to the many people claiming that DX9 was based on ATI hardware.

12. DX9 shift from bandwidth to computation. Pixel Shading is not pixel filling

This may be nVidia's basis for claiming that 256-bit was "overkill." I wonder what this means for FSAA performance?

 

[BoardBonobo]

Cool...? What do they mean by some form of DDR2 memory? And doesn't all this run against the ATI<->DX9 partnership?

 

[Silent_One]

"My current work on Doom is designed around what was possible on the original Geforce"

Interesting...does he mean the original GF1 or the GF1 & 2 core?


"My next generation of work is designed around what is made possible on the NV30"

Now THAT will be really interesting....

 

[BoardBonobo]

For some really strange reason I feel this thread will cause a disturbance in the force

 

[Qroach]

Oh boy this thread is gonna get messy. Very interesting info Ben, it answered some quesitons I had..

 

[Qroach]

Ben,

Any idea if they will be posting more papers from siggraph?

 

[Geeforcer]

LOL, seems I am not the only one who is dreading what is going to happen here, considering recent events on the forum.

 

[BoardBonobo]

It seems we could all be coming down with a serious case of the DTs here

 

[Qroach]

I just need to get this striaght. I saw many people screaming murder because Nvidia wasn't DX 8.1 compatible, and when they are fully DX 9 compatible they still screamed murder becuase they NV went beyond the the DX9 spec? When in reality, they are fully DX9 compliant and it appears that ATI isn't?

That's a little too ironic imo. Ah well we needed to hear as much info as possible.

 

[Gollum]

I think its a little early to assume that ATI will not meet DX9 spec, their claims of DX9 compliancy appear fully confident to me. Maybe they'll be the ones ending up not being fully DX9.1 (if there will be such a thing), but could MS really raise the DX9 specs again so late in the game to favour Nvidia? That'd cause a major mess!

As for the other info, thanks a lot Ben! That's some truly interesting information straight from the horses mouth. 1024 instructions per pass? Question of speed still remains, but still ... uhm wow...

BTW, so far this thread still seems okay.

 

[LittlePenny]

1024 Pixel Shader instructions in a pass is quite amazing. I couldn't imagine anyone using this up any time soon(with the exception of NV technology demos ). Chalnoth mentioned in the 3dArch forum NV30 may be able to do some ray-tracing. Then, MfA pointed me in the direction of this paper http://graphics.stanford.edu/papers/rtongfx/ If NV figured out how to do real time ray tracing using their pixel shaders, it could be very interesting indeed!

edit:

The architectural decision in the NV30 to allow full floating point precision all the way to the framebuffer and texture fetch, instead of just in internal paths, is a good example of far sighted planning

I forget, does the R300 goto the framebuffer? I am pretty sure it doesn't goto the texture fetch. Please someone correct me.

 

[Geeforcer]

There is no evidence that R300 in not fully DX9 compliant. Should Microsoft make any last-minute changes, you can hardly blame ATI for that.

 

[KimB]

Interesting...does he mean the original GF1 or the GF1 & 2 core?

The GeForce 1 & 2 are identical from a programming perspective.

 

[stevem]

...1024 instructions per pass? Question of speed still remains, but still ... uhm wow...

How many clock cycles? Their papers specify 16 texture units & a previous thread summised an 8x2 arrangement, but what if it's a 16x1 arrangement - "Texture Computer" anybody...? I recall Anand relaying something about other DX9 architectures going with an 8 pipeline layout...

 

[Doomtrooper]

1. It will be marketed as the first Cinematic GPU

Nvidia is 5 months late on this claim, the R300 ran Lord of the Rings in Real Time along with technology demos also done in real time..scratch that one off.

DDR2 will not compete with a 256 bit bus, and I see no mention of one here. 1024 intructions is over kill....it would be dead slow running all 1024 in one pass.

Bens information is interesting but from past experiences has been 'dead' wrong 50% of the time.

I was expecting this type of rebuttal from Nvidia from the 9700 launch..they have nothing to compete with it so they will throw their PR into high gear.

 

[jjayb]

As an Ati user for my last 2 cards. First the Radeon, then the Radeon 8500, I say:

More power to them. I'm all for advancement. You won't hear any complaining from me. Some of us Ati fans aren't zealots. I won't get into reasons why, I'm just not a big fan of Nvidia. Ati gives me an alternative. If Nvidia comes out with some mindblowing tech, that means that Ati will adopt it eventually, so that's good for me also. If the Nv30 truly rocks the 9700's world, that just means I get it a lot cheaper. Definitely a win win situation for me.

 

[Mulciber]

Nvidia is 5 months late on this claim, the R300 ran Lord of the Rings in Real Time along with technology demos also done in real time..scratch that one off.

lol, and you screamed bloody murder when nVidia rendered a scene from Final Fantasy: The Spirits Within

DDR2 will not compete with a 256 bit bus, and I see no mention of one here.

Possibly, but it may save them a lot of money not only on ram prices but also board manufacturing prices. If the architecture isnt so limited by bandwidth then it wont really be necissary will it? Isn't it ATI who preaches a "smart" design over brute force?

1024 intructions is over kill....it would be dead slow running all 1024 in one pass.

Thats kinda what I get from the 8500 and Doom3. JC says that by all accounts the 8500s ability to run his effects in 1 pass should mean its faster than the GF4, but it isn't.

I was expecting this type of rebuttal from Nvidia from the 9700 launch.

So was I. It remains to be seen if they actually have a part in time to back it up.

 

[Doomtrooper]

There is nothing on those specs that the 9700 doesn't have...except for the 1024 intructions and manufacturing process of .13.
I was hoping to see some memory configuration information as DDRII would point to 128 bit bus IMO.
Looks like a interesting Xmas this year...bring it on Nvidia I need some competition to get my 9700 cheaper.

 

[Doomtrooper]

Nvidia is 5 months late on this claim, the R300 ran Lord of the Rings in Real Time along with technology demos also done in real time..scratch that one off.

lol, and you screamed bloody murder when nVidia rendered a scene from Final Fantasy: The Spirits Within

I've seen the Demo shown a Siggraph..maybe one should wait and see it themselves. I never screamed anything..must be thinking of someone else you stalk.