Memory Transfer Rate with PCIe 1.0 vs. 2.0

hybridremix

Distinguished
Mar 30, 2009
5
0
18,510
I'm trying to upgrade my video card on my somewhat old, but still powerful, motherboard. I'm not in the market for a full system upgrade so I'm not going to replace the mobo, not for a while. All I'm looking to do is maximize the performance of the system I have.

My question revolves around card performance when limited by the PCI bus speed. My mobo has PCIe x16, and I'm contemplating upgrading to a PCIe 2.0 card. I'm debating several cards, trying to balance cost and performance.

I know the 2.0 cards are perfectly backwards compatible to my mobo, but how much of a performance hit should I expect? Looking at detailed specs [http://www.hardwaresecrets.com/article/132] I know that the 2.0 cards I want have more than double the memory transfer rate of the 1.0 card I want, but what I can't figure out is if those numbers will be halved by my 1.0 bus.

Specifically I'm debating between an 8600 GTS [the 1.0 card I like], several 9600s and a couple 9800s. The 9 series cards range from 1.8 to 2.2 times the transfer rate of the 8600. According to the above table, the specs provided for the 9 series cards are based on 2.0 bus, which I obviously don't have. Should I expect, for example, a 9800 GTX to only use half it's potential because of the bus, or is the performance cut going to be small?

I guess the simplest way to word my question is this: Is my mobo going to limit a 96 or 9800 card so much that I won't see much better performance, and therefore the 8600 is a better buy?
 
It really just isnt an issue. If you was benchmarking and wanted ever ounce of performance from the PC then it might give you a very small differance but in the real world you wont notice any differance.

Mactronix
 
Somewhat old but still powerful? The PCIe slot 1.0 as everyone has already stated won't hold you back hardly at all, not enough to give it a second thought.
But, depending on the processor you have on that old but powerful motherboard, you could really be bottlenecking your video cards performance with a weak CPU.

What older but powerful processor do you have, and how much memory?
We will let you know where, what, and why you should direct your worries about causing the new video card to underperform.
 

hybridremix

Distinguished
Mar 30, 2009
5
0
18,510
My processor isn't what's old but powerful, my mobo is. It's an MSI K8N Diamond Plus from about 3 years ago, with it's stupid socket 939. I guess I never thought to look up AMD's plans for 939 back when I was building this system. It's really tough to find a cheap CPU upgrade when everything is socket AM2 and beyond. Upgrading from my Athlon 64 3000+ is also on my list, since it's only got 1.8 Ghz, but it's been tough. Sure wish I had looked into it more and I could work with a more modern socket like AM2 or something.

I appreciate all the responses. I'm glad to hear I can focus strictly on the graphics card's power without worrying about the mobo slowing it down.
 

mrmmacd77

Distinguished
Mar 26, 2009
27
0
18,530





Wow I'm kind of in the same boat as you hybrid... I'm upgrading my video cards aswell from twin 7800GT OC to 9800GTX + just waiting on a few responses from people on this forum... I also have a similar board as you... But I'm buying for future so I wouldn't worry about limiting your cards now... I would be worrying about getting enough for your next system to make you happy down the road.... and F AMD for getting rid of the socket 939 lol.... guess its time to retire the K8n for both of us and move over to intel's I7 series huh....
 

rewindlabs

Distinguished
Mar 7, 2009
1,181
0
19,290
I was also in the same boat as you with the whole having 1.0 pci etc

But i ordered a GTX 260 and no problems so far...of course the 260 will be transferred to new hardware soon enough...paired with an amd 6000+ atm
 

maximiza

Distinguished
Mar 12, 2007
838
3
19,015
I did allot of reading on this. 1.0a or 1.1 pci-e slot verses a pci- 2.0 slot. You will get only 1-5% loss in performance using a new 2.0 card in a older 1.* slot. No matter what you will still take advantage of the new technology and speed. Timing is everyting, pci-e 3.0 comes out by end fo this year so all 2.0 cards will be going down in price. From what I read 3.0 slot cards will be mind blowing amazing bandwidth with allot less overhead compaired to 1.0 and 2.0 pci-e slots. Check out this very old article.

http://arstechnica.com/old/content/2007/08/pci-sig-announces-transfer-rate-of-future-pcie-3-0-specification.ars

I got a socket 939 too. It works fine in SLI.
 
G

Guest

Guest
im waiting to upgrade my mobo which is 1.0... i currently have a gtx 260 and it is clearly limited by the board. i did also by coincidence just switch from the gs8600, yes the 8600s are crap. i was inspired by a friends 9800gtx lim ed (with the full gig of ram) on his 2.0 board. his actually out performs mine by a noticable margin even with the better card. his proc is a lil better but i feel that as we have identical ram the 2.4 gig proc to the 2.8 gig proc isnt enough to explain the difference in our performance. i feel its the board. that said, compared to my 8600, ill enjoy this for a few months before i start sli on the 260s and max out farcry! till then

Ren
 

cblair

Distinguished
Dec 10, 2008
9
0
18,510
Sorry to perhaps revive an old thread but I was just curious how you tell if a card's bandwidth is under or over the spec of the slot. For instance my 7950 says Memory Bandwidth is 44.8 GB/s but the specs say that my PCIe x16 has 4 GB/s bandwidth. Uhhh 4 <<<< 44.8 ?? Am I missing something? Where is the spec on the card that shows that 4 GB/s on the PCIe x16 slot is higher than the bandwidth of the card?
 

turboflame

Distinguished
Aug 6, 2006
1,046
0
19,290


That's the bandwidth from the GPU to it's onboard memory, not the bandwidth from the card to the rest of the system.
 

cblair

Distinguished
Dec 10, 2008
9
0
18,510


Is there a way to tell what the bandwidth of the card to the rest of the system is from the specs?
 
Not what you are asking exactly but when this type of question comes up the answer is generally that there is no single GPU card that needs more than a PCIE 1 interface can deliver.
So you wouldnt get a scenario where a single card is being held back by the slot.
I dont think anyone knows or understands the actual numbers as this type of thing gets asked from time to time and never gets answered properly.

Mactronix