55 replies to this topic

[auniqueid]

Quote

My results:
Min, Max, Avg 52, 99, 77.427

5820K @ 3.8GhZ
GTX 660 Ti 2GB
Canyon Network
1920x1080

Nice.

[auniqueid]

OK-- I just did a pug, Alpine peaks, 12 on 12. (yes, I was the jerk with the DW and 4x gauss cannons)

All settings on high.one monitor. vsync off. I touched the overclock on my processor back to 4.35 ghz. 980s on stock.


Here's the AB graph from start of match to finish:

Edited by Jimbobbob, 29 September 2014 - 07:39 PM.

[auniqueid]

And here's an HPG manifold PUG-- again, 12x12, same settings as previous... cards at stock, processor at 4.35 ghz, synch off, all settings at high.

Game starts at the red arrow on the left; killed at the red arrow at the right (early death, I know). The sharp drop in FPS happened when I started spectating.

Edited by Jimbobbob, 29 September 2014 - 07:24 PM.

[auniqueid]

last one for the night.. same settings as before, Tourmaline desert.
FPS dipped down to 51 at one point... this was a very close quartered fight, I actually lasted awhile this time

for whatever reason window indexing kicked in somewhere in the vally where the "51" is...

Edited by Jimbobbob, 29 September 2014 - 07:41 PM.

[Flapdrol]

If you look at the lines you see it went below 40 for a few moments, still, not bad for everything on high.

[auniqueid]

Quote

If you look at the lines you see it went below 40 for a few moments, still, not bad for everything on high.

How do these results compare to other systems? I would assume with a clock at 4.35 ghz, and the other hardware, this should be at or near the top . . .

[Catamount]

Jimbobbob, on 30 September 2014 - 06:11 AM, said:

How do these results compare to other systems? I would assume with a clock at 4.35 ghz, and the other hardware, this should be at or near the top . . .

For comparison, you're probably doing about 50% better than my 3570k at 4.2, which I was doing some FRAPS runs on last night (hitting minimums around 30-40, but usually closer to 40 with averages about 2/3 yours). Your system's clockspeed is equal to about 4.6 on mine since Haswell has something like a 5% IPC advantage over Ivy, so the rest ought to be coming from more cores, maybe caches (but I'm thinking cores, since MWO likes to have lots of FPUs to work with).

Edited by Catamount, 30 September 2014 - 08:01 AM.

[ninjitsu]

I average 58 fps with settings maxed out. The lowest I've dipped while monitoring fps is 37. I'm running a 4670k and a 4 year old video card.

[auniqueid]

Quote

For comparison, you're probably doing about 50% better than my 3570k at 4.2, which I was doing some FRAPS runs on last night (hitting minimums around 30-40, but usually closer to 40 with averages about 2/3 yours). Your system's clockspeed is equal to about 4.6 on mine since Haswell has something like a 5% IPC advantage over Ivy, so the rest ought to be coming from more cores, maybe caches (but I'm thinking cores, since MWO likes to have lots of FPUs to work with).

very interesting. I may run some rigorous tests later with background services turned off. I did notice on tourlamine desert the first two cores started to ramp up noticeably more, right around the time windows indexing decided to kick in.

[auniqueid]

and for anyone else that's interested... AB graph for forrest colony, 12x12, this was pretty close quartred fighting .
ran through the graph manually and marked the lowest fps identified by AB, lowest dip was 52fps. highest (labeled) was 146fps.
All settings on high, overclocked to 4.35 ghz, 980s at stock speed.

Edited by Jimbobbob, 30 September 2014 - 11:21 AM.

[Reno Blade]

I tested my 3rd monitor today in nvidia surround now running 5760x1080 or something.
Decent 35 fps on ultra high so far.
Not sure if I will keep playing that way
i5 3570 @ 3.4GHz, 16GB ram, GTX 770 OC 2GB, win7 64bit.

[auniqueid]

Quote

and for anyone else that's interested... AB graph for forrest colony, 12x12, this was pretty close quartred fighting .
ran through the graph manually and marked the lowest fps identified by AB, lowest dip was 52fps. highest (labeled) was 146fps.
All settings on high, overclocked to 4.35 ghz, 980s at stock speed.

Anyway, probably not going to waste time with more fb graphs, if you guys want o see more let me know.

One interesting thing Idid notice is that the bus utilization on the first 980 is hovering around 65-70%. I'm running both card with 16 3.0 PCI lanes... when I get my third card, because of the way asus lays this board out (and because I'm using an M.2 card that shares pci 3.0 lanes), I'm either going to have to go with 16/16/4 or 8/8/8.... wondering if the cards will saturate the bus with only 8 lanes each.

[Catamount]

Well you certainly don't want 16/16/4 whatever you do, because that last card will saturate the x4 bus and the most likely result will be microstutter (one of things that tends to happen when cards start loading or running unevenly). Running 8/8/8, yes, it's possible that there will be times when the bus will get saturated, barely, but it's going to be absolutely insignificant to your performance. People have been testing the impact on GPUs of having a saturated PCIE bus for years with little observed effect. If you savagely overloaded it, sure, you'd probably have a nasty performance hit, but barely doing so, maybe, occasionally, won't matter.

[auniqueid]

Quote

Well you certainly don't want 16/16/4 whatever you do, because that last card will saturate the x4 bus and the most likely result will be microstutter (one of things that tends to happen when cards start loading or running unevenly). Running 8/8/8, yes, it's possible that there will be times when the bus will get saturated, barely, but it's going to be absolutely insignificant to your performance. People have been testing the impact on GPUs of having a saturated PCIE bus for years with little observed effect. If you savagely overloaded it, sure, you'd probably have a nasty performance hit, but barely doing so, maybe, occasionally, won't matter.

Yeah, I think I'll stick with 8/8/8. Was thinking of dumping the m.2 drive so I can go with 16/16/8, but the thing is so blazingly fast I'd hate to lose it.

On a different note.. ran in surround mode, at 5760x12000 resolution.
FPS dropped, as expected... I lived within 40-50 fps the entire time. But whats interesting is that this was far more consistent than any of the other runs ... this is fps on forrest coloney, 12x12, in surround, three monitors, highest settings, vsynch off -- the lowest the FPS dropped was 40.7, the highest was 61.2 -- this is interesting because on surround, I'm pushing roughly the same number of pixels (6,912,000 pixels) as a 4k monitor (8,847,360 pixels), and it definitely pushes these cards (I'm assuming that means no dual 4k monitor setup until I trade up for quad-sli?)




very interesting metrics here-- BOTH cards were hovering between 98-100% usage -- on previous tests on just one monitor, gpu1 was in the 70-80%, gpu2 was practically idle:





I've done a few tests on caustic valley, manifold and tourlamine desert, and am seeing the same graph --- no more than 60, no less than 50 fps on surround 3x monitors.

Edited by Jimbobbob, 30 September 2014 - 02:34 PM.

[xJohnWolf]

Jim, have you had any issues with draw distance? Like mechs or terrain disappearing at distance?

My setup:
Asus R4E mobo
Intel 3960x
2x Asus 780 Ti
32 Gb Ram @ 1600MHz
5760x1080 (3x1 surround) + 1 1920x1080 accessory

When I go single screen I can get a max of 260 fps with max settings and awesome graphics quality. Min around 200 with averages around 220. There is no tearing or any glitches at all. But when I try to do surround (triple screen) with DX11 I have great frame rates still 100+ but I get tearing and draw distance issues. I've tried VSync and a number of other fixes, but the only thing that seems to work is to swap back to DX9. But in DX9 I get poor graphics quality and frame rates only around 50-90 averaging around 60. Which is adequate, but I feel like I should be getting more from twin 780 Ti...

[Missing Spartan]

Jim you shouldn't have any problem running your cards @ 8/8/8 on PCI 3 lanes. http://www.techpower..._Scaling/7.html The test was with BF4 which is probably way more demanding then this game.

Edited by Missing Spartan, 10 October 2015 - 10:23 AM.