cancel
Showing results for 
Search instead for 
Did you mean: 

Is it worth disabling XMP to overclock processor?

Qwinn
Level 11
When I first put my system together, I didn't overclock anything other than that I enabled my 2666 MHz memory XMP profile. I never once disabled it during my last couple months of overclocking attempts. But I've been thinking, what are the realistic chances that I could achieve higher stable processor overclock if I turn off XMP and just run memory at 2133? And would 100 MHz processor overclock more than compensate for that level of memory downclock?
38,493 Views
60 REPLIES 60

Chino
Level 15
Only way to find out is experiment. And yes. I would gladly prefer the extra 100MHz over running memory at higher frequency.

Nate152
Moderator
Hello Qwinn

I suppose you could do a little testing. 🙂

If you're after another 100Mhz I don't see why you couldn't do it with your ram at 2666MHz, you have good cooling and you should be able to pull it off rather easily.

Qwinn
Level 11
You'd think, but no. I have failed to get any stable manual overclock at all, though I haven't tried below 4.5Ghz, because I already had a stable adaptive 4.5Ghz (1.2v, offset +.75, gets up to 1.312v during stress testing). Trying 4.6 Ghz adaptive wasnt stable, but I didnt try disabling xmp.

Nate152
Moderator
When I overclock my cpu I actually do it with the ram at default speed then once I get my max overclock that temps allow then I set the ram.

The reason I do it that way is to take any instability with the ram out of the equation, so if I get a blue screen I know the cpu needs more voltage and I'm not guessing that the ram is unstable.

Some people might do the ram first, I guess it's all a matter of preference.

chrsplmr
Level 18
Qwinn - - Welcome to ROG.

This is a great article. (If you haven't seen it yet.)
http://www.hardocp.com/article/2014/11/13/asus_rampage_v_extreme_lga_2011v3_motherboard/7#.Vl_XrPmrT...

A quick review of ROG RealBench for the Rampage V Extreme might give you some clues at running your rig at peak.
Even though the 5960X seems to own the top of the LeaderBoard > (HERE)
The 5930k is no slouch as seen below:
http://rog.asus.com/realbench/show_comment.php?id=8623
http://rog.asus.com/realbench/show_comment.php?id=4965
They take over > (HERE)
You will notice that none are running @ 2666mhz to achieve these peak performance scores and stability.
((This may be because those running DDR4 haven't RealBench'd. What are they waiting for ?))
It all depends on your goals with this build. Do you want it to perform at Peak Performance and Stable ?
((If you click on the score you will get the stats for that run. btw))
Could you provide screenshots of CPU Z <(download page), with CPU tab and MEMORY tab showing ? (open it twice) This can be very helpful.
((I mean no disrespect or to imply that you are unaware of tools and tricks of the 'Game', Please keep in mind you
and I are not the only ones reading this, they may not be. humbly offered.))
You may find this helpful or at least interesting as well: http://www.ocaholic.ch/modules/news/article.php?storyid=10713

Best of Luck in this. Please keep us Posted. .c.

Qwinn
Level 11
chrsplmr:

No offense taken! I do still consider myself a noob at all this, even though I've been tinkering for a month or two. Thanks much for those links!

"You will notice that none are running @ 2666mhz to achieve these peak performance scores and stability."

At first I thought you meant they were all downclocking memory to get the better processor overclocks and scores, but looking at them, aren't the first two running at 2850Mhz and 3000Mhz respectively? I thought you doubled whatever those readings are. (Confirmed in my Realbench score posted below - I'm running at 2666 and it's saying I'm at 1333).

My CPU-Z stats.

53927

Note - I'm running adaptive which is why core speed is showing low. My turbo multiplier is 45, and the highest the vcore has gone to is 1.312v, which I believe is perfectly safe? Please let me know if that's edging into degradation territory... I'm assuming it's fairly safe since it only hits that on peak load, and usually it doesn't need to go over 1.296v even during stress testing. What I've gleaned from my reading is to not go over 1.35. During stress testing highest I ever saw package temp go was 73c in Intel XTU. In gaming (Witcher 3 in 3D Vision on mostly ultra settings as my benchmark) I rarely see it go over 60c. Technically max operating temperature of the 5930k is listed as 67-68c (I forget which), but I've also seen dozens of threads saying 80c is just fine, so I figure I'm doing ok there.

See this thread for my description of my recent efforts:

http://rog.asus.com/forum/showthread.php?79884-Very-odd-overclocking-results

Here's my best Realbench so far:

http://rog.asus.com/realbench/show_comment.php?id=10075

Would love your thoughts on what I've mentioned in that thread. Those links were excellent reading, btw, thanks kindly.

EDIT: Also thought I'd share a Firestrike I just ran on my current 24/7 super-stable settings:

53928

I've actually gotten Firestrike up to about 25,300 previously, but it wasn't super-stable at those settings. I can only get about 1418 Mhz and 3800 memory clock on the two 980 Ti's totally stable, though I can get to 1455 Mhz and 4000 memory *mostly* stable, will crash in Witcher 3 after a few hours though. That's at stock voltage for the cards, but if I tweak voltage *at all*, it's very unstable almost regardless of whatever other settings I apply. Upping the voltage doesn't help whatsoever. Probably just as well though, on stock voltage the GPU's can get as high as 79c with the fans just shy of being annoyingly loud, so probably wouldn't want to increase temps on them anyway. I've also discovered that voltage goes way up on these cards when you run 3D Vision anyway, which I do constantly, so upping it even more is probably a bad idea regardless.

chrsplmr
Level 18
Qwinn,
Yes, you are exactly right .. my error. ((I have no excuse, my apologies.)

Your score is right up there from what I see of the 5930's ..
Although your memory is different you can see by the top 5930 score
http://rog.asus.com/realbench/show_comment.php?id=8623 how well your doing ..
the difference that jumps off the page is the Multi .. and they bumped up the Bus a touch
and just a little tighter timing ..
Pretty sweet FireStrike as well ... you have been bitten my friend. The OC ROG bug strikes again. lol
From what I've read I think you are correct and more than in the saftey zone .. keep in mind that any voltages over
manufacturers specs 'can' degrade .. stock can degrade. (it's the luck of the draw of cpu) It has been my experience that they are much more resilient than we think. Just keep an eye on your temps.
Before I got my current cpu it had endured 7.2 on LN2 ... and I have .. well, it's obscene. lol
(thanks again HiVizMan)

You are doing great.
You may want to post a new thread .. ie .. "Help my Rampage V ROG to the next OC level. Please." .. or some such posting your
progress so far as above, you will find a quite helpful lot ..
Well done Qwinn .. glad ur here. .c.

Hawkstorm
Level 7
Your best bet is to get your kit working at the frequency you want first then go try overclock your cpu.

If your Kits setup right in the first place and you memtest the hell out of it etc and its fine then it really shouldn't make much difference really on or off, you may have to add a tad of system agent.

Though only difference you may find is that depending on your kit you might 'loose' a bit of vcore to keep your kit stable if your kit requires that in the first place.

Shouldn't make any difference really.

Max temp on the 5930k is 115deg if I remember correctly so 80 is fine, as for voltages that comes down to your cooling and how much heat it can handle, with the Auto volting I've seen my board exhibit mine goes over that under heavy load, which screws over my clock as I could clock my system at 4.6ghz on that voltage instead I have to go for 4.4ghz on a lower voltage to cope with the 'Auto' unwelcome increase.

1.35v isn't really too high if you have a really good custom loop, with multi radiators, as you will seriously pull some wattage and the heat output will be over 80deg, and pull around 240watts, mine pulls that at about 1.33v which it jumps to for the set 1.21v I set it at (with the stupid Auto bugged Turbo Volting on this board which I don't seem to be able to stop.)

Personally I have turbo disabled and speedstep, though there seems to be a error in that under heavy load Turbo voltages are still apllied to this board, making it impossible to get a stable voltage which pisses me off, as it cuts into my overclock, in fact I just confirmed it there's no difference between Turbo being on or off it still does it and its not LLC that's the culprit though TBH that seems to do little also when adjusted from level 1-9 nothing no change at all.

My system is at 1.21 vcore manually, which 'Jumps' to 1.33 underload kind of pointless setting a voltage really isn't it?

Because of this I can only get 4.4Ghz out of my cpu without it going to 1.39v and way out of my TDP for my closed loop water cooler. Currently OCCT max's out at 87degress before it trips on AVX linpack. The excess heat buildup coming for the insane 0.120 odd volts being applied for me that I don't want or require.

Tried adaptive (as it has a turbo offset I set it next to 0 trying to stop it) but it still happens and tbh there seemed to be no gain to using it doing a static overclock.

87degrees is under extreme load so I'm happy with it though personally I don't like going over 80.

So ASUS 1# thing ANY overclocker wants is STABLE voltages, this board you don't get them it seems, all those lovely extra pins on the socket so you can have lower voltages nice! THEN IT GOES AND ADDS 0.120 FOR YOU! (Rant)

I get 10036 with my 2 760's in SLI, so that's a nice score, but then your GPU's will effect that score more than anything.

I'd be interested to see what you get when you run OCCT CPU test as far as your voltages and temps go, is your voltage jumping like mine?

Because your doing better as far as CPU speed than me, though manually I'm on 1.21v at 4.4ghz and your at 1.28 at 4.5ghz but my voltage spikes at 1.32v under OCCT load, I'd be interested to hear if you board is also adding 0.120 volts under load.

Hawkstorm wrote:
I'd be interested to see what you get when you run OCCT CPU test as far as your voltages and temps go, is your voltage jumping like mine?

Because your doing better as far as CPU speed than me, though manually I'm on 1.21v at 4.4ghz and your at 1.28 at 4.5ghz but my voltage spikes at 1.32v under OCCT load, I'd be interested to hear if you board is also adding 0.120 volts under load.



Hi Hawkstorm,

No, don't go by my CPU Specs list, I put 1.285 because that's what I thought I *should* be getting, but my system does indeed go over that on load. I have fixed my specs accordingly.

The only stable overclock I've managed to get to is in adaptive mode. It's set at 1.2v turbo with +0.75 offset. You'd think that would mean I'd max out at 1.275v, but instead under stress testing, monitoring with HWINFO64, vcore can actually get as high as 1.312v, so yes, I am having similar voltage spike issues, though at least to a tolerable degree. I have my cache overclocked to 3.8Ghz with static 1.2v cache voltage. I think I could raise the cache a bit and remain stable with more voltage, but I've been given the impression 1.2v is the maximum cache voltage recommended for 24/7 use. Note that I have *not* witnessed any voltage spikes on that static 1.2v cache setting, that one appears to be working as intended.

I was unable to get a stable overclock in manual mode, although honestly I didn't try for anything less than 4.5 Ghz (because I already have that much with the settings above), and I didn't try tweaking system agent or anything besides the vcore, because I am noob and nervous about touching those without a firmer grasp of what is safe and what isn't. I don't notice much fluctuation with those other voltages in HWINFO64 with my adaptive 4.5Ghz settings though... I took a look just to make sure auto settings weren't driving those through the roof, but no, they are in line with what I'd expect them to be at stock, or close to it, even under stress testing. I did try running 4.5Ghz with manual 1.296v (which is what I originally thought my adaptive was maxing out at) and that did not work, so I can't see what I could gain by further pursuing manual settings, if my result would have to be either slower speed or higher voltage than my adaptive is maxing out at.

Incidentally, I'm not using a custom water loop. GPU's are on air, CPU does have a very nice AIO liquid cooler with the biggest radiator I've seen on market, 360" with 3 120" fans on it. Radiator is so damn big it actually made me unable to fit the OC panel in a 5.25" bay, unfortunately, heh.