PCI Express 3.0 Has Zero Performance Incentive for Radeon HD 7970

No surprise ...

Over the last few months, motherboard manufacturers have been raising a big hoopla over how it's important to pick their products that feature PCI Express 3.0 (Gen 3.0) slots. There was even some drama between competing motherboard manufacturers over who was first to the market with this technology, even when consumers couldn't really make use of the technology. To begin with, you needed a next-generation Ivy Bridge CPU, then you needed a compliant graphics card. Sandy Bridge-E, fortunately, formally introduced the technology, complete with motherboards and processors that support it.

GPU maker AMD wanted to be the first to be out there with a GPU that's compliant with this interface, and so one thing led to another, and VR-Zone got to set up a test-bed using Core i7 "Sandy Bridge-E", ASUS Rampage IV Extreme (which allows users to change PCI-Express standard mode in the BIOS setup program, by forcing Gen 2 or Gen 1 mode), and an HD 7970, to see if running the GPU on PCIe 2.0 and PCIe 3.0 modes made any worthwhile difference. The results are in: zero, nada, zilch, sunna (zero in my language).


9

United States I.M.O.G. says:

Anandtech reported a 9% increase I heard, so initial analysis is conflicting. May depend on other factors like mobo/bios, or someone made a mistake in testing.

Belgium Massman says:

[QUOTE="Borandi @ Facebook"]PCIe 2 to 3 changes the encoding scheme and CPU/PCIe bandwidth - therefore it will only help in CPU/PCIe bandwidth limited situations (as the latency is still the same). No games are CPU/PCIe bandwidth limited (all textures are usually preloaded), so the only things that do improve are in GPU Compute (science simulations) where it is needed. And that's exactly where certain websites found PCIe 3.0 did improve performance over PCIe 2.0 - in the pure compute.[/QUOTE]

Canada Eldonko says:

I thought the first round of SB-E CPUs were not gen 3 compatible?

EDIT - I guess they are but not officially certified by Intel yet?

ps - more tests here also showing no improvement - http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/49646-amd-radeon-hd-7970-3gb-review-21.html

Netherlands wesjuhdabomb says:

Might show some improvement on a Ivy bridge platform with SLI/CFX (2x 8 lanes)

United States sin0822 says:

the VRZone tests didn't show 0% improvement, just very tiny amount. But yea who expected the opposite of this anyways? The GPu doesn't perform that amazing, why would it use that much bandwidth?

Belgium HybridChiller says:

funny, so much commotion about it, and then........

No improvement

Canada Vinster says:

Chiller said: funny, so much commotion about it, and then........

No improvement


Thinking the same thing...

Vin

United Kingdom borandi says:

I.M.O.G. said: Anandtech reported a 9% increase I heard, so initial analysis is conflicting. May depend on other factors like mobo/bios, or someone made a mistake in testing.


Yup, but only in computationally heavy benchmarks where the CPU/PCIe bandwidth is the limiting factor. Ryan at AnandTech knows his stuff about GPUs, and having been given only 6 days to build a new test bed and rerun all the cards he could, he deserves a holiday tbh.

However, limiting is the key word here. The only thing that PCIe 3 changes is bandwidth and encoding. How much change do we see for games using a single GPU going from PCIe 2 x8 to PCIe 2 x16? Maybe a few percentage points (in games)? That was a double bandwidth jump. So what happens if you double the bandwidth again? Nothing. Hence the only result to increase was the bandwidth heavy compute benchmark.

Analogy:
You have 5 cars all wanting to occupy the same space on a motorway.
Let's start with PCIe 2.0 x8, which is a 3 lane motorway. You put 3 in a line, then 2.
Now move to PCIe 2.0 x16, which is a 6 lane motorway. You put 5 in a line, with a space.
Now move to PCIe 3.0 x16, which is a 12 lane motorway. You put 5 in a line, no difference.
Only in compute heavy CPU/PCIe transfers do you have more 'cars' wanting the same space.

People who were expecting an increase in single card FPS don't really understand what PCIe 3.0 is or what it is about. We may see some increase in multi-GPU (3 or 4 cards), where we're at x16/x8/x8/x8 PCIe 3.0, but even PCIe 3.0 x8 is faster than PCIe 2.0 x16 (bandwidth change + encoding scheme changes as well for less overhead per transfer).

Canada Vinster says:

^That makes a lot of sense. thanks for that explanation. Vin

Please log in or register to comment.