171 replies to this topic

[xWiredx]

ninjitsu, on 27 September 2014 - 12:54 PM, said:

Ping doesn't have much to do with the computer.

30fps is pretty bad.

Amen. I don't like having less than my refresh rate.

[Goose]

I want to keep above 45, but never quite pull it off …

[Goose]

Lordred, on 26 September 2014 - 05:13 PM, said:

I was using HWiNFO64 and GenericLogViewer prior to this. I switched to using Afterburners log and compiling the data on my own.

Is there an accuracy difference between Afterburner/RTSS and HWiNFO?

There used to be something about FRAPS was more accurate then RTSS, but now that HWiNFO reads RTSS, I don't feel the need for FRAPS anymore.

I do think It HWiNFO reads more stuff:


Actually, I've figured out within the last month that GLV will trim the head and tail off of a log:

Edited by Goose, 27 September 2014 - 02:47 PM.

[Lordred]

Data for Caustic added.

Spoiler

[Dark DeLaurel]

Is that using Afterburner? If so what were you using to graph it with, I am still have a hard time finding a way to make the results look pretty lol.

[Flapdrol]

Nah, afterburner looks like this:


works too

[Lordred]

I am event logging with MSI Afterburner, importing the log to open office, and graphing it.

[Lordred]

Odd, I have switched to 1CpM for aditional testing now, and I am running into an odd bug, the game is capping at 50fps when I have the maxfps command set to 70, when I disable the command the game shoots up into the 130-180's.

I am just going to run the repair tool and wipe everything and test again in a little.

[Wyzak]

I have an anecdotal experience to share. I would like to preface with my system specs:

Gigabyte 990FX-UD3P
AMD FX8150 (Turbo 4.2, cruise 3.6GHz)
12GB RAM @ 1333MHz (1600 rated but this causes mobo to hang)
running off an SSD array with peak read speeds of 700MB/s
liquid cooled CPU
Gigabyte Radeon 7970HD 3GB 1GHz Windforce
Windows 7 Professional 64

So obviously this stuff was pretty good when I bought it, a little longer in the tooth now but still way ahead of most computers running MWO - right?

But the game is slowly becoming unplayable. Switching to DX11 is a totally unacceptable experience. DX9 works, but I have to disable some of the best features like texturing to get acceptable performance. Recently the video card has been spooling up and causing controls to go to mush at innopportune times (i.e., when you are under fire and most need full control.)

This is 1776x1000, motion blur off, postAA off, damage glow on.

The brute force of this processor is excessive by any reasonable scale. It's too late to divert from CryEngine, but maybe there could be features added in-game to turn off some eyecandy or some more streamlining. It's really hard to have such high end hardware and see it suffer on such a simple graphics experience. Other games have much more intricate physics experience and the computer doesn't bat an eyelash. I know there are hidden math costs with the projectile density in a multiplayer experience and the multiple limbs tracking damage. But other games can do this without the chips melting down. My experience is only anecdotal; I don't graph how bad certain maps get. But there is a problem. Maybe my parts just don't mesh well.

Edited by Wyzak, 28 September 2014 - 06:51 PM.

[Summon3r]

Wyzak, on 28 September 2014 - 06:41 PM, said:

I have an anecdotal experience to share. I would like to preface with my system specs:

Gigabyte 990FX-UD3P
AMD FX8150 (Turbo 4.2, cruise 3.6GHz)
12GB RAM @ 1333MHz (1600 rated but this causes mobo to hang)
running off an SSD array with peak read speeds of 700MB/s
liquid cooled CPU
Gigabyte Radeon 7970HD 3GB 1GHz Windforce
Windows 7 Professional 64

So obviously this stuff was pretty good when I bought it, a little longer in the tooth now but still way ahead of most computers running MWO - right?

But the game is slowly becoming unplayable. Switching to DX11 is a totally unacceptable experience. DX9 works, but I have to disable some of the best features like texturing to get acceptable performance. Recently the video card has been spooling up and causing controls to go to mush at innopportune times (i.e., when you are under fire and most need full control.)

This is 1776x1000, motion blur off, postAA off, damage glow on.

The brute force of this processor is excessive by any reasonable scale. It's too late to divert from CryEngine, but maybe there could be features added in-game to turn off some eyecandy or some more streamlining. It's really hard to have such high end hardware and see it suffer on such a simple graphics experience. Other games have much more intricate physics experience and the computer doesn't bat an eyelash. I know there are hidden math costs with the projectile density in a multiplayer experience and the multiple limbs tracking damage. But other games can do this without the chips melting down. My experience is only anecdotal; I don't graph how bad certain maps get. But there is a problem. Maybe my parts just don't mesh well.

U have a liquid cooled 8150 oc that thing into high 4's and will huge improvements.

Edited by Summon3r, 28 September 2014 - 07:12 PM.

[Lordred]

Ok, so here is the first 1CpM run (One Core Per Module)

Tourmaline Desert, Surprisingly the Avg FPS were nearly the same

OP Updated with new data

54.75fps Avg for Tourmaline on 8/8 Cores
54.72fps Avg for Tourmaline on One core per Module

Spoiler

Edited by Lordred, 28 September 2014 - 07:30 PM.

[Golrar]

I too have noticed in my testing that isolating the cores or shutting off the four even numbered cores doesn't make much difference. I just use process lasso to eek out a few more FPS.

I too have noticed that with each patch my average seems to be going down a bit. My dips have gotten lower than they were a month ago. Never down to unplayable levels, but getting close.

The new map, though a lot of people are complaining about their performance, doesn't seem to be a factor.

I tried graphing results for a few maps, but I just don't have that kind of time. Thanks Lordred for taking the time.

[Alreech]

Wyzak, on 28 September 2014 - 06:41 PM, said:

The brute force of this processor is excessive by any reasonable scale. It's too late to divert from CryEngine, but maybe there could be features added in-game to turn off some eyecandy or some more streamlining. It's really hard to have such high end hardware and see it suffer on such a simple graphics experience. Other games have much more intricate physics experience and the computer doesn't bat an eyelash. I know there are hidden math costs with the projectile density in a multiplayer experience and the multiple limbs tracking damage. But other games can do this without the chips melting down.

It's not the CryEngine.
When Crysis 3 was released a German PC Magazine did a CPU benchmark, and the FX CPUs performed well, especially when rendering open areas with lots of grass.

[Goose]

Alreech, on 30 September 2014 - 11:01 AM, said:

It's not the CryEngine.

When Crysis 3 was released a German PC Magazine did a CPU benchmark, and the FX CPUs performed well, especially when rendering open areas with lots of grass.

Emphasis Mine

[awdwikisi]

ty for the work lordred, those fancy graphs
I patiently await 4k results, could you test 105fov then at 110fov?

[Lordred]

Fightin the 3rd, on 02 October 2014 - 07:47 PM, said:

ty for the work lordred, those fancy graphs
I patiently await 4k results, could you test 105fov then at 110fov?

Still testing @ 80fov, however, here is the first of the 4k res runs, same test methods, just 4k res.

Caustic Valley @ 4k Res


OP Updated

Edited by Lordred, 03 October 2014 - 10:22 AM.

[Alreech]

thanks for the work, Lordred (what should have been done by PGI...)

So overclocking the hell out of a FX 8xxx (and it's northbridge) would avoid low fps.
Sadly it's only an option if you have a good mainboard and a good cooling for the CPU and the case.

How do you made the 1 core, 1 module test ?
Disabling all cores with an even number using a 3rd party tool or assigning cryengine threads to specific cores with the user.cfg ?
Do you plan to use the user.cfg to force MWO to use the almost idle cores 7&8 ?


-----------------
* my board only supports Phenom IIs up to 125 Watts, so then next upgrade will be a Intel i5 or i7

[Flapdrol]

I think he used a feature in his bios to get one active core per module.

Should reduce powerconsumption at high loads, so you can overclock further and still do a proper stability test without breaking the cpu or the mobo.

[Remarius]

Yeah the Fatal1ty BIOS allows that setting.

[Thorqemada]

Gigabyte Bios too allows to set 1 Core per Module on the Gigabyte GA990XA rev1.x so imho that should be a pretty common Bios Option (if you have an updated Bios that supports the FX CPU).

I would wonder if that is not the case in any established brand Mobo.