What GPU's do the experts recommend? |
![]() |
| log in |
Questions and Answers : GPU applications : What GPU's do the experts recommend?
1 · 2 · Next
| Author | Message |
|---|---|
|
Hey guys, I have 3 systems that I would like to put a new GPU (or 2) into. | |
| ID: 1223657 · | |
I have been unsuccesful in getting windows xp, or windows 7 (i upgraded the system to Windows 7 64bit to see if that would help) to recognize the card. I have ran the Radeon drivers, the catalyst suite, etc.. and they keep giving the message that no compatible hardware is found. Funny thing is, the fan on the GPU is running when the computer is on so I'm not sure what that's about. I read somewhere that it could mean the RAM on the GPU is fried. Because I have the computer connected via the integrated graphic card's VGA port, when I boot the computer up, immediately after BIOS a screen comes up that warns me that I am using the integrated graphics and that I should connect to the DVI ports if I want to use the other graphics card. When you boot up the computer with the video cable connected to the video card (and not the integrated), do you get a display or not? I also have a core 2 duo Vostro 220 that has no additional graphics card so I'd like a recommendation on what to put in there. First things first: 1) Are the power supplies in these computers up to snuff? Make sure they are powerful enough to handle a newer video card that may draw more power. 2) What type of connection do you need for each computer: AGP or PCIe? 3) How much are you willing to spend for the video cards? | |
| ID: 1223660 · | |
|
1. I will try to connect to the non-integrated (DVI) port tomorrow, as the computer is at my house and the cables are at my apartment I didnt get to try it yet. | |
| ID: 1223665 · | |
2. For Vostro 220, PCI-E (here is the specs from Dell) For these two machines, I'd recommend any nVidia GPU that doesn't require an external PCIe power connector. Most cards that don't have the extra power connector are spec'ed to only draw the maximum wattage supported by the PCIe x16 slot itself and therefore should be fine on a 300w power supply. 4. The Dell Dimension 8250 has AGP and PCI, 250W power supply. This one is going to be "iffy". The power supply would likely need to be replaced to support any newer card. You might be able to get in a low-powered PCI nVidia card, but I'm worried about the PSU. 5. HP p6250t PCI-E capable, and 300 W power supply Same as 2 and 3 above. I'd spend around $100 each...I assume more expensive cards crunch faster? Does the size of the card really matter since BOINC only uses a certain amount per WU? More expensive cards are going to have faster GPUs, more "cores", and wider video RAM buses. Essentially, yes, they crunch faster, and if they have enough spare power, can even be configured to crunch more than one workunit per card. As far as "size" of the card, you unintentionally brought up a good point. Make sure that any card you buy can physically fit in the computer you want to put it in. More powerful cards can be quite large and not fit into smaller cases, though since you are not buying the high powered cards, this shouldn't be an issue. But I'm assuming by "size" of card, you either meant the video RAM or the power of the GPU. For nVidia cards, 256MB is the stated minimum, but 512MB is rather common these days. Going with 1GB is OK, but it won't be used unless you configure BOINC to crunch more than one workunit per card - but you wouldn't want to do that on lower end cards. So in summary, buy a card with at least 512MB of video RAM that doesn't require additional power connectors. If you're focusing on SETI@Home, you should probably buy nVidia-based cards and you'll really boost your workunit production tremendously. | |
| ID: 1223677 · | |
|
thanks for your help my friend, much appreciated :) | |
| ID: 1223683 · | |
|
by the way, if when i use a DVI cable to connect to that questionable Radeon 2600, find that it works, what would that mean? is it even possible that the display would be output without the drivers having recognized the card? (in device manager it is not listed, neither as an installed and properly working device, nor as hardware that needs a driver- no mention of it at all.. | |
| ID: 1223687 · | |
|
If you get an image on the Radeon 2600, then that indicates that the card is physically working. If you get an image and its messed up, then likely the video RAM is bad. If you don't get an image, the GPU could have fried or the card is bad otherwise. | |
| ID: 1223691 · | |
|
gotcha. guess i can take it from here, thanks again | |
| ID: 1223696 · | |
|
| |
| ID: 1223749 · | |
|
Hi BilBg, what is the main difference between the first and second cards you linked me to? They both have a GT 430 chipset I see, but it only looks like the second one is CUDA enabled? | |
| ID: 1223770 · | |
|
I'm sorry, I just re-read your post, I see that the first and third cards have faster memory. | |
| ID: 1223773 · | |
|
Additionally, I might be willing to add a little bit of a better card into the Vostro 220 model system, as the core 2 duo E7500 @ 2.9GHz actually performs quite decently. If you could point out which card is a "Step up" (say in the range of $130-$150) I might go that route for that one system. | |
| ID: 1223776 · | |
I am presently only familiar with CUDA because I have a GeForce 8800GTX in the system I am using right now, and that crunches using CUDA. You can't really find a "difference" as they are names of different things. "Fermi" is the codename of NVIDIA GeForce 4xx/5xx GPUs CUDA is the name of NVIDIA's computing technology present in all NVIDIA GPUs since 8xxx http://en.wikipedia.org/wiki/GeForce_400_Series http://en.wikipedia.org/wiki/CUDA EDIT: The problem for 2 GPUs in one system or more powerful GPUs to buy is the 300/250 W PSUs and space on the motherboard/case. Wait for other people to comment on this. (or post/ask in "Number crunching" (with link to this thread), you will get more/faster answers) ____________ - ALF - "Find out what you don't do well ..... then don't do it!" :) | |
| ID: 1223781 · | |
|
Ok as far as the older p4 with the AGP/PCI is concerned, they actualy have a PCI version of the GT 430 you suggested for the PCI-E systems, as well. | |
| ID: 1223783 · | |
|
I see, Fermi is essentially the new generation of CUDA enabled GeForce cards, which assumably are improved. I will click off the same checkbox in the lunatics optimized app as I did for my 8800M GTX then. That was mainly why i was wondering :) | |
| ID: 1223784 · | |
On your signal :) ;) My signal is "Yellow", wait for several other people's responses before you get "Green" signal :) Yes, the Lunatics' optimized CUDA apps are compatible with all generations of NVIDIA CUDA-capable GPUs (unlike stock SETI which supplies 3 different apps to different CUDA GPUs/drivers) ____________ - ALF - "Find out what you don't do well ..... then don't do it!" :) | |
| ID: 1223791 · | |
|
so far I ordered 3 of these http://www.newegg.com/Product/Product.aspx?Item=N82E16814121397 | |
| ID: 1224020 · | |
|
Will a given GPU perform "better" on one system compared to another? for example, will the same exact GPU model accomplish tasks and such at a more efficient rate on a Core 2 Duo E7500 compared to a C2D E4500 etc..? Or is the computer's specs rather irrelevant in regards to how an independent GPU will function on one system vs another? Knowing this answer will help me delegate which systems I put GPU's in first, or if it just doesnt matter. | |
| ID: 1224085 · | |
|
| |
| ID: 1224684 · | |
|
Thank BilBg, also that chart was very useful. | |
| ID: 1224729 · | |
Questions and Answers : GPU applications : What GPU's do the experts recommend?
| Copyright © 2013 University of California |