PCI-3.0 x16 on a 2.0 16x motherboard

[Xexion]

i'm looking to upgrade my pc, with a Gtx 980 wich uses PCI-3.0 x16, yet my motherboard supports PCI Express 2.0 x16, will this work ???
and how much can i lose on fps.

i have read that it wont make that much a difference, but those were all topics of 600 series 3.0 cards. i'm not sure if the same applies on newer cards.

Motherboard:
ASUS P8P67 DELUXE
Gtx 460 SLI

[Bad 💀 Motha]

Yes it works; No you wouldn't notice a difference between 2.xx and 3.xx

[Azza ☠]

PCIe 2.0 x16 = 8,000 MB/s = 64,000,000,000 bits
PCIe 3.0 x16 = 16,000 MB/s = 128,000,000,000 bits

GTX 980 Memory Clock = 7,000 Mb/s = 7,000,000,000 bits

That's still below PCI 2.0 Bus Width, but the GTX 980 has a 224 GB/sec Memory Bandwidth. While PCIe 3.0 would be recommended, you would find little to no performance difference on a PCIe 2.0.

With some other graphic cards, the performance drop would be low anyways, only ~1.5 FPS differences.

[FluffyPinkDecoyBunny]

Here's a review on techpowerup.com (http://www.techpowerup.com/reviews/NVIDIA/GTX_980_PCI-Express_Scaling/) about PCI-E scaling using 1.0, 2.0 and 3.0 with various different lanes. You won't feel any kind of performance hit using 2.0 16x over 3.0 16x

[Xexion]

Thanks everybody, my question is answerd.

[Azza ☠]

引用自 vadim：

引用自 Azza ☠：

GTX 980 Memory Clock = 7,000 Mb/s = 7,000,000,000 bits
That's still below PCI 2.0 Bus Width, but the GTX 980 has a 224 GB/sec Memory Bandwidth.

Very strange math.... The clock frequency is measured in megahertz, not in megabit per sec.
PCI-E 3.0 clock rate is variable from 100 MHz to 1000 MHz (with 8 bits data width per lane. Or 500 MHz if data width is 16 bits). Only UI (UI stands for "unit interval" which has nothing with voltage transition in differential pair ) corresponds to the conventional clock frequency 8 GHz (UI = 125ps for PCI-E 3.0).
Thus, in any case PCI-E clock rate is several times less than GTX 980 VRAM clock rate (3.5 GHz, because of DDR).
More than that, PCI-E has much bigger overhead than RAM bus and real bandwith equal roughly 400 MBps for PCI-E 2.0 lane in each direction and about 60% more for PCI-E 3.0 (you can notice that overhead greatly increased for PCI-E 3.0).

Brief: even for PCI-E 3.0 each lane has less bandwidth than single bit width RAM bus. But videocard supports 16 lanes at most, while top grade GPUs have 256, 384 or 512-bits VRAM bus.

http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-980/specifications

GTX 980 Memory Specs:
7.0 Gbps - Memory Clock
4 GB - Standard Memory Config
GDDR5 - Memory Interface
256-bit - Memory Interface Width
224GB/sec - Memory Bandwidth

PCIe 2.0 16x is 8.0 Gbps... what's so hard about the maths?
1 Gbps = 1,000 Mbps (actually it's 1024, but to make it simple)

PCIe 3.0 16x is 16.0 Gbps and the new NVLink (yet to be released) will be a whopping 80Gbps. It's entirely overkill now, but in the future Nvidia Pascal cards would be able to run UltraHD 4K on a single card and stream an entire bluray disk within a second (so they will be working up to that by 2016+).

[Azza ☠]

引用自 vadim：

引用自 Azza ☠：

PCIe 2.0 16x is 8.0 Gbps

Can you explain what do you mean by that, please?

http://en.wikipedia.org/wiki/PCI_Express#PCI_Express_2.0

The PCIe 2.0 standard doubles the transfer rate compared with PCIe 1.0 to 5 GT/s and the per-lane throughput rises from 250 MB/s to 500 MB/s.

So each lane has up to 500 MB/s throughput right.

500 x 16 (lanes) = 8000 MB/s = 8GB/s.

[ARmodder]

引用自 xexion：

i'm looking to upgrade my pc, with a Gtx 980 wich uses PCI-3.0 x16, yet my motherboard supports PCI Express 2.0 x16, will this work ???
and how much can i lose on fps.

i have read that it wont make that much a difference, but those were all topics of 600 series 3.0 cards. i'm not sure if the same applies on newer cards.

Motherboard:
ASUS P8P67 DELUXE
Gtx 460 SLI

If your jumping from a low end GPU all the way to the GTX980 I'd strongly advise against that as there is not that much a difference performance wise between the 970 and the 980 yet the 980 costs almost twice as much! I have heard people who have used both the 970 and 980 say that the extra $200+ you pay for the 980 is not worth it, just thought I'd let you know.

[Azza ☠]

I think your just making it more complicated than it needed to be, however the question has been answered. It will still have enough bandwidth (on a PCI 2.0 16x) to make not much of a difference.

FluffyPinkDecoyBunny had posted an excellent real-world benchmark webpage of them all. On average it would just be ~1.5FPS difference, if not that.

As for what graphic card to get... depends on your monitor / resolution.

If your going for 1080p (1920x1080), then a GTX 970 would be optimal.

If you getting 1440p or higher, else tri screen monitor setup (Nvidia Surround) or something like that, thats when the GTX 980 would come more into play, else it's quite a serious overkill!

[CmdrPwn]

引用自 ARmodder：

If your jumping from a low end GPU all the way to the GTX980 I'd strongly advise against that as there is not that much a difference performance wise between the 970 and the 980 yet the 980 costs almost twice as much! I have heard people who have used both the 970 and 980 say that the extra $200+ you pay for the 980 is not worth it, just thought I'd let you know.

Sorry for bumping an old thread. I read that the 970 actually has a serious hardware defect that causes performance to drop (up to 20%) if it uses more than 3.5GB of VRAM. This might not be a big problem today, but it very well could be in a few years. I have also read people having issues on the 970 with games like Dying Light. So in my opinion it's worth spending the extra $200.