Will an ATI PCI Express 2.1 graphics card work in a...

[RadCazz]

... 2.0 slot? I was at my local Best Buy and I was looking at the latest ATI graphics cards and they all said PCI Express 2.1 on their packaging. I have an available 2.0 slot. Would these cards work? I had the sales/computer associate stumped, lol.

Also, I don't think my case/motherboard has the room for these 'wide' cards. Can you recommend a premium (near top-of-line) graphics card that has a lower profile? My psu is 500W. I want to really be able to crunch projects and play the latest games.

[rgathright]

Please let us know what the motherboard model is.

[Slicker]

... 2.0 slot? I was at my local Best Buy and I was looking at the latest ATI graphics cards and they all said PCI Express 2.1 on their packaging. I have an available 2.0 slot. Would these cards work? I had the sales/computer associate stumped, lol.

Also, I don't think my case/motherboard has the room for these 'wide' cards. Can you recommend a premium (near top-of-line) graphics card that has a lower profile? My psu is 500W. I want to really be able to crunch projects and play the latest games.

I'm pretty sure the version has very little to do with anything other than the theorhetical max data that can be sent/received and no GPU comes even close to approaching the limit of any of the standards. In fact, a person (OK, a brave person) can take a PCIe x16 card and cut it down to where it fits into a PCIx x1 slot and it will still work.

You mention 'wide' and then 'low profie' which are two completely different things. Length, width, and height. Low profile = height and often needed in rack mount servers or thin desktops. Most mini-towers can use normal height cards. The drive cage sometimes gets in the way of long cards. The position of the PCIx slot and the other components near it often determine whether a double width card (most high end GPUs take up two slots). Maybe the easiest thing would be to pop the cover and measure the space you have to work with. Hopefully you don't need low profile since there aren't a lot of options for low profile GPUs that are high end.

[Fire$torm]

The 2.1 in PCI Express refers to the PCI-SIG consortium revision of the PCI Express Specification Standards. As part of that, all PCIe 2.1 devices are backwards compatible, which means 2.1 works in 2.0 which works in 1.0. BTW, PCI Express 3.0 has been finalized and PCI Express 4.0 is now in development.

Wikipedia: PCI Express (http://en.wikipedia.org/wiki/PCI_Express)

[RadCazz]

rgathright, the Motherboard is an Intel G41, Intel Core 2 Quad Processor Q8200 2.33GHz 1333MHz 4MB LGA775, in a medium tower. Hope that helps.

Thanks Slicker and Fire$torm for clearing things up for me! I was also worried the double-width cards wouldn't fit. Like you said Slicker, I'm going to have to measure the space I have to work with. It does look pretty cramped in there. I think the double-width card is going to cover/obscure/block my other available slot. Time to say, "No duh," lol.

I apologize for my ignorance. These highfalutin graphic cards are new to me.

BTW - I was looking at the SAPPHIRE AMD Radeon HD 6870 1GB GDDR5 PCIE Graphics Card (http://www.amazon.com/SAPPHIRE-Radeon-6870-GDDR5-Graphics/dp/B005C8RTTU/ref=sr_1_3?s=pc&ie=UTF8&qid=1339246214&sr=1-3) It's double-width.

Would this card be an excellent choice for crunching projects?

[Fire$torm]

One project the HD 6870 cannot run is WilkyWay#Home. This is because MW@H requires a GPU with Double Precision (DP) capability. AMD only included that in the 69xx series cards. I cannot remember if there are any other projects that need DP.

[Bryan]

The 6870 will get you about 415k/day on Donate ... I'm running one on a PCIe 1.0 machine. As F$ stated it won't run on MW but it will handle anything else.

[rgathright]

rgathright, the Motherboard is an Intel G41, Intel Core 2 Quad Processor Q8200 2.33GHz 1333MHz 4MB LGA775, in a medium tower. Hope that helps.


Sorry, no easy answer from me!

http://www.intel.com/content/www/us/en/chipsets/mainstream-chipsets/g41-express-chipset.html

According the Intel website, the Intel G41 chipset only officially supports PCI Express 1.1. :(

At your own risk, you could buy a PCI Express 2.1 card and plug it in, but I don't think anyone here would tell you it will work with 100% certainty. X_X

[Fire$torm]

At your own risk, you could buy a PCI Express 2.1 card and plug it in, but I don't think anyone here would tell you it will work with 100% certainty. X_X

Actually i would because I'm going it myself. The system is a Dell GX280 (aka Drone) with a single x16 PCI Express 1.0a slot that currently has a Nvidia 8800GT installed. I've also had 2 different ATI Radeon 4850s and a 4870. I've not had any problems with any of those cards.

Drone ---> http://valid.canardpc.com/show_oc.php?id=2400421

GX280 specs (Here (http://support.dell.com/support/edocs/systems/opgx280/en/ug/specs02.htm)) Seventh table down titled Expansion Bus

[Bryan]

In my 2 PCIe 1.0 machines I've run 48xx, 5870, and 5970. Currently running a 6870 in one machine and a GTX 570 in the other. They work fine ... all PCIe versions are backwards compatible.

[rgathright]

Actually i would because I'm going it myself. The system is a Dell GX280 (aka Drone) with a single x16 PCI Express 1.0a slot that currently has a Nvidia 8800GT installed. I've also had 2 different ATI Radeon 4850s and a 4870. I've not had any problems with any of those cards.

Drone ---> http://valid.canardpc.com/show_oc.php?id=2400421

GX280 specs (Here (http://support.dell.com/support/edocs/systems/opgx280/en/ug/specs02.htm)) Seventh table down titled Expansion Bus

Thanks, just glad we were able to answer all of RadCazz's questions. :cool:

[Fire$torm]

Thanks, just glad we were able to answer all of RadCazz's questions. :cool:

Any time. Everything we share in the forum will only help the team. We can make the team better, faster, stronger......... :cool:

[Slicker]

Any time. Everything we share in the forum will only help the team. We can make the team better, faster, stronger......... :cool:


Ching ching ching ching....

[RadCazz]

Thanks for all your replies! :) You've all cleared up my confusion! I'm headed on over to NewEgg to have a looksee at the cards, but Best Buy is having a sale on the PNY NVIDIA GeForce GTX 560 Ti. I think it's a decent price/performance card. If I won the Mega Millions jackpot I'd really have no worries, lol. Then I'd get one of those cool Alienware mobo/pc's with a GTX 690. One can only dream. :p

Update 6/13/2012:
I purchased the NVIDIA GTX 560 Ti. It does take up two slots - one slot isn't being used. Luckily, my psu has two 6-pin connectors because this card uses both of them. Frame rates are very high. I can run games in ultra and high settings with no lag. I'm gonna install BOINC on this computer and see how crunching projects goes. One thing left to do - overclocking the gpu.

And thanks again everyone for your help/input! I enjoyed reading your replies.

[Aux10]

Basically in theory any PCIE card will work in any PCIE slot of the proper size but as stated there is no way of knowing if the MOBO will take it until you try. The only thing known for sure is that you won't be able to get the benefits of the faster buss speeds made available by the newer generations of the technology.

[Slicker]

AMD's docs have the potential throughput calculations for their GPUs. None even come close to the PCIE x16 limits, so I really doubt you would ever reach the bus max speed. That having been said, you can still think of it like a Porsche 911 vs a school bus. Both can do 60 mph easily, but the Porsche can get up to speed much faster than the school bus.

Also, the top speed is in perfect lab conditions. In reality, the throughput is considerable less. That's been true with Ethernet cards for years. Most 100 Mbit cards can only maintain about 40Mbits of constant data. With really good cards and well optimized drivers (e.g. the type that go into servers which are over $100 each) the best ones can get up to 90-95% of the rated thoughput. Most Gigabit ethernet cards/devices peak out at around 300-400 MBits.

My guess is that the same is true for the PCI Express busses. A faster bus @ 50% throughput will outperform vs a slower bus at 100% because it isn't nearing its theorhetical max. That, and the newer ones are often made with smaller transistors which means the data can travel at a faster rate. The biggest bottleneck is probably the speed of the memory on the GPU itself. Even with DDR5 RAM, moving 300MB of data to and from the device takes a little while.

[kmanley57]

A good analogy - Your funnel is only so big. Your GPU is the bucket of water, and the PCI-e bus is the funnel. It(the funnel) can only do as much as it can no matter how big a bucket you use. So PCI-e 1.0 -> PCI-e 4.0 buss, use a PCI-e 2.1 GPU in a PCI-e 1.0 buss it does PCI-e 1.0 speed, on a PCI-e 2.0 it does PCI-e 2.0 speed, and in a PCI-e 3.0 it does PCI-e 2.1 speed.