MSI afterburner stats and OR in extended mode

[Desert63]

Thank Dark, this was exactly what I was looking for. A diagnostic view of the problem and that maxxmem tool certainly showed it. Yes BIOS settings first.
Although I am sure you are right and the tool was build smart enough to not care about the amount of ram I would love if some other people here gave the tool a whirl especially if they have single stick 8 gig configurations.

EDIT - Also, I am a tiny bit concerned the tool doesn't identify my RAM....not the biggest issue, but certainly when you don't know the cause of a problem, every symptom matters.


Oh, are you saying when you run MSI with your dual 970s you show 100% GPU utilization? That is what I would expect if you are working to take advantage of every bit of power you have invested in.

Thanks

Wondering if I am just fighting a benchmark that doesnt actually represent real world performance. For instance when I go to the Maxxmen comparison site:
http://www.maxxpi.net/pages/result-browser/top15---memory.php
All the top performers across each DDR class are triple or dual channel. My work laptop and your machine are dual channel. This benchmark may be simply showing a top result that doesnt represent the real world or a real problem. Why do I think this? Because plenty of people over in the ED hardware thread are single channel and that doesnt seem to represent a real world problem.
It would be a bit odd if it was, but I suppose its possible. ED (and most games) dont use more than a few GIGS of memory...so if this benchmark represents reality that would mean i would need to add another 8 GIGS (that would not be used), just to gain the benefit of dual stick arbitration for dual channel. So then this benchmark is happy, but does that make ED run better? I am leaning away from this being the problem. It would help if anyone else that has a single stick of RAMM and is generally satisfied with thier performance would test and upload their results.
Here is the site to download - its a tiny download and about a 60 second test:
http://www.maxxpi.net/pages/downloads/maxxmemsup2---preview.php

Oh, the forums dont want me dumping in URLs, got it.
See if I can upload itView attachment 57052
1.2 megs - tiny!

Thanks!

- - - Updated - - -

Thank Dark, this was exactly what I was looking for. A diagnostic view of the problem and that maxxmem tool certainly showed it. Yes BIOS settings first.
Although I am sure you are right and the tool was build smart enough to not care about the amount of ram I would love if some other people here gave the tool a whirl especially if they have single stick 8 gig configurations.

EDIT - Also, I am a tiny bit concerned the tool doesn't identify my RAM....not the biggest issue, but certainly when you don't know the cause of a problem, every symptom matters.


Oh, are you saying when you run MSI with your dual 970s you show 100% GPU utilization? That is what I would expect if you are working to take advantage of every bit of power you have invested in.

Thanks

Wondering if I am just fighting a benchmark that doesnt actually represent real world performance. For instance when I go to the Maxxmen comparison site:
http://www.maxxpi.net/pages/result-browser/top15---memory.php
All the top performers across each DDR class are triple or dual channel. My work laptop and your machine are dual channel. This benchmark may be simply showing a top result that doesnt represent the real world or a real problem. Why do I think this? Because plenty of people over in the ED hardware thread are single channel and that doesnt seem to represent a real world problem.
It would be a bit odd if it was, but I suppose its possible. ED (and most games) dont use more than a few GIGS of memory...so if this benchmark represents reality that would mean i would need to add another 8 GIGS (that would not be used), just to gain the benefit of dual stick arbitration for dual channel. So then this benchmark is happy, but does that make ED run better? I am leaning away from this being the problem. It would help if anyone else that has a single stick of RAMM and is generally satisfied with thier performance would test and upload their results.
Here is the site to download - its a tiny download and about a 60 second test:
http://www.maxxpi.net/pages/downloads/maxxmemsup2---preview.php

Oh, the forums dont want me dumping in URLs, got it.
MAXXMEM2
See if I can upload itView attachment 57052
1.2 megs - tiny!

Thanks!

 

[Darkmark07]

are you saying when you run MSI with your dual 970s you show 100% GPU utilization? That is what I would expect if you are working to take advantage of every bit of power you have invested in.

i can hit 100% on both GPUs if I really push the settings, although if truth be told, I struggle to see a massive difference beyond 1.5 SS with most things at high or ultra which actually is about the sweet spot for my machine. Typically I try to keep < 90% in the hangar which gives me a bit of headroom in combat zones and Res sites.

If you mobo supports dual channel and you've only got one stick in there, that might be the next thing to try.

 

[Darkmark07]

Because plenty of people over in the ED hardware thread are single channel and that doesnt seem to represent a real world problem.

How many of those people running single channel RAM are also running VR or are trying to output at an effective 150 FPS?

 

[Desert63]

How many of those people running single channel RAM are also running VR or are trying to output at an effective 150 FPS?

Very good point.

Still I am not convinced that this benchmark is representing the gaming reality. For instance, what is the known required memory bandwidth to get those 150 FPS as 1920? Its a math calculation that somebody has somewhere. "Trying" an additional 8 is $75 or so, for RAMM I will never use. I would rather spend that $75 towards a new MB, or upgrading my video. Yes, I am trying to solve a problem here, but per my thoughts above I am concerned this is a Red Herring - or whatever that thing is.
Appreciated the time, and am not minimizing your help here, as we may be on to something. But the TOP 15 on that site is all dual channel. So top throughput for the sake of throughput may not represent that ED developers spec'ed for their needs in this game.

Thanks

- - - Updated - - -

i can hit 100% on both GPUs if I really push the settings, although if truth be told, I struggle to see a massive difference beyond 1.5 SS with most things at high or ultra which actually is about the sweet spot for my machine. Typically I try to keep < 90% in the hangar which gives me a bit of headroom in combat zones and Res sites.

If you mobo supports dual channel and you've only got one stick in there, that might be the next thing to try.

Thanks for the confirmation. I would expect that no matter what the video configuration if you are pushing the limits, the GPU should max.

 

[Darkmark07]

We're not talking about needing to add capacity here, we're talking about needing to add bandwidth.

The proposal to add a second stick was to increase the throughput (which is what maxxmem measures) and, in your case, shows up as a concern for the CPU and GPU that you have - hence the system experiencing poor performance even though CPU and GPU are not fully utilised.

If having a spare 8GB RAM 'doing nothing' (other than possibly doubling your throughput) concerns you, then perhaps you could swap your 1 x 8GB for 2 x 4GB - that should have a similar effect. Article that explains dual, triple and quad channel memory architectures and why they exist here: http://www.hardwaresecrets.com/ever...triple-and-quad-channel-memory-architectures/

Assuming that the memory settings in the BIOS have been checked and optimised and performance is still not up to snuff, this looks like the easiest and cheapest way to address this in my opinion highly likely reason that your machine is not performing as it should.

I can understand not wanting to throw money at a problem. I'm just talking you through the same thought processes and diagnostics that I used to diagnose an almost identical problem on my old machine. Unfortunately, I was saddled with using double-buffered, ECC DDR2 on my old rig (triple-whammy for performance, albeit it that it was triple-channel) and had no choice but to upgrade. It's possible that there was another hidden problem that went away when it was swapped out, but I couldn't find anything else and I didn't have any further upgrade options available to me.

Other than the proposals that I have already made I have nothing more useful to add except 'good luck'

 

[Desert63]

We're not talking about needing to add capacity here, we're talking about needing to add bandwidth.

The proposal to add a second stick was to increase the throughput (which is what maxxmem measures) and, in your case, shows up as a concern for the CPU and GPU that you have - hence the system experiencing poor performance even though CPU and GPU are not fully utilised.

If having a spare 8GB RAM 'doing nothing' (other than possibly doubling your throughput) concerns you, then perhaps you could swap your 1 x 8GB for 2 x 4GB - that should have a similar effect. Article that explains dual, triple and quad channel memory architectures and why they exist here: http://www.hardwaresecrets.com/ever...triple-and-quad-channel-memory-architectures/

Assuming that the memory settings in the BIOS have been checked and optimised and performance is still not up to snuff, this looks like the easiest and cheapest way to address this in my opinion highly likely reason that your machine is not performing as it should.

I can understand not wanting to throw money at a problem. I'm just talking you through the same thought processes and diagnostics that I used to diagnose an almost identical problem on my old machine. Unfortunately, I was saddled with using double-buffered, ECC DDR2 on my old rig (triple-whammy for performance, albeit it that it was triple-channel) and had no choice but to upgrade. It's possible that there was another hidden problem that went away when it was swapped out, but I couldn't find anything else and I didn't have any further upgrade options available to me.

Other than the proposals that I have already made I have nothing more useful to add except 'good luck'

Hello again , No argument with you at all on what bandwidth means here and what I get by going dual channel. So agreement in what all this means from a pure technology standpoint. Where I have a question is if I am now investing in making my machine good on a benchmark, or good in ED. The two are not necessarily the same thing. For instance, I have a computer store five minutes from me and I could go match and double my ram to put me at 16 GIGS, dual channel. So then I rock the benchmark. That might do nothing for me if the max throughput that can be handled by ED is 5GIGs a second. I don't know that is the limit and I don't know there is any limit in ED, but often there is.
So, I have a performance problem in ED, but there could be a different cause or multiple causes. I really like measuring hardware to know what I get and what it means and what I am paying for, but in the past I have chased benchmarks to see no actual improvement in a target application.
Again - anyone know what the memory bandwidth requirement is for ED? No limits, or point of diminishing return when you hit 8GIG a second, etc...?

Thanks again, I am going to do some more testing with ALL my configurations while traveling around the spacestations in ED. So far I have just been showing it off to family members in the training missions. Although I have spend a lot of time in BIOS and upgraded my BIOS, I have not spent enough time configuring ED and testing, so I am going to do a bit more of that before I decide where to go next.