• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Overclocking Help: What Am I Doing Wrong?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
As I said every cpu is different but the multi had a hard wall on SB. Literally, one would work, the next (and thereafter) wouldnt NO MATTER WHAT. This wall was typically 5ghz+ and with those cpus, top notch air or water would find them. This is how they were binned by hardcore overclockers... Myself, mandrake, mr scott, etc all did this back in the day! literally on POST you would see a cursor blinking in the upper left corner and that was your sign you hit the multiplier limit for the 2500k/2600k cpu. Just because there is a 5ghz club doesnt mean there isnt a hard multi wall on every. single. 2500k/2600k. As I said above...it was 5ghz+. I think the highest was 55 or 56. Some where 52..53...etc. it all varied but this wall was unlike any other with its blinking cursor.

Any more about this, PM me or start a thread and I'll drop these off topic posts in here. Last warning. :)
 
OP
J

Jems

New Member
Joined
Feb 15, 2019
Well . . .

I learned a lot about the debate on hard wall vs. no limit so far! :)

The strange thing is that when I say stable, I mean Prime95 stable. I can boot into 5.35 and even benchmark with it. It's just not stable for prime.
9iy06k.png
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
Well . . .

I learned a lot about the debate on hard wall vs. no limit so far! :)

The strange thing is that when I say stable, I mean Prime95 stable. I can boot into 5.35 and even benchmark with it. It's just not stable for prime.
I wouldnt worry about that if it is actually stable for your uses... :)
 
OP
J

Jems

New Member
Joined
Feb 15, 2019
Have you tried to increase the core voltage?

As I said in my OP, it doesn't matter how far I increase vCore, it just seems not to be stable. I've gone up to about 1.36 or so, which will let it sit just under 1.4 under load.

Also, to answer another poster, yes, I updated the UEFI a couple of days ago. It helped the temps.

Earthdog: true, but I do research and write. Knowing my luck, It'll hit that glitch randomly right at the time when I've just written a brilliant paragraph, which almost never happens, and then lose it with a BSOD. Been down that road before, actually. Was I ever peeved! :mad:
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
Have you tried using it as is yet? Again, p95 isnt The Gospel... what are you running in p95 and how long until a thread drops?
 
OP
J

Jems

New Member
Joined
Feb 15, 2019
When it's beyond 5050 Ghz, it doesn't matter whether it's small packets or the blended test, it doesn't last much beyond the workers ramping up for the first time. Drop it back down to 5020 or so, and it goes 15 hours.
 
OP
J

Jems

New Member
Joined
Feb 15, 2019
Okay,
So I went ahead and played with it some more. Turned vCore to auto and let 'er rip (well, with a couple programs open to monitor vCore, of course). I also messed with the memory, moving it up and down a bit to see what might help. As I thought, To boot at 5.3 Ghz was taking 1.43-45. (I quickly shut down). 5.15 under load will top out at 1.38 vCore on auto, but I still get BSODs running prime Blend. So, perhaps I have hit the sharp bend in the volt curve at 5.03 (I think I can get it close to 5.1 stable, but really, it's not much difference). I did get the memory OCed to almost 3600, though, which was nice.

Anyway, I'm going to chip away at this some more. Let me know if anyone has any other thoughts.
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
You need to overclock one thing at a time...work on cpu. When that is done, then go memory. P95 blend tests more memory than cpu too... so its likely memory causing it to fail. If you want to test cpu... use small fft.
 

Woomack

Benching Team Leader
Joined
Jan 2, 2005
I would set 5GHz and try to lower voltage as much as possible. On my asus motherboard I had to set 1.27V to make it work without issues. Under load it had ~1.32V at 5GHz. There is one more thing. Real voltage depends on more factors. When SVID is enabled then even if you set LLC close to desired voltage then under load it can be something else. On every motherboard I had to set different voltages to keep my 9900k stable. It was between 1.27V and 1.35V while under load it was ~1.32V ... again regardless of LLC and other things. I had no chance to test 9900k on MSI so I can't say much about how it's acting.
 
OP
J

Jems

New Member
Joined
Feb 15, 2019
You need to overclock one thing at a time...work on cpu. When that is done, then go memory. P95 blend tests more memory than cpu too... so its likely memory causing it to fail. If you want to test cpu... use small fft.

That was my intent, actually, and what I was doing. However, when I started getting BSODs that pointed to memory error, I started wondering if the disparity in speeds was a possible candidate. That's why I began bumping the memory as well. I used the XMP, then any difference from there was usually just the difference in the multiplier at first when I was still fiddling with that. Then, I'd include a slight increase or decrease as part of my problem solving to see if I can get the overall system stable. I went that way because I'm admittedly working from ignorance on the particulars of OCing and whether memory being too slow might case a BSOD. If you or anyone can clear up my ignorance in this area, I'd greatly appreciate it.

As for Prime. Small FFTs crash almost instantly on anything higher than what I have now. Blended seems to be solid.
 
OP
J

Jems

New Member
Joined
Feb 15, 2019
I would set 5GHz and try to lower voltage as much as possible. On my asus motherboard I had to set 1.27V to make it work without issues. Under load it had ~1.32V at 5GHz. There is one more thing. Real voltage depends on more factors. When SVID is enabled then even if you set LLC close to desired voltage then under load it can be something else. On every motherboard I had to set different voltages to keep my 9900k stable. It was between 1.27V and 1.35V while under load it was ~1.32V ... again regardless of LLC and other things. I had no chance to test 9900k on MSI so I can't say much about how it's acting.

My goal is to see how fast I can get with a vCore under 1.4 at load. Failing that, your comment is exactly what I plan on doing.

So you have core stability issues if small FFT is crashing.

Yep, which is what led me to start posting here that I can't get anything stable above 5.02.
 
Last edited:
OP
J

Jems

New Member
Joined
Feb 15, 2019
I hit it! 5.1 stable!

So, here's what I did . . .
1. RMAed my MSI ACE MOBO (Wifi was dropping repeatedly while my laptop two feet away would keep the connection).
2. Went back to the boards I know - Asus Maximus XI Formula (MSI ACE wasn't in stock so I'd have to wait for it - decided to go home to Asus).

Settings on MOBO
XMP 2 (3200 Ram).
Bclock @ 100 and Core Ratio @51
Load Line Calibration level 8
Sync Loadline w/ VRM loadline - enabled
CPU current capability 120%
CUP VRM switching Frequency - Manual (500 khr - thank you Asus for the water block on the VRMs)
Power Duty (? can't read my own writing at the moment) CPU - Extreme
Bclk amplitude: 900 mV
Ring - Auto (haven't dialed that in yet)
Bclk aware - Enabled
vCore 1.285
AVX Offset - 0 (I'm testing with non-AVX at the moment. Wanted to get a solid OC outside of AVX, then downtune for it after).

It survived 30 minutes of Prime Small FFTs (but with AVX turned off, again, was wanted to stabilize the reg. overclock before messing with AVX), 20 minutes of OCCT (had to bump the vCore to what you read above from 2.80), and 15 minutes of Real Bench.

For prime, the temps were toasty, in the 90s, but no throttling (see first screen capture). My VRMs are staying right under 50c (SC2). Of course, my ring (uncore) is only 4300, but again, I haven't OCed it yet. I'm planning on running prime for 16 or so hours with a blend just to make sure. But even if it fails, I'm pretty dang happy with where I'm sitting right now.

aarwwg-4.png
Capture.PNG Capture1.PNG
 

batboy

Senior Moment
Joined
Jan 12, 2001
Location
Kansas, USA
I don't remember a brick wall when I had a Sandy Bridge. I swear that I remember hitting 5.4 gig (but that was probably not stable). I was running benchmarks at 5.2 though.

2600K-5200.jpg
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
You don't have to remember it... but it was there!!! We lived it! Its all over the web/OC guides of the time. This was ironed out over a month ago in this thread and doesn't have anything to do with the OP... please stick to the topic. :)
 

DaPoets

Member
Joined
Aug 23, 2007
From what I have seen w/ the 9900k in the end it's just a hot chip and that's what the brick wall is. If I ever go the 9900k route I already know I'll end up going through the long and difficult delidding process because I honestly think it would be fun and you would see a bit better temps. Not amazingly better like the 8th gen get from delidding but at least I would know that I then had the best possible ambient based temps from a custom loop.

Keep things in perspective though w/ the 9900k, it is already fast as fast can be today so pushing it much further is really only for the sake of doing it and not really any noticeable real world benefit.

(We do this for the love of the game)
 

Alaric

New Member
Joined
Dec 4, 2011
Location
Satan's Colon, US
Keep things in perspective though w/ the 9900k, it is already fast as fast can be today so pushing it much further is really only for the sake of doing it and not really any noticeable real world benefit.

(We do this for the love of the game)

That's why the name on the shingle is "Overclockers" and not "Meh, Good Enough" :thup: