• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

NV30 tid bits of information 48GB/sec using 1GHz DDR-II

NFS4

NFS4

No Lifer
First some impressive information on how fast it is : whilst playing Quake 3 at 1280x1024 with FSAA 4x NV30 gets 2.4 times the score GF4 TI 4600 gets; Playing Doom 3, NV30 gets 3.5 times the performance Ti4600 shows; NV30 gets 3 times more points over at 3dmark 2001 then that ol' card 4600.

Also, the pixel fill while using AF x2 is 2.7 faster then 4600, and NV30 can run 200 million triangles per second, which is quite impressive. Finally : The banwidth on the new board gets to the speed of 48GB/sec using DDR-II that will run at 1Ghz; but it's in doubt that samsung will supply them any time soon.
Click to expand...

http://www.neowin.net/

😛😀
 
ati will need a maxx for sure now 😛

But will it be available this year? So much is dependent on manufacturing process and samsung getting DDR-2 in quantity at 1 Ghz speeds(250 mhz QDR)
 
yeah sounds a little exaggerated
DDR2 is not a wise decision to go with yet unless they plan on putting this thing off for awhile
watch um put some ddr2 on there
then the ddr2 prices jump and nvidia is in trouble
bet they would cost like 450+ bucks at release
1ghz speed mem sounds fun tho
 
It all sounds promising, but I personally would rather not see the hype machine go into overdrive until NV30 is actually released. This gives nVidia's detractors less ammunition to fire.

I tend to agree with a newspost from Mike at NV News myself
More NV30 Remarks - 8/06/02 5:42 am - By: MikeC - Source:
Looks like Scott any myself are on the same page in regards to the NV30 story that Reactor Critical made available yesterday. I ran across information from what seems to be a reliable source, but would rather not provide a link since the person mentions being briefed on the status of the NV30.
I appreciate the comments as they were provided as an attempt to clarifiy certain points. I guess there are times when one reads enough nonsense and can set the "record straight" without breaking a non-disclosure agreement. Of course I could be way off here, but my gut-feeling is that this person is telling the truth.


The graphics processor throughput of 200 million triangles per second is not correct. The sources may have mistaken the figure.

The 3DMark performance was exaggerated. As Scott mentions below the results are more than likely based on using extreme graphics settings. However, there's a good chance that the NV30 will significantly outperform the GeForce4 Ti 4600 in the next version of 3DMark.

The Doom 3 performance estimate was based on using the latest NV30 beta board and was not done via simulation.

NV30 beta chips and boards have been around for some time and the final chip should come back any day.

The 48GB/sec of memory bandwith is based on a calculation which includes factoring in overdraw removal (occlusion culling).

The German web site K-Hardware also has put together an artice which covers the type of memory the NV30 can use based on current market conditions. Click the headline to check out the English translation via Google.
Click to expand...

Greg
 
48Gb of bandwidth cannot be achieved using standard 256bit memory bus, even when using fastest DDRII memory. The max you can do is 30Gb/sec. 48Gb/sec figure is grossly exagarated.

Leon
 
I read elsewhere (yes, good source but hey...) That the 48GB/s is effective memory bandwidth, taking occlusion culling and various other technologies into account. Meaning that the actual bandwidth is considerably lower, although they didnt say how much lower.
 
hardocp news article on this

NV30 Facts?
Reactor Critical has some comments about NVIDIA and NV30. Then they share something a "friend" told them:

"In Quake III Arena in 1280x1024 with 4x FSAA enabled, NV30 is going to be 2.5 times faster than the GeForce4 Ti4600. In The Next Doom the board based on NV30 will be able to show 3.5 times or or even more of the performance the current Nvidia`s flagman has to offer us there. NV30 will score three times more than the GeForce4 Ti4600 in 3D Mark 2001."

Hmm, 36,000 3DMarks or so? I think not.

I must agree with them, this is pretty far fetched.
 
Here is something From Nvnews.



quote:
--------------------------------------------------------------------------------

Looks like Scott any myself are on the same page in regards to the NV30 story that Reactor Critical made available yesterday. I ran across information from what seems to be a reliable source, but would rather not provide a link since the person mentions being briefed on the status of the NV30.
I appreciate the comments as they were provided as an attempt to clarifiy certain points. I guess there are times when one reads enough nonsense and can set the "record straight" without breaking a non-disclosure agreement. Of course I could be way off here, but my gut-feeling is that this person is telling the truth.


The graphics processor throughput of 200 million triangles per second is not correct. The sources may have mistaken the figure.

The 3DMark performance was exaggerated. As Scott mentions below the results are more than likely based on using extreme graphics settings. However, there's a good chance that the NV30 will significantly outperform the GeForce4 Ti 4600 in the next version of 3DMark.

The Doom 3 performance estimate was based on using the latest NV30 beta board and was not done via simulation.

NV30 beta chips and boards have been around for some time and the final chip should come back any day.

The 48GB/sec of memory bandwith is based on a calculation which includes factoring in overdraw removal (occlusion culling).

The German web site K-Hardware also has put together an artice which covers the type of memory the NV30 can use based on current market conditions. Click the headline to check out the English translation via Google.

--------------------------------------------------------------------------------

It appears it was just more exagerated rumours floating around the web in an attempt to overshadow the 9700's thunder. How can it (the quake 3 scores) be done with a beta board if the damn thing " Has not taped out yet"?
 
Originally posted by: Dean
Here is something From Nvnews.



quote:
--------------------------------------------------------------------------------

Looks like Scott any myself are on the same page in regards to the NV30 story that Reactor Critical made available yesterday. I ran across information from what seems to be a reliable source, but would rather not provide a link since the person mentions being briefed on the status of the NV30.
I appreciate the comments as they were provided as an attempt to clarifiy certain points. I guess there are times when one reads enough nonsense and can set the "record straight" without breaking a non-disclosure agreement. Of course I could be way off here, but my gut-feeling is that this person is telling the truth.


The graphics processor throughput of 200 million triangles per second is not correct. The sources may have mistaken the figure.

The 3DMark performance was exaggerated. As Scott mentions below the results are more than likely based on using extreme graphics settings. However, there's a good chance that the NV30 will significantly outperform the GeForce4 Ti 4600 in the next version of 3DMark.

The Doom 3 performance estimate was based on using the latest NV30 beta board and was not done via simulation.

NV30 beta chips and boards have been around for some time and the final chip should come back any day.

The 48GB/sec of memory bandwith is based on a calculation which includes factoring in overdraw removal (occlusion culling).

The German web site K-Hardware also has put together an artice which covers the type of memory the NV30 can use based on current market conditions. Click the headline to check out the English translation via Google.

--------------------------------------------------------------------------------

It appears it was just more exagerated rumours floating around the web in an attempt to overshadow the 9700's thunder. How can it (the quake 3 scores) be done with a beta board if the damn thing " Has not taped out yet"?
Click to expand...



Ummm, this is posted already about 4 posts up
 
Finally : The banwidth on the new board gets to the speed of 48GB/sec using DDR-II that will run at 1Ghz; but it's in doubt that samsung will supply them any time soon.

So they say DDR-II transfers 4.8 bytes per clock? How do you transfer the ".8"?

Obviously this "article" is bung. Move along people, nothing worth reading here...
 
48 GB/s!!! :Q

how much actual real world performance will DDR-II add to a graphics solution over DDR? Is it about the same 20-30% of claimed improvement when used as RAM?
 
"how much actual real world performance will DDR-II add to a graphics solution over DDR? Is it about the same 20-30% of claimed improvement when used as RAM? "

That's easy go back to some reviews from then the GeForce first came out and look at the SDR version vs the DDR version.

Thorin
 
Originally posted by: thorin
"how much actual real world performance will DDR-II add to a graphics solution over DDR? Is it about the same 20-30% of claimed improvement when used as RAM? "

That's easy go back to some reviews from then the GeForce first came out and look at the SDR version vs the DDR version.

Thorin
Click to expand...
Yes, but this isn't the same kind of leap.

DDR-II still only transfers date twice per clock just like standard DDR memory. It's basically a revamped DDR architecture to allow for higher clock speeds and on-die termination. The 1 Ghz DDR-II speed you guys are seeing is actually only 500 Mhz double pumped.

(500mhz * 2) * (32 bytes memory width) = 32 GB/sec.

A 128 bit memory interface would give you 16 GB/sec, and a marketing person could easily multiply the 16 x3 with overdraw and occlusion culling to get the theoretical 48 GB/sec.

Edit: I meant x3, not 4.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top