Official Intel Haswell (K) Overclocking thread

Well X5670 has 12threads, I'm sure those extra 4 threads will help a little by such rendering, even if cpu is a little older

Yeah do try lower VCCIN


Ah ok, cache is at somewhat ideal voltage for x43, mine was stable at 1.20v as well, but ~1.22v made it more consistent, although I didnt really see any difference x42 vs x43, also its for memory controller and can be a bit "unpredictable" at higher stress, e.g. by games have fps drops if its not ideal voltage. I usually keep it at x42; x43 or x44 just for brute force benchmark to squeeze that extra point or two :nerd:


About cpu power, 130% is more then enough

 

Hi TJ

Agree,this X5670 has been beast,but right is residing in my friend PC and he using this PC for rendering,I'm using Poser Pro and render engine used in poser Pro doesn't use AVX(Intel Embree) like V-RAY using and many other and there what counts is cores/threads,I will do few tests in 3DS MAX later and I'm curious what times will be in V-RAY or Corona,tried Corona bench and my render time is around 3:25 with 4.7GHz OC,X5670 at 4,2GHz render time has been 3:45

I will lower VCCIN when I finish rendering and try then if my temps will drop bit further

I will probably lower cache frequency and test that in AsusBench,tried with x42 and then with x44 and results are minimal,I will post my results later


Thanks again,Jura

 

Hi TJ

Seems my stable OC at 4.7GHz is no more stable,has been stable for 3 renders in row,then BSOD with 124 which is not enough vCore,tried everything and seems my chip is not as good as I thought so

Tried whole night and day play with the BIOS and stable vcore I found is around 1.31v with 1.79 VCCIN and cache at x43 with 1,17v,temps are in 70's or low 70's on hottest core,lowered too clocks on RAM,now running 2200MHz

On other hand is stable in any bench what I've tried like AsusBench,OCCT,IBT or x264,plus tried Handbrake to do quick conversion of BD

Here are my new OC settings at 4.6GHz



Realbench results







What is safe vCore for 4790k,is still around 1.35v or its more?

Thanks,Jura

 

Hey,

hm are you sure it was also Cache fault causing it? Btw 1.17v is really good for that cache :nerd:

I remember testing voltage on mine and it scaled linear up to 4.3Ghz then a big jump occurs

x40 @ 1.10v
x41 @ 1.125v
x42 @ 1.153v
x43 @ 1.218v (at first I tested ~ 1.18v -1.20v but later found out instability)
x44 @ ~1.26v (1.24 or 1.25v no go), btw auto voltage sets it even higher..


About cpu freq. Im able to be stable at 1.287V @ 4.7GHz, did you test in small steps or you went straight to 1.31v?


Also do you OC in bios or you use AI suite just to show results? Well OC'ing in AIsuite is not really recommended for stability reasons.

About ram freq. system agent voltage can help by stability too, extra 0.020 - 0.050v offset at higher ram freq.

What LLC do you use in DIGI+, Auto or manually set?

 

Hi TJ

I'm not sure what is causing that,I'm just guessing there,really this is my first newer CPU,previously I've used only i7-920 or Xeon X5670 or X5675 etc

I will try to play again in BIOS later this day and will see if will be stable

I tested with small steps,like with 1.26v and at 1.31v PC has been stable for rendering and do several benches,but temps has been bit higher in 70's

I just need to play with vCore and test that again

Yes I do OC in BIOS,I wouldn't do that in AI Suite,no way,its too much risk for me to do,I'm using AI Suite only for showing voltages etc

Aha,OK this I will try with System Agent Voltage and will post later on what happen,maybe this will help me with higher frequency RAM although mine are CL11(I would say they're lazy RAM)

LLC I think its set at Auto,I'm thinking go route with LLC Level 8,but not sure on those CPU,on my previous X58 there I've set at LLC level 3(I think this has been max)

Thanks again for yours help there,Jura

 

OK tried to again to rework my BIOS settings and tried to finally finish my OSX installation which has been pretty much very easy and bought finally Phanteks Enthoo Primo SE (Red/Black) which I wanted to at least to test this case

What I can say regarding this case,its awesome case,big and very heavy(almost 20kg empty),working with this case has been great,just I can't put there my second card GTX780 without the removing bracket and not sure right now

But my temps drops under idle are now in 28C,previously with 750D my temps has been at 34C,with Phanteks I've dropped RPM for H100 V2 fans to 750RPM,with 750D I've run same 750RPM

Under load with 750D my temps has been in high 60's or earlier 70's at 750RPM,with Phanteks temps are in low 60's under load

HDD temps are in 20-30's in Phanteks and in 750D my temps has been always in 36-38C,off course in SSD temps has been consistent in 24C

In therm of temps of GPU(Titan X SC) on idle are right now 33C(I've seen lowest as 28C with all fans turned on full),under load 76C,with 750D my temps has been on idle at 36C and under load 83C with all fans turned on

As right now I'm very happy and regarding to OC,been playing with vCore and seems right now I've found stable vCore at 1.24,I've returned RAM speed to 2400MHz,cache is right now at 43x and cache voltage is 1.16v

And here are pics of the Phanteks Enthoo Primo SE





Thanks,Jura

 

Fast forward 3.5yrs and some extra guide info

VCCSA - cpu system agent does help by cpuv and lowers it. I saw this debate once, but when I tried it it didn't look promising, now I see I didn't try high enough.


I always tested max up to 0.060 - 0.070v, making it total ~ 0.880v, safe water apparently up to 1.30v, air ~ 1.10 -1.20v.
For me 0.130v+ offset (total 0.940v) did the trick and I was really able to lower voltage from 1.288v to 1.268v

I tested 0.080v with cpuv 1.274v and it failed, passed at 1.278v, now extra 0.050v (0.130v+ offset) and I was able to pass 4.7ghz @ 1.268v.
Also now I saw up to 2-3C lower per core temp. vs old 1.288v and vccsa 0.050v+ offset :nerd:


http://valid.x86.fr/k6e8l0

 

If anyone doesn't care to, could you please shed some light on this or gimme an idea what to change here.

I hit a wall yesterday and noticed some erroneous? readings. AUXIN, which I hid in HWINFO and I want to say ring voltage.

I panicked thinking ring voltage was at 1.8, so I set it to 1.2 manually which is exactly what auto sets it to. Seems high for x39 cache regardless but nonetheless, it still shows 1.8v in HWINFO. At auto (1.2), it shows 1.8 in HWINFO and +0.00 offset in HWmonitor. HWmonitor shows correctly if set manually.

 

Try aida64 monitoring, that will show proper values for sure.

 

I have it downloaded but not installed, which I will definitely do.

I've stressed it a few times around 40mins each with IETU and it has passed. Manual Vcore but with Vring/cache at auto.

In between, I do random things I'd normally do. Play games etc and all was fine, until I tried RPCS3 PS3 emulator. Skate3 made it crash with "whea_uncorrectable_error".

I'm starting to think luck wasn't on my side, got a CPU that's a turd as far as overclocking goes.

Here's HWmonitor after 40mins of stress. Temps are mediocre, Vcore is not acceptable for 4.4Ghz IMO. Really not wanting to pass 1.3v on air. -_-



EDIT: Pushing it hard, doing this as i stress it. lol

For good measure, I flashed the motherboard with the newest bios, after clearing cmos.
Started over

 

Well it will get hot after 1.25v, especially after 1.30v, also don't OC cache too much, try to stay ~40-42x multi. Too high cache can break cpu OC and you have to use lower.


Also dont brute test prime95 etc, try asus real bench or fhd x264 benchmark

Or even game CPU test LostPlanet2 - testB, this will heat the most and show instability too if not right. The division, new tombraider, resident evil 6 bench or gta5 or BF series can show instability quick too..


0X124 whmea bsod, is either cpuv too low or cache OC too high or with not enough voltage, or vccsa (cpu system agent too low), although I saw you already have 0.200v+ offset so that should be more then enough.

Imo 0x124 bsod is not so scarry, 0x101 watchdog is.more problematic and harder to solve.

 

That was at 4.5Ghz. It has yet to crash, which I appreciate the info on BSOD. I'll definitely keep it in mind.

So far, the only problem I have is trying to set the cache. Like as soon as I manually set it, it's unstable, even with 1.2vring.
I may just leave it auto, as it's stable so far and mess with the memory.

As you've probably noticed, I have a z97 motherboard. The plan was to go with a 4790k but I seen this 4770k on craigslist for cheap, so I jumped on it.
By chance I do find a cheap 4790k, I may jump on that as well anyway. That or a GPU, which would make more sense now that I have a decent CPU.

EDIT: Add PSU to that list. I've noticed it dipping below 12v under stress, plus the fact I've had it since 2009.
Still, going the route I have, I feel has been a good idea. I've saved quite a bit of cash over building a new system entirely.
No, it's not an 8700k but it still gets the job done.

EDIT: I apologize for adding so much crap to the thread but it's better than double posting, I guess.

Nevertheless, just as TJ suggested.





https://valid.x86.fr/16cd5t
Monitor overclocked, 70Hz. ^

I'll run it till it crashes.

I'm guessing I'll get watchdog, more than likely. lol

 

Hi
I was able to get a stable overclock of 4.6 on 1.215 volts. However I get bsod on 4.7 even when increasing all the way to 1.35 volts.
My thermals do not appear to be the issue. I have a noctua d15s.
In my bios, I have the multiplier set to 46 sync all cores and voltage at 1.215. Everything else is auto.

It seems strange that I can be stable 4.6 at relatively low voltage but not be able to go 4.7. i use real bench to stress test
I tried increasing the input voltage all the way to 2.0 but this issue still occurs. I also tried increasing the cache frequency and cache voltage but that did not solve the issue either.
Seems crazy that 4.6 is stable at 1.215 but 4.7 is not at at 1.35
Thanks in advance

I have 4790K on Asus Z97A
2 x 8GB 2400 DDR3
EVGA 650W Gold PS
Noctua D15s Cooler

 

1.35V is going to generate a lot of heat on these CPUs and heat reduces the OC potential. The only way to get a decent OC with these CPUs is a delid. I did it with mine and saw a ~20C temp drop. Like that I managed 4.7 @ 1.27V

 

Thanks but heat isn't the issue. I can't get stability at 4.7 even though 4.6 is stable at 1.215 at great temps.

 

If you say so.... I think you'd be surprised.

 

How can one tell if the crash is due to voltage or heat?

 

Well I think the fact you can't get it stable @1.35 speaks for itself.

It's possible there are some mobo settings that could be used to improve power/voltage stability or the PSU itself has issues.

 

@Darren Hodgson

If you're still at z87deluxe

Here is the latest modeed bios
https://www.win-raid.com/t3517f44-OFFER-ASUS-Z-DELUXE-updated-BIOS-v-NVMe-mod.html

Using the UBU 1.76.2.5- with UEFITool v25 I also upgraded the following:
EFI IRST RAID for SATA 12.7.0.1936 ----> 15.1.0.2545
EFI GOP Driver Haswell 5.0.1036 ----> 5.5.1034
EFI Intel Gigabit UNDI 5.4.19 ----> 0.0.29
EFI Realtek UNDI 2.020 ----> 2.053
CPU microcode 19 ----> 28 (latest Spectre/Meltdown & now MDS mitigating microcode) Thx Sonix! [Latest edit: Date: May 7, 2020]


Use usbflashback to update it.