Purchased Poweredge T620, with a NVIDIA Tesla C2075 CUDA card in for use as a high end workstation and CUDA machine.
Due to the programs I must run I have set the machine up dual booting Windows 7 and Linux (this in itself took some doing).
The graphics are currently being served by the onboard Matrox G200.
I have managed, on both systems to install the CUDA graphics card, for use in programming. However I decided to attempt to use another graphics card (NVIDIA Geforce GT610) as on Linux, installing the NVIDIA drivers stops me being able to access my monitors full screen resolution (a problem that I think will be more easily solved if graphics are being handled by an NVIDIA chipset on an external card). Additionally the Matrox G200 is not going to be sufficient for some of the programs I am going to be required to run in Windows.
On installing the GT610, no output was displayed (both before and after driver installation), I discovered this is also the case if I plug a monitor into the CUDA card. It seems there is some BIOS/iDRAC setting preventing output through any additional PCI graphics card.
I have tried disabling the onboard graphics controller in the BIOS, however this merely left me with no display at all, then requiring me to reset the BIOS via the jumper, a process I would prefer not to have to repeat. I also speculatively, adjusted the iDRAC network setting "Enable NIC" to "Disabled" but with no luck (though I have not tried disabling the onboard graphics via the BIOS since this change). Windows sees all graphics cards as installed and working, however does not see any displays connected to the,
If anyone has any ideas as to which settings need to be changed to enable output from PCI graphics cards, that would be amazing!
With most Dell servers, they won't support using a 3rd party video card in the server. The 12th generation servers differ in the fact that you can use the 3rd party video cards in them, but in a GPGPU use. Not for the sake of video display. So they support 3rd party video cards, not for using for display, but as a graphic processor.
Let me know if this answers your question.
In my T610, i installed a XFX GeForce graphics card which i modded to fit in a x8 slot and then within BIOS, set the "Embedded Video Controller" to "disabled". Upon connecting the monitor to the XFX card, i could the view the machine booting up and windows 7 had no issues with aero interface (the onboard matrox g200 would not work well with aero).
So I would be very surprised if the T620 could not handle a PCIe video card and in fact I understood that one could order Quadro cards with their systems which were a supported option. What you may have is some conflict occurring when you use a graphics card in combination with a GPGPU card. If the T620 is not a production machine, i would remove the GPGPU card and drivers and install a graphics card, set the BIOS and connect the monitor to the graphics card. If it works, i would install the GPGPU card again and see if they both work.
Maybe the bods at NVidia can shed some light on this issue as too often you on this forum one hears 'not supported'.
too often you on this forum one hears 'not supported'
Sky, you know as well as anyone why that is the case. Far and away, most of the most Dell-server-knowledgeable people use servers as they were designed, so unless someone has attempted to do something "off-track", there isn't much we can tell them, except to help them understand what they are up against. It IS unsupported, because it was never considered in the design and testing for that system, and in some cases, some versatility had to be sacrificed to ensure a functional and reliable system, which is usually more beneficial than a system that can be tinkered with at will. A lot of people who jump from desktops to servers can't understand why they can't just plop a video card in, why they can't just buy a drive for it from Best Buy, and why they can't just install XP on this "computer". We never say it won't work - simply that Dell didn't intend it and if they want to try it, it may or may not work, and may or may not require some trial and error, and we often try to help them do just that (specific cards are different from installing an unsupported OS though) ... but they need to know that from the outset.
Thanks for the suggestion, I will perhaps try this, the complication (as you may know) is that if, on turning the onboard graphics off, the computer does not find the graphics card, the only way I can see to get the onboard graphics back is through a jumper reset (the jumper in case being buried under all the fans) which is awkward, and also resets the hard drives to not ATA (which I had to set them to to be able to get Linux to see the Windows OS) which the causes me to have to reinstall Linux after resetting the hard drives to ATA (not ideal).
I am not sure it is a conflict however, even without the additional (GT610) graphics card installed, I also cannot get output from the CUDA card, which came with the system. It is this that is leading me to believe there is some configuration/setting preventing output to additional PCIe video cards. As you mention it is a bit of an issue, not only will the matrox g200 not support aero, it also doesn't support many other things, such as opengl etc.
I understand (now) that the graphics cards are not 'supported' in this type of machine for video display. As the cards can be installed and recognised by the system, I am wondering if there is anything that can be done to enable output, or at least to let me understand what about the machine is preventing output to the pcie cards. Other similar posts, seem to suggest it is an issue to do with DRAC/iDRAC causing the conflict.
@theflash, yes you are correct as servers are servers and do a specific job where video is not considered a priority, hence the provision of on-board matrox G200 which integrates rather nicely with the DRAC.
So i really should have re-read what i wrote before posting as it is a little incorrectly and awkwardly worded (cut and paste haste). My comment didn't really convey what i wanted to say. Suffice to say it would have been more accurate to write that graphics cards are not supported by Dell on the 11G and older servers and as such you will usually get the 'not supported reply for that reason. And as i first suspected there is some driver/hardware conflict between TELSA and GeForce, it could be better to initially seek an explanation from NVidia.
Now what is still confusing for me is that I understood that graphics cards were supported on the 12G servers. I even went through the exercise to configure a Quardo cards on a T620 12G servers when they were first released (but at the time of my post i could not even find a TELSA card in the T620 config options let alone a Quardo card)! So considering, Chris' response that graphics cards are "not for using for display, but as a graphic processor" is a little confusing. Is Chris' comment indeed correct and has something changed since the initial release of T620 or have i misunderstood something?
@ Joe.s, the TELSA cards I have seen do not have a video connector on them as they are designed for number crunching CUDA work, not video output. What card do you have?
As i mentioned, i would remove the TELSA card and and try the Geforce card on it's own, but drivers can be an issue. Note that drivers are common between TELSA and Quadro cards but Geforce cards uses different drivers. IFAIK, Geforce cards can not work with Quadro drivers so this may actually be your problem. Do you have access to a cheap NVS card you could try to install as it should save you from first having to remove the TELSA card and its driver. If not, you should find that GeForce card will work with a Geforce driver BUT later you will again have driver issues when you reinstalling the TELSA card.....
And I'm not sure why you have multi-boot issues and have to play within BIOS settings but i'm surprised you have to reinstall linux rather than just set BIOS to ATA before the first boot. With such a powerful machine, have you considered virtualization?
@Baff, the issue is not PCI vrs PCI-X, rather it is that the Matrox G200 can not be disabled in BIOS and thus add on graphics cards result in some form of conflict that stops the machine from booting, hence the hit and miss affair in finding a working graphics card. I have yet to read verifiable multi user confirmation that a specific add on graphics card works in 2600, 2900 2950 series servers. And one must note that server BIOS is constructed for stability, not a consumer focused overclocking culture that gamers are used to. I just wish Dell did release an updated BIOS that would allow disabling the on-board graphics on these older servers. This may just attract a younger crowd to old PE server hardware and familiarize this crowd with Dell which may just help sales in future (cheap investment really)...
@Skylarking Thanks, I will try your suggestion, of removing the CUDA card and drivers later today/tomorrow and see if I am able to use just the Geforce card.
I have the Tesla C2075, which has a single DVI output on it, I was not planning on using it for video output, but merely tried connect to it, to see if that improved matters, given that it came with the system.
What I find confusing, regarding the driver front, is that on disabling on board graphics (in the BIOS) I get no output to any of the video cards during the boot procedure, I thought that if there were a driver conflict issue then this wouldn't kick in until the OS starts, which was leading me to think there is something in the BIOS/iDRAC preventing video output to ANY video card. Please correct me if this assumption is not true!
Virtualization is a possibility, but from my understanding, it is difficult/impossible to utilise the CUDA card, whilst within a virtual machine.
I am confused why a PE T620, which was meant to have graphics capability, now seems not to allow Quadro or Telsa card selection when configuring to buy on-line !
Anyway, remember that you can not set 'integrated video adapter' to 'disabled' within the BIOS unless the BIOS detects a PCIe graphics card is installed in the system as this option would be grayed out (i am assuming T620 BIOS functions the same as T610 BIOS in this respect).
Now I'm not familiar with the Telsa cards and the C2075 specifically, but this NVidia PDF clearly states the card had dual link DVI connector supporting up to 1600x1200 resolution!
On this basis, the T620 BIOS should see this card as a graphics card and thus allow you to set the 'integrated video adapter' to 'disabled' within the BIOS (without any other graphics card installed in another PCIe slot within your system). If this is indeed the case, it would confirm that the BIOS does actually see the Telsa card as a graphics card in which case it may simply be you could have a faulty Telsa card.
To confirm if the Telsa card is faulty or not, you could try installing it in a consumer PC where it should output a signal to the monitor during boot process. Doing this removes some of the uncertainty around what capability the PE T620 actually has and instead focuses on the Telsa capabilities. If your Telsa card doesn't work in a consumer PC, I would suggest going back to the seller and discussing it with their tech support in view of getting the card replaced! If Dell was your seller, then i wouldn't even bother with testing in another PC and lay the issue at their Tech Support.
If the PE T620 BIOS does not allow you to set the 'integrated video adapter' to 'disabled' within BIOS when only the Telsa card installed, then the BIOS does not see the Telsa card as a graphics card, hence why you see no video signal output on the Telsa DVI connector. In this instance, I would still install the Telsa card and confirm it outputs video in a consumer PC before seeing the T620 (BIOS?) as the faulty component. Again, in this instance I would directed this issue to Technical support (seller or Dell).
If you have confirmed that the Telsa card is faulty, then by all means try the GeForce card in the T620 (obviously without the Telsa card installed) and see it that fairs better at displaying video during the boot process which it should. If you still get no video, then assuming the GeForce card works in a consumer PC, the issue is definitely with the T620 and again I would go back to technical support..
Some consideration needs to be given as to which PCIe slots have you tried using. Have you read the T620 HOM in relation to expansion slots which indicates GPU has 2nd priority in suggested slot to use is 4,5,7 and 2 and that the GPU enablement kit must be installed? I don't know what a GPU enablement kit is and have found no info on it but something to consider!
I guess it's worth mentioning multi (SLI) graphics cards in a system need some licensing for it to work, so this could be a cause of concern. It could simply be that the Telsa graphics card will not play together with the GeForce graphics card, with respect to how the SLI licensing is done. Quardo and GeForce have different licensing models from what i have gleaned. It could be that the card BIOSs may have some handshaking to do with the system BIOS with respect to SLI and this is where one conflict might occur. Further, when the OS is running, Telsa card may not work with a Geforce driver and GeForce card does not work with a Quadro(Telsa) driver.So you may still see issues outside Windows as I don't know what the graphics card BIOS does at boot time when multi GPU detected!