Review: MSI GeForce GTX 1070 Gaming X SLI

#5293841 Posted on: 06/23/2016 10:01 AM

---

Great article - as always!
Though I have a feeling something went wrong with "Total War: WARHAMMER" test (or posted result).
According to article - in FullHD (1920x1080) SLI works way slower than a single card. Is this correct?
76 FPS (SLI) vs 90 FPS (Single)

If this indeed have been the actual test result - a word or 2 on that on the page itself would be helpful.

#5293842 Posted on: 06/23/2016 10:07 AM

---

@HH "GeForce GTX 1070 - 8 GB GDDR5X memory" or GDDR5 only

#5293844 Posted on: 06/23/2016 10:15 AM

---

Great article - as always!
Though I have a feeling something went wrong with "Total War: WARHAMMER" test (or posted result).
According to article - in FullHD (1920x1080) SLI works way slower than a single card. Is this correct?
76 FPS (SLI) vs 90 FPS (Single)

If this indeed have been the actual test result - a word or 2 on that on the page itself would be helpful.

That behavior is actually described on the conclusion page. Often with multi-GPU solution you can stumble into a bottleneck once you are not GPU bound anymore. The framerates go up higher and higher whereas the graphics driver needs to do it's work twice as hard and thus enforces a much heavier workload on the processor. If that processor cannot keep up you will see negative scaling in lower resolutions, the problem is CPU related and thus a CPU bottleneck.

#5293845 Posted on: 06/23/2016 10:18 AM

---

No time I'm afraid, in case you missed it .. it's the busiest summer ever with al GFX releases, in order to timely release other articles there is just no time left.

I hear ya though - perhaps in a later stage I'll look into a massive mGPU article with different brands and cards all together. No promises though.
Thank You Hilbert. I know it is busy time.
I have another unreasonable request to ask:
"When you are benchmarking RX-480 for FCAT, can you capture GPU utilization and clock over time too?"

Even if only for one game, so we can see if performance we see is all RX-480 can deliver.
Because at time Fiji released, I measured so many games where GPU utilization was damn low and fps was not great. But with newer drivers, GPU utilization and fps went up.
AMD did not recover all there is to Fiji. GPU is still underutilized in many cases. But at least we can get to know if there is untapped potential.

Even something as simple as I did here:
WH 40k: Space Marine:Maximum details + Adaptive AA + texture LOD settings "-1" enforced via RadeonPRO.
1080p: 120fps (limit set in RP)
1440p: 108~120fps (limit set in RP)
2160p: 96~120fps (limit set in RP)
--------
W10 Driver 15.200.1023.10 Fiji
1080p: 128~205fps
1440p: 130~202fps
2160p: 95~192fps
Improvement on lower resolution is notable. At 2160p card is quite utilized, so no improvement there:

It is 3times same section of game where hordes of smaller orks are sieging camp on ground, and then at end there is 1st ground boss fight.
And, you are right below. I should have put it into PM.

#5293846 Posted on: 06/23/2016 10:21 AM

---

Thank You Hilbert. I know it is busy time.
I have another unreasonable request to ask:
"When you are benchmarking RX-480 for FCAT, can you capture GPU utilization and clock over time too?"

Even if only for one game, so we can see if performance we see is all RX-480 can deliver.
Because at time Fiji released, I measured so many games where GPU utilization was damn low and fps was not great. But with newer drivers, GPU utilization and fps went up.
AMD did not recover all there is to Fiji. GPU is still underutilized in many cases. But at least we can get to know if there is untapped potential.

This thread is not at all about RX 480 so please stay on-topic. I'll indulge one reply though, AMD decided not to include a DVI connector (needed) on that card, FCAT is impossible until an AIB partner supplies one with a DVI connector.