Jump to content

Mwo Fps For My 5960X/2X Nvidia 980S


55 replies to this topic

#21 auniqueid

    Member

  • PipPipPipPipPip
  • Stone Cold
  • Stone Cold
  • 131 posts
  • LocationUSA east coast

Posted 29 September 2014 - 03:04 AM

Quote

You were in the testing grounds? ...


Everything I did is laid out in my original post.

someone responded that they were "skeptical" for some reason of my frame rate, (I still don't understand or his basis for saying so, other than maybe he's a skeptical person), so I suggested that maybe the tests I did were not active enough, and I could try for a more active pug.

#22 Flapdrol

    Member

  • PipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 1,986 posts

Posted 29 September 2014 - 03:23 AM

The reason I don't believe a minimum of 60 in a big fight is this game only uses 60% of a quadcore, which means adding cores will do almost nothing for performance, a quadcore haswell doesn't hold 60, and the haswell-e isn't faster per core, just has more cores and cache.

#23 Catamount

    Member

  • PipPipPipPipPipPipPipPipPip
  • LIEUTENANT, JUNIOR GRADE
  • 3,305 posts
  • LocationBoone, NC

Posted 29 September 2014 - 05:25 AM

Even if there was some modest advantage to more cores, and who knows maybe mwo makes some minimal use of cores 5/6, MWO is very fond of per core performance. There is no reason to expect a stock 5960x at 3ghz (maybe 3.2 max all core turbo, at most) to even perform on par with my 3570k at 4.2, which doesn't come close to absolute minimums of 60. That low per core performance would, I think, severely bottleneck the hungrier threads. Threading only goes so far if everything has to wait for the slower thread. This is the basis of Amdahl's Law and MWO displays it in spades. My guess is this is just an artifact of non-careful measuring. I bet a proper log of FPS across multiple games (FRAPS benchmark?) will show minimums nowhere near there.

Edited by Catamount, 29 September 2014 - 05:31 AM.


#24 xWiredx

    Member

  • PipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 1,805 posts

Posted 29 September 2014 - 06:09 AM

Using the exact same testing methodology, the minimum frame rates with my 5820K @ 3.6GhZ were 10x better than the minimum frame rates with my 2600K @ 4.4GhZ (literally, 30-35 min vs 2-4 min). There's an 800MhZ deficit there. I'm not going to go through rigorous testing to rule out each individual improvement between the platforms (plus at least one of them can't be tested). I can tell you that memory bandwidth doesn't do much for MWO, though, so it's down to IPC improvements (which only accounts for 10-15% improvement), cores, and cache. I can also tell you I'm GPU limited with my 660 Ti for sure. If I had a GTX 980, I'm guessing my mins would be up there in the 50-60 range. I really don't see how this is hard to understand or believe.

#25 Flapdrol

    Member

  • PipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 1,986 posts

Posted 29 September 2014 - 06:16 AM

I've seen footage of dips to 25 fps on a 4770K, maybe he should've lowered settings a bit, but still.

#26 xWiredx

    Member

  • PipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 1,805 posts

Posted 29 September 2014 - 06:22 AM

...so if you're comparing the same architecture, and the one with 2 more cores and 7.5MB more cache has better minimum fps on very high everything @1920x1080, there are only a few factors to consider: the slower one has a slower GPU, the extra cores help, the extra cache helps, or a combination of 2 or all of these. I used FRAPS to benchmark. When I turn on my desktop tonight, I'll do a couple test runs. My OC on the 5820K is now at 3.8GhZ.

#27 Catamount

    Member

  • PipPipPipPipPipPipPipPipPip
  • LIEUTENANT, JUNIOR GRADE
  • 3,305 posts
  • LocationBoone, NC

Posted 29 September 2014 - 06:43 AM

Your 2600k is bombing out there. That's hardly typical. OTOH, my 3570k matches your 5820K's minimums at 3.6. In fact, I think I might even beat that out by a bit, though I'd want to run a more recent fraps run to verify (a couple months ago my minimums were in the low 40s, haven't had reason to test since). Surely you don't think my 3570k, or any typical i5, is somehow ten times better than your 2600k. Your 5820k actually performs about where I'd expect. It's perfectly in line with the trends we've seen across hundreds of samples of performance over the last year or two, doing just about what I'd expect for MWO's current state (you'd have gotten 60 easily when 12 man's first came along, and for awhile after, probably on either chip :( ). The OP's chip remains the outlier here.

It's an outlier and nothing we know about MWO's performance gives it any apparent physical mechanism to be such, so yes it's "hard to believe".

Edited by Catamount, 29 September 2014 - 06:50 AM.


#28 Catamount

    Member

  • PipPipPipPipPipPipPipPipPip
  • LIEUTENANT, JUNIOR GRADE
  • 3,305 posts
  • LocationBoone, NC

Posted 29 September 2014 - 07:07 AM

That isn't to say the OP's performance is impossible, of course. If the OP's chip is giving some kind of miracle performance it'd be great news, since we might be able to get a better idea of what constrains MWO from that if someone can get the same setup and look closely at what's going on (Goose, get a couple thousand together and get on it :P), or help the OP do the same. That's precisely why this merits examination. Skepticism is, however, very warranted.

Edited by Catamount, 29 September 2014 - 07:08 AM.


#29 auniqueid

    Member

  • PipPipPipPipPip
  • Stone Cold
  • Stone Cold
  • 131 posts
  • LocationUSA east coast

Posted 29 September 2014 - 07:15 AM

View PostCatamount, on 29 September 2014 - 05:25 AM, said:

Even if there was some modest advantage to more cores, and who knows maybe mwo makes some minimal use of cores 5/6, MWO is very fond of per core performance. There is no reason to expect a stock 5960x at 3ghz (maybe 3.2 max all core turbo, at most) to even perform on par with my 3570k at 4.2, which doesn't come close to absolute minimums of 60. That low per core performance would, I think, severely bottleneck the hungrier threads. Threading only goes so far if everything has to wait for the slower thread. This is the basis of Amdahl's Law and MWO displays it in spades. My guess is this is just an artifact of non-careful measuring. I bet a proper log of FPS across multiple games (FRAPS benchmark?) will show minimums nowhere near there.



I'll do more rigerous tests this weekend. didn't realize this was so controversial -- I've been using a gtx 660ti for the past year, so I thought 60fps was decent, but not stellar (didn't realize that v-synch capped me at 60 mhz), or I would have paid attention more -- written down the map, the number of people playing, etc. like I've said a few times in this thread, it's entirely possible that the PUGs I played were not active enough or the maps not complicated enough to stress the haswell-e.

On an entirely different note, I mucked around with overclocking this weekend, and got stable results at 4.5 ghz on the chip, and upped the core clock on the 980s to 1,400, without any bumps in the voltage. Ran Prime 95 for an hour, and not a blip on the chip --- temps stayed under 110 F. reran firestrike (non-extreme) and got a modest bump to high 21,000, again, no major bump in temps. May try this weekend to go further with some voltage adjustment.

#30 Catamount

    Member

  • PipPipPipPipPipPipPipPipPip
  • LIEUTENANT, JUNIOR GRADE
  • 3,305 posts
  • LocationBoone, NC

Posted 29 September 2014 - 07:33 AM

Well like I said, if you really are getting such good FPS, it might give some insight into why MWO hates life in general, and how to make it happier.

Also, that's an incredible stock-voltage OC. You could probably just keep it there :) (unless you just feel like playing around, which I wouldn't blame you for :D)

#31 Summon3r

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,291 posts
  • Locationowning in sommet non meta

Posted 29 September 2014 - 07:36 AM

good god man 4.5ghz on stock voltage!?!? congrats that is awesome!!!

#32 auniqueid

    Member

  • PipPipPipPipPip
  • Stone Cold
  • Stone Cold
  • 131 posts
  • LocationUSA east coast

Posted 29 September 2014 - 08:12 AM

View PostSummon3r, on 29 September 2014 - 07:36 AM, said:

good god man 4.5ghz on stock voltage!?!? congrats that is awesome!!!


Not trying to pat my back here --- MY results are pretty typical for these chips -- people have been getting 4.3 - 4.5 ghz with no issues. The brick wall on these chips seems to be hit anywhere from 4.4 to 4.6 ghz, and it's a dramatic drop off... I've seen dozens of verified benchmarks with people up to 4.5 ghz with stock voltages an simple cooling, butvery very few at 4.6, and none at 4.7 (which doesn't mean no one has done it, just not all that common)

All I did was increased the multiplier, left bclk alone, and didn't touch voltages. Soon as I hit 4.5gz, any further increase to multiplier and BSOD as soon as I launched prime (although it did boot, which leads me to believe I haven't hit a ceiling yet).
Even using the asus mobo has an overclocking feature that will get me to 4.4 with zero issues and zero bump on the voltage.

Edited by Jimbobbob, 29 September 2014 - 08:17 AM.


#33 Catamount

    Member

  • PipPipPipPipPipPipPipPipPip
  • LIEUTENANT, JUNIOR GRADE
  • 3,305 posts
  • LocationBoone, NC

Posted 29 September 2014 - 09:31 AM

Yeah don't touch bclk; that can get dangerous on modern Intel platforms.

There are a few things you can do to try to stabilize besides increasing core voltage. If it booted, but crashed on Prime, you could try more aggressive loadline calibration for starters (regulate vdroop better). Still, easy 4.5 on that chip sounds sick. If I wasn't transitioning to notebooks soon, I'd consider getting one :/

#34 Summon3r

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,291 posts
  • Locationowning in sommet non meta

Posted 29 September 2014 - 10:51 AM

View PostJimbobbob, on 29 September 2014 - 08:12 AM, said:


Not trying to pat my back here --- MY results are pretty typical for these chips -- people have been getting 4.3 - 4.5 ghz with no issues. The brick wall on these chips seems to be hit anywhere from 4.4 to 4.6 ghz, and it's a dramatic drop off... I've seen dozens of verified benchmarks with people up to 4.5 ghz with stock voltages an simple cooling, butvery very few at 4.6, and none at 4.7 (which doesn't mean no one has done it, just not all that common)

All I did was increased the multiplier, left bclk alone, and didn't touch voltages. Soon as I hit 4.5gz, any further increase to multiplier and BSOD as soon as I launched prime (although it did boot, which leads me to believe I haven't hit a ceiling yet).
Even using the asus mobo has an overclocking feature that will get me to 4.4 with zero issues and zero bump on the voltage.


excellent news on the OC'ability of these chips ive been waiting to hear some actual results from people... if u havent even touched voltage i would think you could hit some crazy OC levels with good cooling... and as Catamount said steer clear of messing with bclk

#35 Flapdrol

    Member

  • PipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 1,986 posts

Posted 29 September 2014 - 11:22 AM

Be careful overclocking these, if you up the voltage and increase the power the cpu is allowed to use in the bios they can get really hot.

#36 auniqueid

    Member

  • PipPipPipPipPip
  • Stone Cold
  • Stone Cold
  • 131 posts
  • LocationUSA east coast

Posted 29 September 2014 - 11:32 AM

View PostFlapdrol, on 29 September 2014 - 11:22 AM, said:

Be careful overclocking these, if you up the voltage and increase the power the cpu is allowed to use in the bios they can get really hot.



I'm going to monkey around with it this weekend, but based on what I've seen, unless I have an uber-magical chip, even with boosted voltage, I'm not going to see anything more than 4.6-4.7 ghz max. And I think 4.7 is even too much to expect.... doesn't seem worth it to fight for another 100mhz, when I can get 4.5 on stock voltages and standard cooling. 4.5 is more than enough ... and frankly, far more than I expected from this chip. I may even back it off a little to give a bit of a buffer from the edge.

Now, memory is another thing... I have ddr2800, and I can't get it to budge over 2133. from everything I've read its a bios issue all x99 boards are facing, but I'd love to bring that number up a bit.

And still having issues with the new m.2 drive... I have four pcie lanes devoted to it, nothing in the "shared" pci slot, and its running along at the same speed as a high quality sata III device... should be getting 750gbs read/600ish gbps write, but I'm no where near that ...

Edited by Jimbobbob, 29 September 2014 - 11:36 AM.


#37 Catamount

    Member

  • PipPipPipPipPipPipPipPipPip
  • LIEUTENANT, JUNIOR GRADE
  • 3,305 posts
  • LocationBoone, NC

Posted 29 September 2014 - 01:54 PM

Another thing to keep in mind is that chips will wear out when OCed to their more extreme limits. I mean, they'll wear out even with modest OCs, or no OC at all, but the lifespan difference between a super-high voltage-fueled OC barely kept working by a beefy cooler and the lifespan on a good practical OC, something on stock or near-stock voltage, is very big. Real high OCs will often just nuke the stability of a chip instead of outright killing it, too.

So you might be able to get to 4.7 if you push (higher?), but in a year it might destabilize at anything over 4 (damage), whereas it could have hummed along at 4.4 or 4.5 happily for many years. Given that we're probably all going to keep our chips longer than we're used to from here on out (or maybe I should say from Sandy Bridge on out), I'd try to settle into a good practical OC that's easy and stable and leave it there, not that going all out once will harm it.

Edited by Catamount, 29 September 2014 - 01:54 PM.


#38 xWiredx

    Member

  • PipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 1,805 posts

Posted 29 September 2014 - 03:54 PM

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

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

#39 ninjitsu

    Member

  • PipPipPipPipPipPip
  • FP Veteran - Beta 2
  • FP Veteran - Beta 2
  • 402 posts

Posted 29 September 2014 - 04:34 PM

View PostxWiredx, on 29 September 2014 - 03:54 PM, said:

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

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


What game settings? Custom CFG?

#40 xWiredx

    Member

  • PipPipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 1,805 posts

Posted 29 September 2014 - 04:37 PM

View Postninjitsu, on 29 September 2014 - 04:34 PM, said:


What game settings? Custom CFG?

I never run a custom config. All settings at Very High but only using postaa.

Edited by xWiredx, 29 September 2014 - 04:37 PM.






1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users