NV48 Cancelled

Well, Dave is most likely right about it being IBM 130 nm. Doesn't make much sense to take the NV40 and totally convert the entire thing to TSMC's 110 nm just to get PCI-E support.

Now, one person who might know is somebody with one of the new SLI rigs from NVIDIA with the two 6800 Ultras running at 425 MHz. If they could pop off those coolers and see if the HSI chip is there, then we would have a more definitive answer... Dave?
 
I had the Ultra's for a little while but they have been shipped to Hexus to use in an Intel event over the weekend, hopefully I'll get them back later. BTW - the default clock is 412MHz, with the ASUS PEG BIOS thing it pushes it up to 437MHz (but proceeds to lock the system).
 
Tim said:
Firingsquad has some HDR benchmarks, they show the 6800GT outperforming the 6600GT with the margin expected:

http://firingsquad.com/hardware/nvidia_nforce_4_sli/page13.asp

More HDR benchmarks from firingsquad:

http://firingsquad.com/hardware/far_cry_1.3_midrange/
Click to expand...
That's interesting, the results don't match the xbitlabs results. Either someone made a mistake when benchmarking, or something else is going on (there is one difference between the xbitlabs and firingsquad setup, xbitlabs used 16xAF).
Edit:
The 6800 HDR performance problem might be a AGP-only problem.
Click to expand...
I fail to see why AGP vs. PCI-E should matter, the data pushed over the bus with HDR or not is pretty much the same, isn't it?
It's also interesting to see that SLI didn't do much for performance when HDR was enabled neither...
 
PatrickL said:
Follow the link provided by josh and look at the chip.
Click to expand...
I did. It looks exactly like the NVidia logo on the NVidia page, and the distance between the letters is equal.
 
DaveBaumann said:
I had the Ultra's for a little while but they have been shipped to Hexus to use in an Intel event over the weekend, hopefully I'll get them back later. BTW - the default clock is 412MHz, with the ASUS PEG BIOS thing it pushes it up to 437MHz (but proceeds to lock the system).
Click to expand...

Beware when you're told that by local PR :LOL: Mine (the boards we were going to use) were just supposed to nip to Future for a couple of days :devilish:

Plus, default clock on the pair I had is 440MHz and PEG Link wasn't even an option for my mainboard. Yay for consistency when getting this stuff out to reviewers.
 
Rys said:
Plus, default clock on the pair I had is 440MHz and PEG Link wasn't even an option for my mainboard. Yay for consistency when getting this stuff out to reviewers.
Click to expand...

The Ultra's you had were XFX's though, weren't they? As for the PEG Link, I now have a BIOS update to apply that I believe removes it, just no 6800's as I accidentally sent the GT's to The-web-hack-formerly-known-as-mullet-man.
 
DaveBaumann said:
Rys said:
Plus, default clock on the pair I had is 440MHz and PEG Link wasn't even an option for my mainboard. Yay for consistency when getting this stuff out to reviewers.
Click to expand...

The Ultra's you had were XFX's though, weren't they? As for the PEG Link, I now have a BIOS update to apply, just no 6800's as I accidentally sent the GT's to The-web-hack-formerly-known-as-mullet-man.
Click to expand...

XFX as in NVIDIA told me to call them XFX when in fact they're just reference boards with a certain BIOS? In that case yes! Straying off topic a little, are XFX's Ultras blue at all (meaning they're having a go at making them themselves), or are they still shipping green boards with a branding plate applied (NVIDIA supplied stock)?

Looking forward to your SLI article at any rate, hope you're close to finishing it. If you're around tomorrow, I'll have him bring the GTs to the thing for you, if they're not back with you already (or on their way back).

Back on topic, I'm not entirely convinced that NV48 is a product we'll never see. I think we should see it, and I'm still pretty sure it might be shown next week, along with NV41, in the States. I could be entirely wrong though, and it wouldn't be the first time!
 
That chip at gamepc is an NV41.

EDIT: It is unfortunate that GamePC didn't dig a little further into this chip and tried testing the video portion of the NV41 and see if NVIDIA did in fact fix it with this product.
 
mczak said:
DegustatoR said:
mczak said:
A good question. If I'd have to guess, I would assume it's to denote the 2nd test system (system 1 is agp, 2 pci-express), though the article doesn't say which cards were agp, which pci-e (it's safe to assume that the 6600GT was pci-express).
Click to expand...
http://www.xbitlabs.com/articles/video/display/farcry13_5.html
Click to expand...
Well, yes, but where in that page does it say which graphic cards were pci-express and which were agp?
Click to expand...
It's fairly evident--given Xbit's previous reviews--that the 6600GT is PCIe and the rest are AGP. So the P4 might favor FC's HDR in some way.

Yeah, I remembered a later benchmark showing the 6800s appropriately ahead of the 6600GT, but I forgot it was FS. Good memory, Tim. (Or fortuitous clicking. ;)) FS's system config page is fubarred, tho: Brandon lists a P4, an A64 MB, and AGP and PCIe cards all in one confusing jumble. Nutty. If the 6800 and 6600GT are indeed on separate systems and buses, their performance difference may be attributed to something other than GPU speed.
 
I hadn't thought of actually verifying the scores myself :oops: , but a quick run-through on my own 6800 GT agp indicates performance results much closer to the ones obtained by FS, at least on the Pier map, with HDR on and set at the default level.

I can't run any detailed tests ATM, though.
 
Tim said:
Edit:
The 6800 HDR performance problem might be a AGP-only problem.
Click to expand...
Or, potentially, the problem was quietly fixed in a later revision of the core. I never tested it in my own machine (benchmarked, that is), but I will say that Far Cry is playable at 1024x768 on my GeForce 6800 with HDR on. Performance isn't great, but it is playable.
 
Luminescent said:
I hadn't thought of actually verifying the scores myself :oops: , but a quick run-through on my own 6800 GT agp indicates performance results much closer to the ones obtained by FS, at least on the Pier map, with HDR on and set at the default level.
Click to expand...
Well, I guess this makes it look like HDR was only broken in the first revision of the 6800's. Good to know.
 
I did some closer benchmarking and realized that my framerates are a bit lower than FS's on the Training Level; the framerate averages 35 fps (1024x768, trilinear, no aniso, HDR set to 2).

When FS benchmarked the agp version of the 6600gt (second FS benchmark link referenced in this thread), it found it to average 22.8 fps on the training map. Peculiarly, in its latest Nforce 4 review, the PCI version of the 6600GT obtains an average FPS rating of 38 fps. These tests were carried out on different systems, however, it seems that in both, the graphics card was predominantly the performance bottleneck (as performance scales with the res going from 800 to 1024) and both tests used similarly performing drivers 67.02 and 66.93.

It seems, then, that there is a difference in performance between PCI and AGP versions of the cards, although I'm not sure this is due to core revisions, as the agp 6600GT sports an already revised version of the NV40 core.
 
PatrickL said:
nVidia is written nvidi a. Seems strange that a machine write so badly on a chip.
Click to expand...
Look at the gaps between serifs on the I and D, and on the I and A. It's the same distance, the A just looks further away because it's tapered at the top.
 
And it isn't a memory bandwidth issue
Click to expand...
Are you sure? Do you have a benchmark that only uses FP16 blending (and not FP16 filtering) that compares NV40 vs NV43?

NV43 had a lot of tweaks to improve texturing performance - it's not just a cut-down NV40.
 
Luminescent said:
Below are Firing Squad's performance graphs of the 6600GT AGP (67.02) and 6600GT PCI (66.93) on the Training map, in HDR mode (which I assume is configured the same for both tests).

6600GT AGP:
http://firingsquad.com/hardware/far_cry_1.3_midrange/images/hdrtrg1024.gif

6600GT PCI:
http://firingsquad.com/hardware/nvidia_nforce_4_sli/images/j1024.gif

There is a 16 fps discrepency between both cards.
Click to expand...

How come there is almost no difference between running one card and running SLI in those examples?
 
You must log in or register to reply here.

Similar threads

S
Echo Point Nova (Steam)
Replies
0
Views
302
Scott_Arm
S
F
RDNA5 Speculation
Replies
2
Views
1K
Erinyes
E
C
Assassins Creed Shadows
2 3
Replies
53
Views
4K
Jupiter
J
Z
Does being messed about by women or people in general, turn you bitter?
Replies
16
Views
1K
Zombietoid
Z
B
  • Poll
Do you see another game other than GTA VI winning 2025 GOTY at TGA?
Replies
9
Views
931
Seanspeed
S
Back
Top