New homebuilt machine not quite cutting the mustard.

"Not cutting the mustard" May be a bit of an overstatement, but my machine was pooping it's figurative pants last night whilst gaming.
Before I go any further let me give you my system details.


CPU: Core i5 2500k over clocked to 4.2 GHz

RAM: Corsair Vengeance CMZ8GX3M2A1600C9R 8GB (2x4GB) DDR3

GPU: HIS HD 7970

PSU: Corsair AX650 Gold

Whilst playing 4v4 SC2 last night it got to the point when it was running at a consistent 8 fps.
This is when a *** ton of units were on the map. Now when the map was crowded with units it was fine (probably 30 fps) but when this special unit called a "mothership" go in on the action is where my pc pooped the bed. Its special ability is to cloak units within a fairly big area around it, when it passed over my forces and my allies forces (probably about 500 individual units) cloaking and uncloaking everything as it moved. Everything started to move like a 1980's security camera.

This was with the resolution at 1920x1080 and the settings ranging from high to low. The only setting I have on highest is texture, but things like reflections and individual character animation (bodies stylishly falling to pieces as energy blades pass though them for example) are turned off.

The gpu was running at about 60c and the CPU around the same at this point.

I just built this machine and I was excited with my false hope that it would handle epic rts moments like these not with ease but competently.

So my question is. What sort of machine would be able to keep this game running above 30fps at a moment like that? With everything set to ultra?

As a side note. This set up is making easy work of every other game I've put it on. So it actually is cutting the mustard...Except for this one time.
15 answers Last reply
More about homebuilt machine cutting mustard
  1. Any game where the computer controls tons of units will be slow no matter what the CPU is. Even on my X6 when I play Civ V on anything above a Tiny or Mid-Size map it struggles.
  2. run 3dmark11
    check your OC stability by using prime95 on small ffts for 12 hours. no worker's should stop.
    run prime95 small ffts along with OCCT GPU test simultaneously to check your temps (OCCT has a temperature graph)

    Your mobo is 4+2 power phases, not the sturdiest around the market. A lot of people with corsair psu are having problems with them lately. Do the tests, compare to other scores online and we see what happens. What is your vcore?
  3. Supporting more than 2 threads will be the only way to save sc2 from its poor performance.
  4. This could also be an issue with the 7970 and early drivers. The 2500K @ 4.2 should eat anything that's thrown at it alive. Even in games like Skyrim which is notoriously cpu hungry it performs side by side with the HT enabled 2600k.

    I'd look around see if anyone else is having 7970+SC2 issues, that's my first thought.
  5. vx53c said:
    run 3dmark11
    check your OC stability by using prime95 on small ffts for 12 hours. no worker's should stop.
    run prime95 small ffts along with OCCT GPU test simultaneously to check your temps (OCCT has a temperature graph)

    Your mobo is 4+2 power phases, not the sturdiest around the market. A lot of people with corsair psu are having problems with them lately. Do the tests, compare to other scores online and we see what happens. What is your vcore?

    This might illicit a sigh, but can you explain what ffts is? I googled it to no avail.
    I did run intel burn test for 5 hours, all cores stayed below 70c except for one that reached 73. This was with high ambient room temp mind you.
    Also prime95 the cores didn't get as hot, all stayed below 70c.

    I ran 3dmark. My overall score is p6659...Lower than what it should be apparently.
    It says my driver is not approved.

    Although the physics score was much higher than what the average user gets: (user score: P7874- target for my hardware: P6700)

    Like a4mula said though. It could be an early driver thing.

    It's a shame about the psu.
    I always have trouble with them. Even though I select well reviewed brands. So I went all out with this one. It cost a bit but every single review was very good, and even though it's only 650w that's still about 100watts more than my system uses at load.
  6. no sighs here. when you first fire up prime95 it gives you 4 options of testing. the first one says small ffts, the third is the default. choose first one and go.
  7. P95 gives you three options for stress testing:
    Small fft's - really stresses the CPU.
    Large fft's.
    Blend. These latter two stress the memory more than the CPU.

    Highly complex math used in signal analysis.
  8. I actually used intel burn test for 4 hours when I first overclocked it.
    No stability problems or heat problems.
    Did the same with prime 95 when I went to work today.
    Just got home and after 9 hours of prime 95. Still running well, got to 74c on one core and it was really hot today (summer time in Australia) 35c according to the forecast, but it felt way hotter.

    So I'm pretty happy with that.
    Ran occt, again no problems the gpu got to 68c

    Also vx53c, my vcore is set to auto. As that is apparently what's required with this mobo for a 4.2 ghz oc.

    I've been testing games with fraps, and they're running as they should be except for that one instance with the uncloaking and decloaking of units.
    The only concern I have now is why am I getting a lower than expected 3dmark score.
    I'm looking at reviews for the 7970 and the overall performance score they're getting is around the mid 7 thounsands while mine is the mid 6 thousands, yet for some reason my physics score is blitzing what the average user gets.

    I suppose I shouldn't complain considering all the games I play are working as desired.

    My mate is buiding the exact same set up as me tomorrow, except he has an antec high current gamer psu and a HAF X case (shouldn't really make a differnce). So I might be able to narrow it down somewhat after we run 3d mark on his machine.
  9. auto vcore is never a good thing. If it goes above 1.35 even for a little while then it is bad news.
    keep monitoring and tell us what vcore it uses. (on occt you can see the diagram of rails and voltages)
  10. vx53c said:
    If it goes above 1.35 even for a little while then it is bad news.

    Why is that...the i5-2500K can take up to 1.52V by spec (not flaming, just curious).

    +1 to the driver issue...the 7970 is a new card, I would imagine that is the issue. Do you have another GPU you could try (borrow a friend's?) to see if the issue replicates?
  11. anything above 1.35 will gradually degrade your cpu. also i can sleep better at night knowing my mobo won't decide to spike the processor with 1.6vcore because my bubbles screensaver along with an antivirus update demand suddenly too much of my 2500k
  12. Speedfan and coretemp are reporting constant vcore of 1.05 on idle and a constant 1.36 at load cpu-z is broken and reporting back 1.5v on both idle and load . However when I did load it the max temps spiked at 82c briefly.
    Compared to the 73c when I first overclocked it a few weeks ago, and that was a hot day. This is a cool night!
    What's worse is that i can't even access the bios now for some reason.
    My keyboard is a logitech g110. The led lights for the keys light up as soon as the ASUS logo appears on boot.
    I've accessed the BIOS by pressing delete plenty of times before and now it's not working, I've also tried two other keyboards in different usb slots to no avail.

    The computer is posing no other problems yet though, but I would really like to get into the bios and change the vcore.
    Possibly even put everything back to factory settings since the cpu is all of a sudden getting way too hot. Even though the temps and voltages are acceptable when playing games 1.33 seems to be the max voltages and 63c seems to be the max temps.

    The 82c has me very worried though.
  13. it is as i feared then. poor power phases and vcore spikes.
    Try a ps2 keyboard if you have one and bash that del key in.
    alternatively, if you don't own a ps2 keyboard then try a different usb port, try getting some usb ports to a card bracket from the mobo headers, clean the CMOS from the jumper or the battery, RMA the board and get a real overclocker.
    All good options! Don't despair, but still, don't run the cpu OC on auto ever again.
    I know! It takes a LOT of time, so many hours of testing on prime95 to find the perfect voltage and then prime+occt to make sure you temps agree with you... meh. That's life.
    If it were me, i would clear CMOS, make sure the board is good to go and try to jump to a 8+2 or better phases from Asus/Asrock/Gigabyte.
    You could also get to boot bios or clear CMOS and leave it running on auto, or give it a lil bit of manual OC after endless trial and error (mostly error) and be happy with it. I recall trying 2 days straight to get a stable OC on a cheap 4+1 LE mobo from asus and finally giving up because i was climbing the vcore mountain a step at a time for a meager 400ghz increase. Yeah, bugger OC on 4 cpu phase mobos no matter what guru's say about gigabytes legendary 4phase vrms OCing a 2500k to 5ghz. Pfffft.
  14. Thanks for the help man.
    I Got an old keyboard and managed to get into the bios.
    Changed the vcore to manual and set it to 1.35votls.
    Unfortunately, The monitoing programs are still reporting back 1.36 volts whilst using intel burn test, and core three still spikes at 82c.

    Even though my temps don't go above 60 whilst playing games and the voltages don't go above 1.33, I'll just get a new board.
    What would you recommend?

    Edit: I discovered that my motherbaord doesn't allow you to set the voltages traditionally, but rather has an offset mode.
    So after setting the offset mode to a negative value, in order for the cpu to not go over 1.35 volts, I ran intel burntest again.
    I got much lower temps... a brief spike of 73c was the highest I saw on one core.
    However the voltages are reporting back the same. Even though I put the offset mode in negative and set the offset to 0.040 which should in theory bring the 1.36volts down to 1.32 maximum, I'm still getting the 1.36v.
  15. mm probably you are trying to set the turbo voltage offset. 1.35vcore is much for 4.2ghz, no matter how bad the cooling might be. You should be looking at something like 1.25~1.30vcore for 4.2ghz.
    I like my asrock z68 e3 gen3. asus p6z68-v pro / gen3 is good too with more flavor and more quality but it will cost you. You will need to play with asrock board's settings a bit coz it will throttle you with it's default intel speedstep/ thermal throttling values stopping you at 4.2ghz no matter what, but after that you will find it to be the bang for buck it is. If you got a CM 212 you will be looking at good temperature/vcore (about 1.25->1.27 @ 65C) ratio at about 4.2ghz, if you got a better cooling solution aim for ~60C and around 1.27 with room abient of 20C. Personally i don't like higher than very early 60's on 20C keeping in mind that summer will give you 40C on a very hot day making it 80's for the cpu. Hotter days might even make your OC unstable but that's what you get. Good luck!
Ask a new question

Read More

Homebuilt Systems