i5 2500K cpu-z voltage

cinnamonandgravy

Limp Gawd
Joined
Jan 4, 2011
Messages
268
Been OCing my 2500K but I have some voltage questions.

does cpu-z report correct voltages?

I'm at 4.8ghz right now and cpu-z reports a max 1.3v during full load (IBT).

From memory, my OC went like this:

up to 4.6ghz with stock voltage
4.7ghz with +0.05v
4.8ghz with +0.3v

if the stock voltage is 1.2v, 4.8ghz should have me at 1.5v:eek:, but it only seems to be 1.3v. maybe my "+0.x" numbers are wrong.

temps are solid, IBT = 70C, normal gaming = 58C. all cores within 5C of each other.
 
Been OCing my 2500K but I have some voltage questions.

does cpu-z report correct voltages?

I'm at 4.8ghz right now and cpu-z reports a max 1.3v during full load (IBT).

From memory, my OC went like this:

up to 4.6ghz with stock voltage
4.7ghz with +0.05v
4.8ghz with +0.3v

if the stock voltage is 1.2v, 4.8ghz should have me at 1.5v:eek:, but it only seems to be 1.3v. maybe my "+0.x" numbers are wrong.

temps are solid, IBT = 70C, normal gaming = 58C. all cores within 5C of each other.

It may depend on your motherboard and the CPU itself. My Biostar reads my 2500K at 1.25V stock, so for a 4.4GHz OC without manual voltage control on the motherboard, it sits around 1.3V. I didn't bother fiddling further and the chip has since been relegated to backup machine duties anyway.

If I recall correctly, Sandy Bridge CPUs have preset voltage levels. Some motherboards, like mine, let you select a voltage increase threshold, then apply increasing voltage levels up to the limit that you have selected. Theoretically, my system could go up to 1.35V with the voltage limit I have selected, but I have not seen CPU-z actually read anything above 1.3V when stress testing or benchmarking.

It could also be a quirk in CPU-z. The software is not 100% accurate with voltages as far as I know, but relies on the VID that the CPU is calling for at the time. There is a bunch of interesting info on Sandy Bridge OCing here: http://www.overclock.net/t/978822/the-truth-about-sandy-bridge-vid-voltage-range
 
yes, CPU-Z report correct voltages, and actually 1.3v for 4.8ghz is very very low, you have a great overclocker chip.. but be sure to read the complete voltage isn't the same 1.306v as 1.398v. you can also try with another programs, like prime95 small FFT or Aida64 only FPU test, Asus RealBench its also a very good option (the best in my opinion) to test stability and voltages.
 
I'm manually increasing voltage via BIOS (asrock z77 extreme6).

max ive seen the voltage in cpu-z is 1.312v.

ill try some other programs to validate cpu-z's report.

if its actually 1.3v... maaaaaybe I can hit my pie-in-the-sky goal of 5.3ghz at 1.45v:D

ps cooler is some used corsair H75 (i think) with two gentle typhoon 120mm fans, push/pull.
 
well, there's a diminishing return point where voltage just start to increase disproportionately to the clock. so you will be probably end at 5.0-5.1ghz at 1.45V. and that's a really golden chip.
 
My 2500k is nowhere near that good, I'm jealous. But you're motivating me to play around with the settings a bit, and that's a good thing.
 
My 2500k is nowhere near that good, I'm jealous. But you're motivating me to play around with the settings a bit, and that's a good thing.

yeah not only you.. a golden chip like that make anyone jealous. :D a 5.2ghz will be better than a 4.6ghz 4690K so a lot of power there...
 
well, there's a diminishing return point where voltage just start to increase disproportionately to the clock. so you will be probably end at 5.0-5.1ghz at 1.45V. and that's a really golden chip.

4.7ghz was where i needed to starting increasing the voltage, but it only required the smallest bump in voltage to be stable. 4.8ghz, on the other hand, required quite a bit more voltage comparatively.

if 4.9ghz will also require quite a bit more voltage (odds are it will), ill be hitting the limit pretty quickly :( i want the OC to be used 24/7, so even 1.45v is getting a bit up there.

well, ill give 4.9 and maybe 5.0ghz a shot tonight.
 
yeah not only you.. a golden chip like that make anyone jealous. :D a 5.2ghz will be better than a 4.6ghz 4690K so a lot of power there...

Are there that many people getting 4.6GHz out of their Devil's Canyon CPUs? I was under the impression that the chips were binned from Intel much closer to their peak, so there wasn't a lot of overhead to play with. Also, there is the terrible TIM issue.
 
Are there that many people getting 4.6GHz out of their Devil's Canyon CPUs? I was under the impression that the chips were binned from Intel much closer to their peak, so there wasn't a lot of overhead to play with. Also, there is the terrible TIM issue.

you are probably referring to the i5 4570K and 4770K the first iteration of haswell chips, yes those are bad, however 4.6ghz was absolutely common if people were able to keep temperatures down.. but devil's canyon (haswell refresh) i5 4690K and i7 4790K have a better TIM and are very good overclockers, with special mention to vietnam batches. isn't nothing strange to see those at 4.7-4.8ghz at low voltages and acceptable temperatures.. specially the 4790K which are already at 4.4ghz Turbo Frequency.. and 4.5ghz can be achieved under stock voltage, the TIM issue is can be easily fixed by delidding it which also have its benefits like running direct die cooling with insuperable cooling performance even with low end coolers.
 
two bits of news:

first, i was wrong with my original voltage numbers by an entire decimal space. here's what i originally wrote:

up to 4.6ghz with stock voltage
4.7ghz with +0.05v
4.8ghz with +0.3v

its really..

4.7ghz with +0.005v
4.8ghz with +0.030v

now the voltage cpu-z is reporting makes more sense. if stock is 1.25v, 4.8ghz should be 1.28v with my settings, which resulted in 1.312v max measured. i dont like how it reached 1.312v when i selected 1.28v, but whatever. will investigate later.



second bit of news is a little more fun. when trying to get 4.9ghz stable, i got tired of increasing the voltage in 0.005 increments, so i said screw it, lets go from +0.03v to +0.1v. passed the "very high" IBT with a max reported voltage of 1.376v.

bumped it up to 5.0ghz, same voltage, passed the "standard" IBT with the same max reported voltage.

so.... 5.0ghz is showing promise at ~1.376v, and the current voltage is due to me being lazy... if it does prove to be stable, i'll dial down the voltage in 0.005 increments.

definitely too soon to say or not that my 2500k at 5ghz at 1.376v is truly stable, but so far so good. IBT got my temps up though... low 80s :(

proof: http://i.imgur.com/pEQDPTr.jpg
 
I would use hwinfo 64, It looks like you might be throttling back with the increase in time per calculation and decrease of Gflops. HWinfo will show you each core speed and the voltage as well and you can even have it show you graphs. Nice chip either way though mine wants 1.39 for 5ghz but my motherboard throttles from the voltage draw, z77 chipset doesn't have enough oomph for it.
 
hmm i think you might be right about throttling.. 4.9ghz was yielding ~126gflops

ive got the z77 chipset too... maybe same issue. will do more testing.
 
cpu1 was always the coolest core for me, and cpu3 was always the warmest.

up to 4.8ghz or so all cores were within 5C of each other... todays the first time ive seen a ~20C difference between the cores.
 
How is cpu1 at 65c and cpu3 at 81c

It is not that uncommon with the multi-core CPUs. It might be because of very slight differences in core quality that manifests when pushing things with overclocking. There can also be TIM-related issues, but as far as I know they shouldn't be too pronounced on the older chips, since the IHS was soldered on instead of relying on whatever manner of thermal paste they have been using in each successive generation.

Technically I guess it could also be possible that the IHS is in better contact with the heatsink assembly at some spots than others, but I doubt you would see a 20C difference per core resulting from that since the surface area of the IHS, while larger than the surface area of the die, is still pretty small; however, I am not a physicist or engineer, so maybe one of them can chime in with a better explanation.
 
i used IC diamond carat 7 TIM. when i first got the CPU i was a bit alarmed by the disparity in core temps, so i double checked the TIM; it was fine. ive heard a difference in 20C is fine and somewhat common, but i cant help but be a bit alarmed by it. even 5C got me a bit worried at first.

anyhow, had to bump up the voltage a bit to get 5.0ghz stable. now at 1.41x or so. played Ark last night for hours with my CPU at 100% usage practically the entire time... a single core didnt break 60C. not a single crash.

regarding throttling during the IBT, that's worrisome. I ran it again at 5.0ghz and got a minimal variance in GLOPS, and no steady decrease in GLOPS like last time. i stated earlier 4.9ghz yielded higher GLOPS... i think thats because i ran 4.9ghz using the very high test, whereas 5.0ghz used the standard test. I'm going to retest 4.7ghz - 5.0ghz using both the standard, very high, and extreme tests. also gonna run other CPU "gaming" tests at every frequency.

if it throttles during the IBT, yet gaming tests show increased CPU scores for every freq. boosts, ill carry on past 5.0ghz.

edit: also got my mobo/chipset wrong - its the asrock z68 extreme7 gen3. sorry - bought all this stuff off ebay not all too long ago
 
Last edited:
so doing an IBT at every frequency was taking too long, and was too boring to watch, so i switched to the newest 3dmark (basic edition) and the catzilla benchmark. here are the results.

catzilla
3.3ghz - overall 23,261; cpu - 382
4.6ghz - overall 24,963; cpu - 968
4.8ghz - overall 25,015; cpu - 1136
4.9ghz - overall 25,131; cpu - 1181
5.0ghz - overall 25,262; cpu - 1339

3dmark
3.3ghz - overall 7053; cpu - 5987
4.6ghz - overall 7423; cpu - 8313
4.8ghz - overall 7468; cpu - 8665
4.9ghz - overall 7485; cpu - 8863
5.0ghz - overall 7503; cpu - 9032

5.1ghz was a fail no matter the voltage. :( 5.0ghz was at +0.11v, so "1.36v", but it got logged as going as high at ~1.41x

for catzilla, running the cpu at 5.0ghz compared to stock resulted in a +251% cpu score increase. for 3dmark, it was a +51% cpu score increase. that aint bad.

how does that translate into actual gaming performance (~1440p and beyond)? too lazy to check right now, but id wager a handful of fps at best.
 
Try Asus RealBench IMO the best stability software out there for CPU. (specially for newer chips). as I said before there's a point where high amount of voltages required are inevitable, a diminishing return point and I was close with the calculations about your chip, 1.41v at 5.0ghz its still a golden performer and much much better than the average of great 2500K overclockers, so you have a reason to be proud of your chip.

Now its time to check temperatures, if you have issues with your cooler you can try lapping both the cooler and the CPU Heat spreader, that way you will have a perfectly flat and even surface to allow a even Heat Dissipation between CPU and cooler, decreasing overall temperatures.

about Gaming performance I have to say that the Higher the Resolution the faster GPU bottleneck you will be, far before CPU. you may still have some "Thread bottleneck" in some games as they scale flawlessly with more Threads and the difference between an i5 and an i7 can be huge as Crysis 3, Hitman Absolution or Watch Dogs, Crysis 3 is probably the greatest exponent of CPU performance, but you will be perfect for 99% of the rest of games. I can hardly Think in a recent game where a [email protected] can present any problem.. =)
 
only time temps broke 80 were during an IBT. gaming only sees one core (#3) breaking 60 @ 5.0ghz 1.41x v.

personally, im a bit adverse to CPU stress tests; they produce loads and temps your CPU will never see while gaming, so i really dont see the point given i use my cpu pretty much exclusively for gaming.

if my CPU never crashes with gaming/benchmarks, yet fails an extreme stress test, thats fine by me. :cool:

ill try for 5.1ghz next weekend, but it sounds like 1.45v wont cut it. :( you were pretty spot on with your estimates.

i always play games above 1080p, so ill do some benchmarks to see just how much a fps increase 5.0ghz nets over a few lower frequencies (3.3, 4.0, 4.5, and 4.8ghz).
 
that's the reason I love RealBench over anything else, it's based in real world usage and applications, so the temperatures you will see there aren't un-typical or unrealistic.. everything it's real and trust me work flawlessly. if you pass RealBench 5 consecutive Benchmark test (not Stress) you will likely be green light for anything you throw to that chip in the real world.

you can also test Crysis 3 (which is my favorite Stability Gaming test). if you can play 1 continue hour of crysis 3 you will be also green for anything game out there in the market, also when you are testing games and FPS try to see Minimums over anything else as averages and maximum variance may be low, but as always Greater Minimums mean a smoother gaming experience which matter much more than raw FPS numbers.
 
RealBench sounds right up my alley then. Ill give crysis3 a shot too. My friend convinced me to download Ark (unreal engine 4), and because its so unoptimized right now (i hope is the reason), all 4 cores were at 100% about 99% of the time. cpu was at 5.0ghz and i played like this for hours, so not bad. if it continues to be stable like this, ill be perfectly happy keeping this OC 24/7, since only one core breaks 60C and voltage tops out at 1.41v.

and im completely with you when it comes to fps minimums. my favorite thing to see is a super small delta between lowest fps and avg fps.

so for my grand ol' CPU benchy extravaganza, ill test 3.3, 4.0, 4.5, 4.8, and 5.0ghz, and ill test at 1080p, 1620p, and 2160p.
 
Back
Top