PCIE v1.0 mobo bottlenecking question.
Moderator: Moderators
- bobthedinosaur
- Blood & Steel Developer
- Posts: 2702
- Joined: 25 Aug 2004, 13:31
PCIE v1.0 mobo bottlenecking question.
I have a mobo with a PCI express v1.0 x16 slot, which needs a GPU (since the last one died). Would getting a GPU with GDDR5 be pointless? I can't seem to find the relation between the x16 speed pcie slots, v1.0 pcie motherboards, and gddr speeds.
Re: PCIE v1.0 mobo bottlenecking question.
If I recall PCI-E 2.0 has twice the bandwidth per lane, so what you have is equivalent to PCI-E 2.0 x8. Quite a few motherboards do not ship with all PCI-E 2.0 x16 nowadays. For example, without an addon chip the intel socket 1156 boards can only power 16 pci-e 2.0 lanes total. That means if you want to sli or crossfire the pci-e slots will both be 8 lanes.
They really don't see all 16x slots being mandatory because the performance impact is minimal. see this link.
I would say go ahead and get a modern card, you wont have a really big difference unless you try to run a behemoth such as the GTX 480 or 5970hd.
They really don't see all 16x slots being mandatory because the performance impact is minimal. see this link.
I would say go ahead and get a modern card, you wont have a really big difference unless you try to run a behemoth such as the GTX 480 or 5970hd.
- bobthedinosaur
- Blood & Steel Developer
- Posts: 2702
- Joined: 25 Aug 2004, 13:31
Re: PCIE v1.0 mobo bottlenecking question.
hmm
From what I've read my pci-e, even though it is a v1.0, is a x16 speed which I think has a transfer rate of 8 GBS and the DDR3 has a maximum transfer rate of 6.4 GBS so using DDR5 will be bottlenecked on a x16 since they say it can get up to 20GBS to 28 GBS. So if I use a DDR3 I won't be using the full potential of a x16 PCIE speed, but GDDR5 is over kill and wont be able to get close to it's max.
This is all based off of wikipedia info, and my poor understanding of technology mixed and assumptions.
From what I've read my pci-e, even though it is a v1.0, is a x16 speed which I think has a transfer rate of 8 GBS and the DDR3 has a maximum transfer rate of 6.4 GBS so using DDR5 will be bottlenecked on a x16 since they say it can get up to 20GBS to 28 GBS. So if I use a DDR3 I won't be using the full potential of a x16 PCIE speed, but GDDR5 is over kill and wont be able to get close to it's max.
This is all based off of wikipedia info, and my poor understanding of technology mixed and assumptions.
Re: PCIE v1.0 mobo bottlenecking question.
Like I said, don't worry about it. I am running my GTX460 on a PCI-E 1.0 x16 slot. My performance is right where reviews and benchmarks say the card should be. The smaller the card, the smaller the performance hit.
- bobthedinosaur
- Blood & Steel Developer
- Posts: 2702
- Joined: 25 Aug 2004, 13:31
Re: PCIE v1.0 mobo bottlenecking question.
I got a ddr3 graphics card, but mainly because of $. If I could afford it I'd probably just build a new rig.
Re: PCIE v1.0 mobo bottlenecking question.
What gfx card in particular did you get? just curious
.
- bobthedinosaur
- Blood & Steel Developer
- Posts: 2702
- Joined: 25 Aug 2004, 13:31
Re: PCIE v1.0 mobo bottlenecking question.
Sapphire Radeon HD 4670. It is just to hold down the fort, since the rig is almost 5 years old.
Re: PCIE v1.0 mobo bottlenecking question.
gddr speed will affect transfers within the card - remember that's the memory it uses for all general operations, not just transfers over pci-e
you can run a high-end card decently on pci-e 1.0 8x
you can run a high-end card decently on pci-e 1.0 8x
- CarRepairer
- Cursed Zero-K Developer
- Posts: 3359
- Joined: 07 Nov 2007, 21:48
Re: PCIE v1.0 mobo bottlenecking question.
I believe my mobo has only pcie 1.0 and my vid card is a gt240 with DDR5 (nice small low-power card for my sff rig). It works very well.
Re: PCIE v1.0 mobo bottlenecking question.
My confident answer from experience and way too much research into the subject:
A faster graphics card will always be faster than a slower graphics card, regardless of your motherboard interface. Not everything's about I/O, it's not like CPU bottlenecking where the game simulation just simply can't be computed any faster. It's more like the performance decrease you might get from increasing your anti-aliasing setting; the performance reduction is nonlinear and doesn't cause an actual cap in real-world performance or cause stuttering, artifacts, instability, or anything overly negative like that.
That said, starting with the Geforce 8xxx and Radeon 4xxx series, a performance penalty began to show when you were using PCI-E v1 instead of v2. However, the difference still remains less than 10% in virtually all scenarios, even with modern Geforce 4xx and Radeon 5xxx series cards.
My advice is that it would be unwise to purchase a replacement motherboard until mid-2011 if at all possible due to major technology improvements that are just around the corner. The effect these improvements will have is not yet fully known, but they will be quite significant and likely incompatible with existing motherboards.
Examples: USB 3, SATA 3, new-architecture CPUs from AMD with corresponding CPUs from Intel around the same time, faster DDR 3, PCI-E v3, and various accompanying architecture and chipset changes. All of these will be common starting in early 2011.
Finally, there will also be a major architecture improvement for GPUs from ATI and NV in the first half of 2011, probably in late Q1. This is not the incremental GPU improvement ATI has planned for later this year either, but a whole new architecture. Whether you want it or not, it will mean lower prices for the things you might want today.
I'm running a 5850 in a standard-speed PCI-E x16 slot, it can run virtually anything I throw at it... within reason.
Short version: Better off to stick with what you've got, save up and buy a whole new long-term system in mid-2011 when major new technologies will have come on the market.
A faster graphics card will always be faster than a slower graphics card, regardless of your motherboard interface. Not everything's about I/O, it's not like CPU bottlenecking where the game simulation just simply can't be computed any faster. It's more like the performance decrease you might get from increasing your anti-aliasing setting; the performance reduction is nonlinear and doesn't cause an actual cap in real-world performance or cause stuttering, artifacts, instability, or anything overly negative like that.
That said, starting with the Geforce 8xxx and Radeon 4xxx series, a performance penalty began to show when you were using PCI-E v1 instead of v2. However, the difference still remains less than 10% in virtually all scenarios, even with modern Geforce 4xx and Radeon 5xxx series cards.
My advice is that it would be unwise to purchase a replacement motherboard until mid-2011 if at all possible due to major technology improvements that are just around the corner. The effect these improvements will have is not yet fully known, but they will be quite significant and likely incompatible with existing motherboards.
Examples: USB 3, SATA 3, new-architecture CPUs from AMD with corresponding CPUs from Intel around the same time, faster DDR 3, PCI-E v3, and various accompanying architecture and chipset changes. All of these will be common starting in early 2011.
Finally, there will also be a major architecture improvement for GPUs from ATI and NV in the first half of 2011, probably in late Q1. This is not the incremental GPU improvement ATI has planned for later this year either, but a whole new architecture. Whether you want it or not, it will mean lower prices for the things you might want today.
I'm running a 5850 in a standard-speed PCI-E x16 slot, it can run virtually anything I throw at it... within reason.
Short version: Better off to stick with what you've got, save up and buy a whole new long-term system in mid-2011 when major new technologies will have come on the market.
