Games run fine.. then lag

Robert0211

Commendable
Dec 12, 2016
154
0
1,710
Intel S500PSL
2x Intel Xeon X5450 3.0 GHZ 8-core total
24 GB FB-DIMM DDR2
2x 1 TB HDD
EVGA GTX 1050ti SC
Win 10 pro

SO, every game requiring 4+ core to run, work fine for like... 15-20 min, but then start lagging, sound glitch, visual lag, every 5 sec or so, why? Rocket League and Elite Dangerous do that, while World of Tanks HD client doesnt? (wot is a 2 core only}
I havnt ruled out overheating, which shoulsnt happen, but oh well, cpu temp max at 65-68 and gpu temp never above 50, all temps in Celsius
 
Solution
Your case seems fine and you have done a good job. It would help if you were able to mount a fan on top of the case that would suck some air above the RAM or on the side panel, that would blow some air to the CPUs and RAM. But since your case doesn't allow you to do one of the above, you could do one last thing as a test. Disable the top intake fan that is next to the CPUs. That fan is creating a big amount of airflow above the CPUs but you aren't using tower CPU coolers. Since they are down-draft coolers, that fan may disrupt the amount of air that they can use, thus limiting their cooling ability. Disable that fan and close the case. Now monitor the temps. Maybe that little trick helps. I can't think of anything else at the moment...
Only CPU overheating could explain that behaviour. The max temp you reported is the max core temp or CPU package temp? On this site => http://ark.intel.com/products/34446/Intel-Xeon-Processor-X5450-12M-Cache-3_00-GHz-1333-MHz-FSB, you'll find the specs of your CPUs and it says the the max temp (Tcase) is 63°C. Based on that it seems that your 2 CPUs (or at least one of them) do overheat and you you should do some dust cleaning on your coolers and reapply some new thermal paste.
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710


On CPU world it says 70 C, and I literally built the system over Christmas break, cheap thermal paste or misappied it?
And my hard were monitor doesn't have a package temp I dont think... So its core temp
 
If it's core temp it's OK. Core temp is higher than the actual CPU package temp. Since you have 2 CPUs you should be seeing 2 different temps. Are you seeing the same temp from both? I am asking because you mentioned only one temp. The max temp I mentioned is according to Intel and usually xeon CPUs have lower temp limits vs consumer/mainstream CPUs. These CPUs are designed for increased stability and reliability and not for overclocking, so they have a lower max temp ceiling.

The thing is that this system is very old and its motherboard may have stability issues, especially its VRM circuit. The CPUs may not overheat, but if the motherboard and its VRMs are overheating they may lower the current applied to the 2 CPUs, thus dropping their clock speeds. Also the temp sensors may not report the correct temp or may be damaged. You have to monitor the clock speeds of your 2 CPUs and make sure that they don't throttle down thus producing the lag you are experiencing.

Finally you have to take into account that you have 2 CPUs and you have to monitor 2 totally different temps and clock speeds. If one of your CPUs is overheating or dropping its clock speed and the other doesn't then you''ll see lag on your screen. You have to make sure that both are working normally.
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710


The temps I mentioned are all average 8 cores ( 2 cpu) hover around 63-67 C Im using CPUID HW monitor, and it shows the temp of each core individually. So Im seeing 8 temps, and when lagging starts, the temps are about 63-67 degrees
 
Do the core temps have large variations between each other? Maybe one of your CPUs operates at a higher temp that the other. Also what about the clock speed of each core? Does it suddenly drop when the lagging starts? If it only happens when the temp reaches at that levels then it's logical to assume that the temp is responsible for this behaviour. You should also monitor the motherboard temp. Usually motherboards have 2 or more temperature sensors on various motherboard sections. Maybe when the cores reach that specific temp they cause some parts of your board to overheat.

As a test you can run your system with all the fans (especially the CPU fans) at their max RPM. That way you may lower the max temp and observe whether the lag happens again and when it does see if it is time or temperature related.
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710


Im testing again to see clock speed, and on this HD monitor mb temp is not present, and I have no fan control, CPU fans adjust automaticly to a certain degree but all case fand are 3 pin, and I have no way to control them
 
Do the core temps have large variations between each other? Maybe one of your CPUs operates at a higher temp that the other. There must be a setting in your board's BIOS to let the fans run at its max RPM. if your board can automatically adjust the CPU fan speeds then there must be a way to disable that. You should also know that as various electronic components age they become less tolerant to voltage and heat. So a CPU that is specified at a max temp number may not be able to achieve that 8 or 10 years later. The heat transfer material inside your CPU (below the heat-spreader) may also have dried after all those years and there is no way to replace it besides removing the heat-spreader but then you risk damaging the CPU die. What you can do is try to get much better CPU coolers or try to undervolt both your CPUs. That way you'll lower the max temperature of all your cores and may avoid that temp ceiling that is probably causing the lag.
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710


You can de lid them :p I did a test really quick.. and WOW the cpu overclocks to about 3.1 BUT when lagging starts.. it clocks as low as 1.4 GHZ.... insane... the lowest core temp when lagging is 56 while the highest is 63.. the other cores are somewhere in between those 2 temps

 


Hahaha good for you if you can do it. I never had the guts to de-lid a CPU but in you case it may help if you cannot find another way to drop that temp. One or both of your CPUs is definitely overheating. Can you pinpoint the exact cores that are throttling and then find out at which CPU they belong?

You have to find a way to bring that temp down. As I said you can increase the CPUs' fan speeds, reapply thermal paste, get new better CPU coolers, undervolt or underclock those CPUs or finally if all else fail, de-lid them (good luck with that ;)) and reapply new thermal paste there. Be careful though you have to be absolutely certain that the heat-spreader is removable before proceeding because some heats-spreaders are glued to the CPU die and cannot be removed without destroying it.
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710


The only option for any fans is closed circuit and open circuit.. Doesn't make much of a diffrence as this being a server I can't undervolt unless some 3-rd party application can do it.. The heatsink might not be applied properly because I'm using some nuts to hold it in place instead of paying 14$ for 2 back plates.. I'm gonna reapply thermal paste soon.. Use more expensive (insignia brand)
 
Good luck you hands seem tied since you are dealing with server parts. I also forgot about the server motherboard and related accessories I only hand my mind on the xeon CPUs (face-palm). I don't have much experience with them but I believe that you may be able use a 3-pin fan-to-molex adapter in order to make sure that the fans are spinning at their full speeds. On the other hand though I've heard some server fans spinning at full speed and they are too noisy for my taste. You should definitely look at the heat sink's placement and make sure that it sits firmly on the heat-spreader and no air-gap exists between them. Otherwise there is no way to drop the thermal load. Even the best cooler won't help you there.
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710

So other then reapplying the thermal paste and cooler, theres nothing I can do?
 


Since these are server parts and you don't have control over the clock-speed and the voltage settings, your options are limited. Try to reinstall the heatsinks and make sure that you find a way to improve the contact between the heatsink and the heat-spreader. You could also get better CPU coolers than what you currently have but you have to make sure that you able to properly install them and the heatsinks should be firmly seating above the heat spreaders without air-gaps between them. Unfortunately this is a very old server socket and it may be hard to find compatible CPU coolers which also may be too expensive. One last thing you could do as I said above is pinpoint which CPU is getting hotter and try to concentrate on its cooling solution. Maybe you are lucky and only one CPU is overheating.
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710
So I went into my Bios and changed fan control over to open loop so I have more control, and whoope doo, it was set to acoustic (quiet) so I det ir to performance (basically full throttle) so, ill test tomorrow when I have time for a longer play session ill post results
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710


would custom watercooling help at all?
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710


Changing to open loop only delayed overheat by about 10 min
 
It's clear that you current CPU cooling system isn't able to cool your CPUs or it isn't correctly mounted. You should get a new one or find a way to make the heatsink make a better contact with the heat-spreader. You could search your BIOS for power saving mechanisms and make sure that all those settings are enabled. Finally if your BIOS allows you to lower the CPU's multiplier just a little bit thus dropping the CPU's clock speed by about 100-300 MHZ, do it, it may help but I think it's only a matter of time before it overheats again. You have to find a way to further improve your current CPU cooling solution.
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710


I touched the fins of the heatsink while temps showed low 60's and they were cool.... is that a sign of bad contact?? i mean these things should be cooling them easy, the are the recommended coolers and are FULL copper, fins and contact point On any other computers (prebuilt) when i would touch the fins on the heatsink even idleing, it would be warm/hot to the touch, granted they were full aluminum but still...
 
Yes it seems that they aren't making enough contact in order to transfer the heat and a lot of air is trapped between the heatsspreader and the heatsink. That heatsink should be at least warm to the touch if the CPU is at 60 degrees C. If your board didn't throttle your CPUs in order to protect them then the temp would skyrocket to over 80 degrees C. Even the best CPU fan won't help you in that situation. On the other hand if you somehow manage to somehow make a perfect contact between them but you can't bring the max temp down then it's a clear sign of bad/aged/broken thermal transfer material below the heatspreader and between the CPU die and the heatspreader and the only way to fix that is to de-lid the CPUs but I have no clue whether you can do it.
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710


Delidding isnt an option, my die is soldered, took off heatsinks and saw WAY top much thermal paste, put a thinner layer on and secured them on better, this delayed heating to about an hour, HOWEVER if I go and set the affinity so that those games only use one cpu... no problems, why is that? does windows or the game not know how to efficiently use all 8 cores and so it overheats?
 
Windows just sees 8 cores not 2 CPUs so efficiency may be an issue in your case. Usually server boards use server operating systems and not commercial ones. On top of that you system is too old and technologically is considered "ancient". It may have 8 cores but it's on 2 CPUs and each CPU has 2 2-die cores each, which means that your system is in reality 4 2-core CPUs that communicate via a slow and technologically "ancient" bus. Also your memory bus is slow and DDR2 is too slow too serve 8 CPU cores.A modern i5 would be way faster than what you currently have and even an i3 in certain situations. So lets say that your system isn't efficient at all by today's standards. As a result, performance-wise, I don't think you would see a big performance difference if you only used 1 CPU.

On the other hand your system seems to have stability issues even though it's based on a server board. Have you run stability tests like prime 95 and see how it performs and if your entire cooling system is able to handle the excessive heat in a worst case scenario? If I were to bet I'd say it would crash. As I said, in a previous post, your hands are almost tied. That said I think you can somehow manage the situation by doing little tricks like the one you did above with the affinity thing. Also that trick may point to a cooling issue on only 1 of your CPUs. As a test do the same thing with the other CPU. If it now overheats then you have a smoking gun.
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710


It shouldnt be 2 core with 4 threads... Every program says 4 cores with 4 threads
 
Your CPUs don't support hyperthreading so the number of cores and the number of threads, is the same. So in your case you have 2 quad core CPU which means you should see 8 cores and 8 threads. If you only see 4 cores then something is going on with one of your 2 CPUs
 

Robert0211

Commendable
Dec 12, 2016
154
0
1,710


I do see 8 cores. Anyways looking at temps now, its hitting lower by 1-5 C but still laggs.... It doesn't downclock though... I honestly have no idea at this point