PDA

View Full Version : DVI signal quality of bottom connector in 6800U


arielcohen
08-15-04, 02:25 PM
I received my eVGA 6800 Ultra a couple of days ago. I have a problem with one of the DVI outputs. When I connect my monitor to the bottom connector (the one close to the motherboard), I get occasional display problems. For example, when I use Internet Explorer to look at www.3dgamers.com, there's serious corruption if I place the window on the right side of the screen.

I believe there's a problem with the DVI signal quality. I have a Samsung 243T monitor running in Digital mode. I see the problem when using resolutions of 1920x1200 and 1600x1200, but not at lower resolutions. Also, there's no problem with the other DVI output (the one further from the motherboard).

I was surprised to see this DVI signal quality problem with the lower DVI output since I believe that this output is connected to the external Silicon Image TMDS, and I never had problems before with cards that use this TMDS.

Is anyone else seeing this problem?

JGene
08-15-04, 02:39 PM
Have you tried to tweak your resolutions in Power Strip? Maybe fine tuning your display timings could fix it.

arielcohen
08-15-04, 04:05 PM
Maybe, but I shouldn't have to do that. I may have a defective TMDS or a defective board which causes signal noise. I'm wondering if this is just a problem with my particular board, or if it's a board design problem with the 6800 Ultra which causes the signal quality problems. If I just got unlucky with my board, then I should exchange it, but if it's a common problem with these boards, then there may be no point in exchanging it.

JGene
08-15-04, 05:19 PM
Could be that you have a bad DVI port considering the other one works perfectly well. Do you have any other issues, such as stability or strange colors?

Have you tried calling eVGA customer service and see what they have to say?

arielcohen
08-15-04, 08:20 PM
If it was a bad connector, I think the symptoms would be worse. The image is fine most of the time. The problem is always triggered by a particular image, such as having the browser with the web site that I mentioned on the right side of the screen. The corruption consists of flashing horizontal lines and a flashing duplicate of the left part of the picture which appears on the right side. It definitely looks like a signal quality problem. The problem goes away if I reduce the color quality from 32-bit to 16-bit or if I reduce the resolution to 1280x1024 or lower. Even at 1920x1200, I'm using a pixel clock rate of 154MHz which is quite a bit lower than the maximum of 165MHz for the Silicon Image TMDS.

I'll call the eVGA customer service tomorrow.

JGene
08-15-04, 09:15 PM
True, but it still sounds like a bad connector to me. Maybe it wasn't soldered on right or maybe the traces to the DVI plug are faulty. Maybe it's even a bent pin, you never know. In any case, your card isn't capable of sending a high-bandwidth signal over that DVI port and it sounds like a good reason for an RMA.

Be sure to keep us informed of what eVGA has to say about it :)

arielcohen
08-17-04, 10:36 PM
An RMA was approved by eVGA. However, I had the chance today to try another eVGA 6800 Ultra card, and it had the exact same problem, so I don't plan to go ahead with the RMA since the replacement card is likely to have the same problem. The only reasonable choices are to return the card for a refund or to keep it. I plan to keep it since the problem is rare and it happens only on one port. If I start encountering the problem more (such as in Linux which I haven't tried yet), then I'll return it.

By the way, I saw the problem with three different monitors and three different DVI cables, so it's not a monitor or cable problem.

saturnotaku
08-17-04, 10:55 PM
This is a situation where video BIOS flashing would be an option. What is the BIOS version of your card? eVGA has made a new one available that could very well be newer than what you have and potentially fix this issue.

arielcohen
08-18-04, 12:38 AM
No, I tried the latest eVGA BIOS and it didn't make a difference.

This is a very strange problem. I just tried Linux, and the problem is much more serious there. Whenever I put a Firefox browser window on the right side of the screen, I get the corruption. Strangely enough, I don't get the problem when I try Mozilla instead of Firefox. Also, if I don't display the navigation bar on the Firefox browser, I don't get the problem. There must be some pixel pattern that triggers this problem. I tried two different NVidia Linux driver versions (the latest and the one before last) with the same results.

I guess it could be a driver problem, but it seems unlikely because I don't have this problem with my dual-dvi TI4800 using the same drivers. Also, why would the problem appear on one DVI port but not the other if it was the drivers?

I think that some more likely explanations would be that it's a BIOS problem (something is not getting set up right by the BIOS), or a design flaw. The design flaw could be in the board (more likely) or the GPU itself (less likely).

I have to say that I'm quite pissed off about this. I wish I could contact someone at NVidia about this problem. I hope a workaround is found at some point.

JGene
08-18-04, 02:28 AM
Wow wasn't expecting to see that bug show up even with a few different monitors and a different video card. Could it be a design flaw? Maybe, but it makes me wonder what bandwidth the lower DVI port is rated at...hmmm. I've had my own personal quirk with nVIDIA. (www.geocities.com/jgeneedid)

Anyways, here you go: Request support from nVIDIA (http://support.nvidia.com/Content/NVSupport/frmSupport.asp?RequestTypeID=1)

walkndude
08-18-04, 03:16 AM
Jgene its kind of annoying to watch you drag your problem into other peoples threads -you should know better and your time may be better spent at the root of the cause...

a display with firmware containing incorrect edid information(1280*1024 instead of the correct 1368*768) -your time may be better spent going after sharp for a firmware update instead of expecting nvidia's driver to break spec inorder to get you around the problem.

Sorry for going off topic a bit arielcohon, I dont have a fix for you but I do know that the few dual dvi cards manufactured by nv prior to the 6800 were limited to 1280*1024 on the secondary port when driving the primary above said resolution, which happens to be the same problem your running into...

Did evga specifically state this shouldn't be an issue when you contacted them ?

arielcohen
08-18-04, 10:42 AM
DVI port 1 is connected to a Silicon Image TMDS which can handle a pixel clock rate of up to 165MHz. It should be able to handle 1600x1200 and 1920x1200 without a problem. Even at 1920x1200, the pixel clock rate is only 154MHz. I have no problems with my TI4600 running at these resolutions.

The 1280x1024 limitation was for the internal TMDS in some previous-generation NVidia GPUs, not the external Silicon Image TMDS. By the way, it looks like they fixed that in the 6800, and they now have a really good integrated TMDS in the GPU. This is the one feeding Port 2 which works great.

My current suspicion is that NVidia tested the reference boards where port 1 was a dual-link port. There was one TMDS on the front side of the card and one on the back. Both were connected to port 1 to support dual-link DVI. Unfortunately, the TMDS on the back of the card is not populated in the boards that are actually shipping. Perhaps this created a problem which was not seen by NVidia when testing the reference board that had both TMDS chips. In principle, there should be no problem with one TMDS, but there may be a board design issue which is showing up only in the case where the chip on the back of the card is not there.

arielcohen
08-18-04, 11:38 AM
I just tried it with a Dell 2001FP monitor at 1600x1200. There's a similar problem, but the symptom is different. Instead of the image going berserk like it does on the Samsung 243T, it goes blank for half a second and then reappears on the 2001FP. It looks like it's losing synchronization.

I saw other reports on other forums about this problem. Something is definitely screwed up with Port 1 DVI.

ricercar
08-18-04, 07:31 PM
Don't use a Dell 2001Fp for testing video card DVI problems.

The Dell 2001FP has DVI receiver problems at 1600x1200 and higher. I don't want to clutter this thread, but you should google for more details before you trust this display for testing your video card DVI ports.

Code_Fuzion
10-26-04, 12:00 PM
DId you ever get a fix for this? I am running into the same probs. Used a Monster DVI Cable, etc. I am using a 6800 Ultra same as you and the lastest and past drivers. Nothing works! It these shimmering pixels to the right of the window (mostly white windows that create the problem) and in Web Browsing. It also occasionally blinks off and on during these windows. Works GREAT in Games but this annoying problem sux for windows. I am using DVI Port 1 and get the same exact problem. How do I set it up to use Port 2 DVI?

Code_Fuzion
10-26-04, 03:50 PM
No one else has this problem :(

atlas
10-26-04, 11:20 PM
FWIW: I had a problem with dual DVI on a 5700u. The port away from the mobo shows shimmering pixels on fine gradients at 1600x1200x32. I have 2001FP's. Never did get it resolved.

Oh, and a replacement card did the same.