• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

2500K slams into wall

D

debaser*

n00b
Joined
Jan 1, 2012
Messages
9
Config:

i5 2500K
8gb G.Skill DDR3 1600
Radeon 5770 @ 925 / 1400
ASRock Extreme3 Gen3 Z68
Crucial m4 64gb SSD
Antec 300 case
OCZ SilentXStream 600W PSU
Cooler Master Hyper 212 Evo

BIOS Settings Altered:

Turbo Boost Power Limit: 300 (short) 300 (long) 1 (whatever third option is)
Core Current Limit: 150
LLC: 5 (Stock but notable)
Spread Spectrum: Disabled
PLL Overvolatge: Enabled

My system runs perfectly @ 44 multi / 1.33 (1.216 load) vCore (fixed) / 65C Load (hottest core IBT max) / ran 10 IBT passes / 9hr p95 blend

To get it stable at 45 multi, I have to crank the vCore to 1.41 (1.264 load). Here it maxes out at 78C, which is as far as I felt comfortable pushing it before asking here.

I will probably end up running at 44 multi, but I wanna test it out some more.

What are the max safe temps you would recommend for IBT stress tests? Under normal use, the hottest it gets is about 8-10C less than IBT load temps.

Do you see any errors in my BIOS settings? I may go back later to mess with offset voltage, but I don't see it being necessary for this.

Or, I could ignore IBT altogether and try to find stability through p95. I feel like this stresses the system as much as I can through normal use. I'm sure I could lower the voltage a ton to achieve p95 stability.
 
i was kind of disappointed after i clicked on this thread.

i guess i shouldn't take things so literally.
 
I'd say 78C is on the high side of acceptable - it is plenty safe, but most people normally like to keep it in the mid-70s under stress testing. As for hitting the wall, have you tried enabling Internal CPU PLL Overvoltage? That will often buy you several more multipliers at the same or similar Vcore.
 
Mine isnt happy on my Z68A-GD65 G3, past 44x either. Though I dont break 70, (hit 66-68), but BSOD on 45 or higher with 1.4Vcore. Im going to play with lowering my Vcore to 1.35-1.38 for 44 I guess.
 
SonDa5 said:
Wall may be memory related.
Click to expand...

I have run at 1333 CL9, and CL10. I have removed 8GB of ram, and even upped the Volts to 1.6V.

I was hoping for 4.8, and an IBT temp of 80 (since gaming, I never get close to that, so high temps in IBT etc is fine IMO)

Right now im encoding a blu ray rip for my son. Im at 100% CPU load (not a hard load I dont think), @ 4.4ghz, 1.38V, and im hitting 50C, with my AP181s on low, and DH14 in auto
 
Try doing a manual voltage increase for the PLL from 1.8 to 1.85/1.9 and increase the VTT by a small portion to see if that helps stable it out. 1.42v is normal for a 4.8 and 1.375v should do 4.5 as long as the vcore does not suffer from vdroop. Vdroop will kill your stability.
 
I switched from IBT to LinX.

Ran 5 passes of LinX max memory at 47 multi & 1.425 vCore hottest core 74C load

I will definitely will play around with my PLL and VTT settings to see if I can lower the voltage on it.

The vdroop causes it to drop to 1.288 at it's lowest. IBT was causing massive vdroop.

xSD6a.jpg


Will stress testing with just p95 and LinX be good enough?

p95 blend no likey. I suppose I can push it a smidge harder.
 
Last edited:
Linx and IBT are pretty good. If it passes that, it'll probably be fine. Play some games, see how it is - you aren't guiding the space shuttle with it, so it's not like it has to be 100% stable under any circumstances.
 
Mine would pass Linx but not pass Prime 95.. Just depends on your idea of stable. 24 hours of prime is my stable. I like to be able to know I can install windows or do whatever on my OC and know it's not going to mess up. My 2500k wouldn't do 44x well but would do 48x just fine. Just some chips don't run some mulitpliers the same. I would try moving a little up and down and see what happens.
 
I kept pushing to stabilize the 47 multi, and I did. Unfortunately, that was at 1.455 vcore in bios, 1.312 max vdroop which went to 75C.

So I guess for now I will stick with 46 multi. 1.41 bios / 1.280 max vdroop / 70C. Ran an hour of p95 blend and 5 LinX passes so far.

Maybe I will get bored one day and lap my HSF or something, but I guess this is where I'm gonna stop tweaking for now.

Now that this issue is close to being resolved, I wonder what your guys thoughts are on overclocking memory. I'm considering pushing either timings or freq at stock voltage once this is stable.
 
Last edited:
Overclocking memory isn't worth it on Sandy Bridge, as RAM faster than 1600 doesn't provide any real world benefits. That Vdroop seems excessive - have you tried a higher value for Load Line Calibration? Normally Vdroop is on the order of 0.05 or so.
 
Would lower timings offer any benefit?

The vdroop is pretty damn massive, I suppose LLC is the next area for me to tweak once I settle in on my 46 multi settings.
 
Don't run memory at over 1.5V and over 1333MHz on a 2500K. It's not supported by the Intel spec and can void your warranty (as explained by an Intel rep).

I blew the integrated memory controller of my last 2500K that way. It might be a rare situation but it can happen. Don't bother with memory at all. Continuing tweaking your CPU if you can.

So far I've had three 2500Ks. Two Malay ones from July and September and a Costa Rican one from last week. I'll have a fourth to test with soon. All three could do 5.0GHz prime stable at 1.4V to 1.425V so it may be something else in your motherboard configuration that is creating that wall for you.

I have LLC @ Ultra-High (ASUS forum rep suggested Extreme) and CPU Current Capability to 130% (ASUS forum rep suggested 140%). I have Internal PLL Overvoltage disabled. I'm using offset minus vCore configuration rather than set a static manual vCore (but that will work also) because I want to keep Speedstep/EIST for when I am idle. Why the massive vDroop? Setting a stable vCore is probably more condusive to system stability.

I would look to upgrade your cooling solution if you want to push to 4.8GHz to 5.1GHz. Prime small FTT will get the CPU 10° C hotter than Blend test if you want to try it out.
 
Last edited:
I cant understand why people want to run their Sandy's at 5.5ghz on Air @ 1 volt.

Crank that puppy up to LLC Very high/Extreme / 1.48vcore / 50x and watch her rip @ 90c. And she'll be fine. Oh and get a Corsair H80/100 and scrap the fan if you really want to overclock. Or boost your case cooling up significantly.

Get RAM that is rated for XMP @ 1600mhz / 1.5volt, there is only like 2 out there anyways, its dirt cheap. Honestly I would rather you run 1333mhz RAM anyways with tighter timings. As the throughput on socket 1155 memory controller is absolutely astonishing. It devastates even 1366 in a lot of test, not joking (well with 1600mhz).

Spare-Flair said:
Don't run memory at over 1.5V and over 1333MHz on a 2500K. It's not supported by the Intel spec and can void your warranty (as explained by an Intel rep).
Click to expand...

Whilst' he is right, of course you can run 1600 with XMP just do NOT overvolt any ram on this platform past 1.5volts. That will certainly stress the memcon. I think Spare-Flair's issue was due to the overvolting not the ram speed.
 
tangoseal said:
I cant understand why people want to run their Sandy's at 5.5ghz on Air @ 1 volt.

Crank that puppy up to LLC Very high/Extreme / 1.48vcore / 50x and watch her rip @ 90c. And she'll be fine. Oh and get a Corsair H80/100 and scrap the fan if you really want to overclock. Or boost your case cooling up significantly.

Get RAM that is rated for XMP @ 1600mhz / 1.5volt, there is only like 2 out there anyways, its dirt cheap. Honestly I would rather you run 1333mhz RAM anyways with tighter timings. As the throughput on socket 1155 memory controller is absolutely astonishing. It devastates even 1366 in a lot of test, not joking (well with 1600mhz).
Click to expand...

If you want your CPU and overclock to survive for a year or two, 90°C is pushing it a bit too far.

2500K TCASE is 72.6°C which iirc is the highest safe temperature the temp sensor diode embedded under the heatspreader should reach for long-term CPU health and stability. I'd keep it under 80° C personally.
 
Spare-Flair said:
If you want your CPU and overclock to survive for a year or two, 90°C is pushing it a bit too far.

2500K TCASE is 72.6°C which is the highest safe temperature the temp sensor diode embedded under the heatspreader should reach for long-term CPU health and stability.
Click to expand...


The whole point I was making was that the chip is capable. The OP hit a wall because he/she is not willing, or it might appear so, or is not well experienced enough to run the chip with a little more juice. 90c is very hot and yes the chip will not last long at all. The whole point was that it can be done all day long with ease on almost every single K series Sandy. I also made the point that he/she, pending desire to go higher in clocks, should invest in far better cooling as Air will work but no where near what an active liquid type system would do.

For instance, OP, I am running a 6 core 130watt TDP I-7 3930K @ 4.7ghz.1.38vcore all day long stable and tested rock solid. I am using a full water cooling loop and @ this frequency fully loaded this chip is easily pulling 350-400 watts alone. My temp@ idle is 27c avg and fully loaded using Prime95 Large FFT is 55-60C. Better cooling goes a LOOONG way.
 
Spare-Flair said:
If you want your CPU and overclock to survive for a year or two, 90°C is pushing it a bit too far.

2500K TCASE is 72.6°C which iirc is the highest safe temperature the temp sensor diode embedded under the heatspreader should reach for long-term CPU health and stability. I'd keep it under 80° C personally.
Click to expand...

Tcase is a temperature that is only measured by placing a thermocouple on the top center of the heatspreader, and is used by heatsink manufactuers. The temperatures read-out by something like CoreTemp are the temperatures from the digital thermal sensor, and they begin to limit the CPU at Tjunction, which for Sandy Bridge is something like 100C. I still wouldn't run mine at 90C though.
 
It looks like if I run 47 multi I would sit at 68c and lower when I'm encoding.

I run a few encodes a week and they can take between 3 and 10 hours pretty much 100% load throughout.

Do you think I would be damaging the CPU over time at temps like this (at 46 they max at 63c).
 
Stupid_Newbie said:
Mine isnt happy on my Z68A-GD65 G3, past 44x either. Though I dont break 70, (hit 66-68), but BSOD on 45 or higher with 1.4Vcore. Im going to play with lowering my Vcore to 1.35-1.38 for 44 I guess.
Click to expand...


My Z68A-GD65 G3 doesnt like multi's over 44x either, regardless of settings.
 
Spare-Flair said:
Don't run memory at over 1.5V and over 1333MHz on a 2500K. It's not supported by the Intel spec and can void your warranty (as explained by an Intel rep).

I blew the integrated memory controller of my last 2500K that way. It might be a rare situation but it can happen. Don't bother with memory at all. Continuing tweaking your CPU if you can.

So far I've had three 2500Ks. Two Malay ones from July and September and a Costa Rican one from last week. I'll have a fourth to test with soon. All three could do 5.0GHz prime stable at 1.4V to 1.425V so it may be something else in your motherboard configuration that is creating that wall for you.

I have LLC @ Ultra-High (ASUS forum rep suggested Extreme) and CPU Current Capability to 130% (ASUS forum rep suggested 140%). I have Internal PLL Overvoltage disabled. I'm using offset minus vCore configuration rather than set a static manual vCore (but that will work also) because I want to keep Speedstep/EIST for when I am idle. Why the massive vDroop? Setting a stable vCore is probably more condusive to system stability.

I would look to upgrade your cooling solution if you want to push to 4.8GHz to 5.1GHz. Prime small FTT will get the CPU 10° C hotter than Blend test if you want to try it out.
Click to expand...

Huh.. I got a 2500K @ 4.5ghz with the RAM at 2133mhz.. been fine for a long time now.
 
uhhhh.. dude... 100 mhz isn't going to make any difference. I would take the 4.4ghz and be happy. A 1% overclock isn't going to show any significant improvement. Take the overclock and if you're still not happy when Ivy Bridge comes out then buy one of those and see if it will clock further...
 
debaser* said:
It looks like if I run 47 multi I would sit at 68c and lower when I'm encoding.

I run a few encodes a week and they can take between 3 and 10 hours pretty much 100% load throughout.

Do you think I would be damaging the CPU over time at temps like this (at 46 they max at 63c).
Click to expand...

No, temps in the 60s are perfectly fine. In theory, you could run the CPU right at Tjunction (~100C) 24/7 and it should be fine.
 
Your problem is the lack of Load Line Calibration. Change the LLC to very high, and drop the vcore in Bios to 1.3v. 1.264v load is very good for 4.5 ghz. You shouldn't be holding back on tweaking the LLC.
 
debaser* said:
Correct me if I'm wrong, but tweaking LLC won't help with my load temps one bit.
Click to expand...

If tweaking LLC can give you more system stability at lower core voltage, then yes, if only slightly or not at all. Otherwise get some better cooling.
 
Last edited:
But won't my CPU still require the same amount of voltage at load? Tweaking the LLC will do nothing to lower load temps, it will just save a bit of power at idle.
 
debaser* said:
But won't my CPU still require the same amount of voltage at load? Tweaking the LLC will do nothing to lower load temps, it will just save a bit of power at idle.
Click to expand...

load line calibration is designed to limit Vdroop. It does not increase the voltage at load, just ensures a more constant voltage from idle to load. With your massive Vdroop, you may be using more voltage than neccessary to maintain your overclock. You are right though, that for all intents and purposes, you will be using the same power at load. Your system may be more stable though.

Get better cooling.
 
I don't really see any incentive to spend any money for better cooling. At the very best, I could probably hit 5ghz on a nice liquid setup, which would probably run me about $100 for an extra 400mhz.

I'd rather wait for the cheapest hyper-threaded K ivy bridge CPU in a few months (I could probably get that for net +100-125 if I sold my 2500K).
 
My stock temps at load after 3.5 hours of Prime95 were about 82C max...should I be worried? I got 2 case fans running.

Normally, I don't usually run apps that stress all 4 cores at 100%.
 
M'ichal said:
My stock temps at load after 3.5 hours of Prime95 were about 82C max...should I be worried? I got 2 case fans running.

Normally, I don't usually run apps that stress all 4 cores at 100%.
Click to expand...

If those are your peak stress temps you are fine.. you will probably never hit that under normal use. Try Intel Burn in Test.. I bet they climb higher!
 
Copyright said:
If those are your peak stress temps you are fine.. you will probably never hit that under normal use. Try Intel Burn in Test.. I bet they climb higher!
Click to expand...

I am scared :eek:

I did OC it yeterday though, just using the ASUS auto-thing, which pushed it to a nice 4.3 but the temps were more like 87.

I alos ran a 3DMark11 test and noticed NO performance gain. Is that normal? I mean it's like a 1Ghz OC. I read that CPU OC on these procs does not really help games much, so is that the case here? I'm probably not using the best CPU bench and should at least be looking at the CPU score of the 3D Mark test, right?
 
M'ichal said:
I am scared :eek:

I did OC it yeterday though, just using the ASUS auto-thing, which pushed it to a nice 4.3 but the temps were more like 87.

I alos ran a 3DMark11 test and noticed NO performance gain. Is that normal? I mean it's like a 1Ghz OC. I read that CPU OC on these procs does not really help games much, so is that the case here? I'm probably not using the best CPU bench and should at least be looking at the CPU score of the 3D Mark test, right?
Click to expand...

It depends on the game - BF3, for instance, responds very well to overclocking. If you are significantly GPU limited (using a 9800 GTX or something) than CPU will have a much less significant impact on overall performance.
 
You must log in or register to reply here.
Back
Top