Need a sanity check

Lochekey · Sep 13, 2015

So let me introduce myself. MY name is joe but I go by lochekey mostly. I have been benching for about a month now and decided that I enjoy it enough that I intend to make it a normal activity. As such I decided to join the overclockers.com team for some camaraderie. I am mainly looking for a sanity check on what I can expect form my hardware. I am currently trying to improve my XTU score past my personal best of 1101 with an I5-4690K. I am able to stabilize my cpu at 4.9ghz with 1.49 volts but could not get it to complete a benchmark run at 5.0ghz even with 1.575 volts. I am using 16gb(4 sticks) of mushkin blackline 2133 10-11-10-28 ddr3 to back up my cpu. I have pushed the ram to 2400 9-12-11-33. I am unable to get the ram stable at anything past 2425 even with 1.8volts and loosening timings. I have posted a screenshot below of my bios settings as well as my ram timings. I am hoping someone can tell me if I am just expecting to much hoping to push this hardware further, or maybe you can spot something glaringly obvious in my set up that is hindering me. Either way some feed back in the form of a sanity check would be appreciated.

Thanks
Lochekey

Woomack · Sep 13, 2015

Welcome

You probably won't make it pass XTU @5GHz without sub 0 cooling. Even 4.9GHz is usually impossible, even though most other tests are passing.

1. I see you have power saving options enabled, disable eist, c states and enable performance mode in windows power settings.
2. Lower VCCIN when you are not benching on sub 0. 2.3V is too high for some chips and you can see degradation in some time ( it's not a rule but may happen ).
3. If I'm right then your memory is based on Hynix so I would set 1.75-1.85V and timings like 10-12-12-30 and check how high can you set clock. Next make the same at 11-13-13-35. Usually CL10 can run at 2400-2600 while CL11 at 2600-2666. Best chips can make 2800 CL11 or 12-14-14 but lower clock at tighter timings will offer higher performance. Also use 2 memory sticks as 4 will lower max OC.
4. Try higher cache clock - about 300 MHz below CPU clock. Set RING voltage to 1.35-1.40V.

No more ideas right now.

Lochekey · Sep 13, 2015

Woomack said:
Welcome

You probably won't make it pass XTU @5GHz without sub 0 cooling. Even 4.9GHz is usually impossible, even though most other tests are passing.

1. I see you have power saving options enabled, disable eist, c states and enable performance mode in windows power settings.
2. Lower VCCIN when you are not benching on sub 0. 2.3V is too high for some chips and you can see degradation in some time ( it's not a rule but may happen ).
3. If I'm right then your memory is based on Hynix so I would set 1.75-1.85V and timings like 10-12-12-30 and check how high can you set clock. Next make the same at 11-13-13-35. Usually CL10 can run at 2400-2600 while CL11 at 2600-2666. Best chips can make 2800 CL11 or 12-14-14 but lower clock at tighter timings will offer higher performance. Also use 2 memory sticks as 4 will lower max OC.
4. Try higher cache clock - about 300 MHz below CPU clock. Set RING voltage to 1.35-1.40V.

No more ideas right now.

thanks for the advice I will work through these suggestions and see what I get. I was benching sub ambient with these settings if that makes a difference. An ice bath for my rad did wonders.

Mandrake4565 · Sep 13, 2015

Lochekey said:
thanks for the advice I will work through these suggestions and see what I get. I was benching sub ambient with these settings if that makes a difference. An ice bath for my rad did wonders.

Even with -12c water flowing through my chip I couldn't pass XTU higher then 4.8 Lochkey, Even on DICE my best is 5.3 Ghz on XTU. When woomack is talking Sub ambient hes talking DICE/Ln2. I will add Woomack is out resident memory expert so if you need help with ram he's the one to talk to.

Johan45 · Sep 13, 2015

That helps but high voltage is still iffy with these CPUs. I would stick with 4.9 @ 1.49v unless you can get even colder you're not going to get 5.0 stable. Like woomack says XTU like memory/cache speed so work on that. Try using AIDA64 memory/cache benchmark to test and tune your settings.

Lochekey · Sep 13, 2015

Woomack said:
Welcome

You probably won't make it pass XTU @5GHz without sub 0 cooling. Even 4.9GHz is usually impossible, even though most other tests are passing.

1. I see you have power saving options enabled, disable eist, c states and enable performance mode in windows power settings.
2. Lower VCCIN when you are not benching on sub 0. 2.3V is too high for some chips and you can see degradation in some time ( it's not a rule but may happen ).
3. If I'm right then your memory is based on Hynix so I would set 1.75-1.85V and timings like 10-12-12-30 and check how high can you set clock. Next make the same at 11-13-13-35. Usually CL10 can run at 2400-2600 while CL11 at 2600-2666. Best chips can make 2800 CL11 or 12-14-14 but lower clock at tighter timings will offer higher performance. Also use 2 memory sticks as 4 will lower max OC.
4. Try higher cache clock - about 300 MHz below CPU clock. Set RING voltage to 1.35-1.40V.

No more ideas right now.

Hello again Woomack, I took your advice and dropped my VCCIN to be safe as well as adjusted the power saving modes in my benching profiles. I took 2 sticks of ram out and tried redoing my ram settings but still cant seem to stabilize 2600 mhz even at cl12 with volts as high as 1.85. I tried mixing and matching the different modules as well as ram slots on my motherboard to no avail. For now I think I am just going to stick at 2400mhz and try to tighten up my subtimings further. once I hit a wall there I am sure I will be back after 2600 again down the road. Ill be trying the higher cache clock the next time i try a sub ambient run and I will see what results that has on my scores. Thanks again for all the help.

Lochekey · Sep 13, 2015

Johan45 said:
That helps but high voltage is still iffy with these CPUs. I would stick with 4.9 @ 1.49v unless you can get even colder you're not going to get 5.0 stable. Like woomack says XTU like memory/cache speed so work on that. Try using AIDA64 memory/cache benchmark to test and tune your settings.

I downloaded AIDA 64 a few months back and have yet to touch it. I will start trying it out to see what affects my changes to my sub timings will have on the memory. I was hoping to improve my XTU score to improve my rank in the current rookie rumble ( currently 3rd place overall) but I now don't expect I will be able to improve enough to make a significant difference. Thanks for the advice on Aida 64 though I would have never thought to look there otherwise.

Mandrake4565 · Sep 13, 2015

Oh btw Lochekey you're not going to find much sanity here, figure I'd let the cat out of the bag now.

Blaylock · Sep 14, 2015

Welcome Joe!

Lochekey · Sep 14, 2015

Blaylock said:
Welcome Joe!

Blaylock,

I gotta ask, Are you a rodeo clown? I've never been one for the bulls or broncs, but I do dabble with barrels and pole bending.

OptyTrooper · Sep 14, 2015

Mandrake4565 said:
Oh btw Lochekey you're not going to find much sanity here, figure I'd let the cat out of the bag now.

I resemble that remark.

Blaylock · Sep 14, 2015

Lol not literally. It's a term we use here to describe someone who needs to do more sanity checks. There are a few of us that throw our radiators or even the whole rig in the freezer. We hang our rads out the window in the winter. The first time I heard the phrase was in the just Putzen around thread. Linked at the bottom. Great stuff in there if you can get through all the goofing and general having fun.

I believe I saw a thread of yours where you were thawing out your dinner (I think it was you anyhow) that would qualify you as a rodeo clown lol. (AKA rodeo cowboy)

http://www.overclockers.com/forums/showthread.php/731320-Just-Putzen-Round?p=7857861#post7857861

Mandrake4565 · Sep 14, 2015

OptyTrooper said:
I resemble that remark.

Exactly Opty!

ATMINSIDE · Sep 14, 2015

XTU depends heavily on how good of a piece of silicon you have.
Make sure you're not throttling based on the load meters they show on the benchmark.

Tech Tweaker · Sep 14, 2015

Blaylock said:
Lol not literally. It's a term we use here to describe someone who needs to do more sanity checks. There are a few of us that throw our radiators or even the whole rig in the freezer. We hang our rads out the window in the winter.

A few times I've been benching in the middle of Winter with the window wide open and a box-style fan stuck in it sucking cold air into the room. Got down to 50°F a few times. Sitting there waiting on benches to finish while wearing a winter coat, gloves, and a hat. Because it was really bloody cold. I didn't have a water cooling system though, so no radiator to put outside, and couldn't take the whole rig outside because it was snowing, but I could bring cold air inside. I was trying what I could to beat that fellas score though, I was at my voltage limit for ambient, and the only solution was to lower the ambient temperature in an attempt to get a higher core frequency.

Some of us are just on the edge of insanity. When the competitive bug hits you, sometimes you'll do some strange things to get better scores.

I do however think that 1.49v vCore is a little high for that CPU though.

Mandrake4565 · Sep 14, 2015

Some of us are just on the edge of insanity. When the competitive bug hits you, sometimes you'll do some strange things to get better scores.

You mean things like this TT?

Lochekey · Sep 15, 2015

Hey guys no thermal throttling during this run peak temp hit 58 with it normally doing high 40s under load. I normally don't exceed 1.4v but I had to give it a shot too see what I could do. Yes, that was my thread about defrosting food, I have included the pic below for anyone who missed it.

Johan45 · Sep 15, 2015

Nice stuff, so let's see some results.
Just had a look at you scores and it seems that XTU is really pulling you down. There are some pointers for the other benches if you're interested. First is to open TaskManager and kill any unnecessary processes and then set the benchmark to high priority. Also kill explorer.exe. With GPUpi try different combinations of batch reduction sizes. They'll all yeild different results.

Blaylock · Sep 15, 2015

Lochekey, if I may offer a small suggestion. You should try turning the fan over your ram around. I've gotten better results blowing air onto the ram and VRM sinks than trying to blow air away from them. It's minor, but sometimes every little bit helps.

Good stuff, Welcome to the Rodeo.

Woomack · Sep 15, 2015

@Lochekey, I have a theory that frozen peas would be best ...

Need a sanity check

Senior Pink Member

Benching Team Leader

Senior Pink Member

Mr. Clean Senior Member

Benching Team Leader Super Moderator

Senior Pink Member

Senior Pink Member

Mr. Clean Senior Member

"That Backfired" Senior Member

Senior Pink Member

Member

"That Backfired" Senior Member

Mr. Clean Senior Member

Sim Racing Aficionado Co-Owner

Member

Mr. Clean Senior Member

Senior Pink Member

Benching Team Leader Super Moderator

"That Backfired" Senior Member

Benching Team Leader

Similar threads