Recent Periodic Lag Spikes

Recently I have had some major periodic lag spikes. This only occurs in BeamNG, and only recently. I have tried disabling all mods, clearing cache and putting settings all to default and lowest, turning off physics, collisions, sound etc.

This is the performance log below.

PC Specs are quite good, performance has never been a problem. No major driver changes recently, but I did install new graphics drivers as troubleshooting.

Xeon E5 1620 v3 3.5Ghz
16GB DDR4
R9 Nano
SSD
Windows 10 with Anniversary Update

BeamNG 0.6.1.0.2748 x64

 

BeamNG has had issues with the deeper pipeline that Xeons have. That's probably the cause of your problem. AFAIK, there is not solution besides getting a consumer CPU.

 

I'm not totally convinced. The 1620 v3 is not a special Xeon, and I am not using ECC memory. There shouldn't be any other architectural differences between my specific CPU and an Core i series. Besides, it was working fine last week and for the last 21 hours (many months for me)!

Is there any other diagnostic information I can look for?

 

It's the same architecture, but the data is processed a bit differently. BeamNG doesn't get along with Xeons because of that. All Xeons, every single one, not just these "special" ones you speak of, have this difference.

 

Every Xeon has a deeper memory and processes pipeline. This is where all of the issues come from. BeamNG is still unoptomized to hell, and as such, bogs down due to these pipelines. You've always had these framerate problems, you're just now noticing them.

Long story short: Buy an i5/i7 next time.

 

I am referring to E5 vs E7 Xeons, and those with with multi CPU support etc. As far as I know, the lowest end E5 (like what I have) is only slightly different from the (for example) Haswell-E i7s, with a few extra instructions supported and features turned on. I can look at the whitepapers but I really don't think 'm going to find anything interesting.

I've definitely not had this issue before. The lag spikes are pronounced and regular - just look at the CPU time graphs I posted. Looks like this forum isn't a good place for support - marked as "will not fix"? :/

 

Other thing, get HWMonitor, check your temperatures, xeons thermal throttle more aggressively than i5/i7 counterparts and the game with its high CPU load does cause many PCs to start overheating

 

FPS limiter was just there as I was playing with settings. The issue persists regardless of configurations. Attached is the log, with my normal graphical settings.

There is absolutely no throttling or down-clocking going on, and no heat issues.

I believe I found the issue:
The log files showed these repeating few lines

Code:

124.05267|D|DirectInputManager::enumerateDevices| *** Input devices start ***
124.08595|D|DirectInputDevice::create|Mouse [instance: Mouse] detected, created as mouse0 (asynchronous)
124.08596|D|DirectInputDevice::create| - devtype 0x112
124.08597|D|DirectInputDevice::create| - A mouse or mouse-like device (such as a trackball)
124.08598|D|DirectInputDevice::create| + 3 axes
124.08599|D|DirectInputDevice::create| + 8 buttons
124.08605|D|DirectInputDevice::create|created device 'mouse0', took 0.000s: create:0.014ms, caps:0.001ms, txts:0.063ms, enum:0.016ms, fmt:0.004ms, buff:0.002ms, axis:0.000ms, leds:0.000ms
124.08610|D|DirectInputDevice::create|Keyboard [instance: Keyboard] detected, created as keyboard0 (asynchronous)
124.08611|D|DirectInputDevice::create| - devtype 0x413
124.08612|D|DirectInputDevice::create| - Keyboard or keyboard-like:
124.08613|D|DirectInputDevice::create|  - IBM PC Enhanced 101/102-key or Microsoft Natural keyboard.
124.08614|D|DirectInputDevice::create| + 128 buttons
124.08647|D|DirectInputDevice::create|created device 'keyboard0', took 0.000s: create:0.014ms, caps:0.000ms, txts:0.050ms, enum:0.271ms, fmt:0.028ms, buff:0.000ms, axis:0.000ms, leds:0.000ms
124.11604|D|DirectInputManager::enumerateDevices| *** Input devices end ***
124.11705|D|DirectInputDevice::acquire|mouse0 acquired.
124.11712|D|DirectInputDevice::acquire|keyboard0 acquired.
124.14443|D|Win32Window::WindowProc|Win32Window::WindowProc - getting device arrival:
124.14449|D|Win32Window::WindowProc|Win32Window::WindowProc - getting device arrival:
125.15240|D|DirectInputDevice::unacquire|   mouse0 input device unacquired.
125.15243|D|DirectInputDevice::unacquire|   keyboard0 input device unacquired.

I found out how to show the console in game, and this was printed for each pair of lag spikes. I think it is pretty clear what is going on here! And it's definitely not related to my CPU architecture... >.<

EDIT: By disabling all HID devices, the lag spike goes away. Now to find which one is actually causing this problem. However, BeamNG is the only program I've used that seems extremely sensitive to which peripheral is plugged in, there is probably a logical fix to the problem.

EDIT 2: Attached a performance graph of what the normally performing system looks like.

 

Submitted the reports. The ID for the reports are 3282 and 3283.

More information about the HID device that causes problems:

Bluetooth Audio/Video Remote Control HID
BtaMPM.sys
BthAvrcpTg.sys

Related to a Windows 10 Mobile device, but the device is not connected or even turned on when the problem occurs.

 

Not to beat a dead horse, but would you two mind simply editing your posts with some strike-thru? You are both wrong. Pipeline depth is the same between a Haswell desktop product and Haswell server product. See these references if you remain confused:

• Announcement of OP's CPU which mentions that it's a workstation part and does not mention any changes from the desktop part which would affect pipeline depth... for any of the CPU's in the press release. http://www.cpu-world.com/news_2014/...v3_and_E5-2600_v3_CPUs.html?PROCESS=Read+more...
• Wikipedia page which includes a chart with the number of pipeline stages. It's worth noting here that the pipeline depth doesn't seem to have changed since Sandy Bridge. So that's SB->IB->Haswell->Broadwell->Skylake ... 5 or more generations all with the same pipeline depth.
• Maybe you were thinking of the original "Intel Core" on the desktop vs "Nehalem" in servers? (a) that's some really outdated stuff for us to be talking about and doesn't apply to the OP (b) Both microarchitectures were used in both desktop and server parts. So both numbers of pipeline stages were available in both servers and desktops, depending on what products were purchased.
• Here's a short article from Intel explaining how Sandy Bridge's pipeline works: Pipeline Speak: Learning More About Intel® Microarchitecture Codename Sandy Bridge. (also part 2) Note that it refers to Sandy Bridge extensively but does not differentiate between SB implemented in a Core-i vs Xeon.

There's lots of misleading hardware info laying around in these forums. There is no reason for it to persist after proven wrong.

 

Then would you please care to explain the significant performance deficits that users have had with Xeons compared to consumer processors of similar specs? While the OPs issue is not related to this, my post still remains valid, as I didn't say that Xeons' data pipeline was deeper, I said that the data was processed differently.

 

Actually.. you did. See your own post #2 in this thread. You specifically referenced the (non-existent) deeper pipeline.

I'm not familiar with the problems you refer to. Can you link me to some other thread(s) where the Xeon was determined to be the culprit and not some other thing? I'll be happy to elaborate on those cases - the OP's Xeon obviously performs fine and does not cause an issue.

 

First hand experience with a 3.5ghz haswell xeon, CPU bound 15fps 1 D15 on ECA with a quadro GPU, 90+FPS paused.
A few other users have reported xeon performance issues in BeamNG. Xeon reportedly gets issues in quite a few other games too.

 

I think somebody needs to learn a little more about computers before they try to help someone else.

 

Are you saying that you disagree with the points I made in post #13?

 

And then there are others with similarly speced Xeons that bottleneck Quadros to below full physics sim speed (as in, the physics don't refresh at the required 2000 Hz)

 

my xeon seems to run the game fine 35 frames with 6 d series on gridmap is the max i can do real time with 8 chrome tabs and a few other programs