X79 Overclocking Guide

LtMatt · 22 Aug 2014 at 15:14

whyscotty said:
pfft - 1.55v or go home

But I'm on air Lol. Max temp so far 78c running IBT very high.

Nickolp1974 · 22 Aug 2014 at 15:23

LtMatt said:
But I'm on air Lol. Max temp so far 78c running IBT very high.

christ, stay where you are then!!!

make sure speedstep is on and c1e, will give the cpu a little breather whilst loading up next test!!

LtMatt · 22 Aug 2014 at 15:25

I think the H110 is coping pretty well considering, certainly with the Noctua Industrial fans doing a good job.

whyscotty · 22 Aug 2014 at 15:27

LtMatt said:
But I'm on air Lol. Max temp so far 78c running IBT very high.

Bin IBT - use realbench as a guide

LtMatt · 22 Aug 2014 at 15:28

whyscotty said:
Bin IBT - use realbench as a guide

What settings to use in real bench m8? Happy to bin IBT off, it takes ages but i do know it's stable if it passes IBT.

scotsdragon · 22 Aug 2014 at 15:32

On real bench just use the benchmark if it passes that it is game stable (usually), if you want to do more set the stress test to 15 mins and set ram usage to full

whyscotty · 22 Aug 2014 at 15:34

I just use the benchmark.

Never had a crash if it completed the benchmark

LtMatt · 22 Aug 2014 at 15:37

Cheers lads.

Ok Firestrike @4.8ghz nick, 17,300 Physics. Might be able to improve that a few more points when i bench proper. Not sure if this is good or not.

EDIT

Woot 4.8ghz appears stable, or 10 runs of IBT stable. I normally do 20, as sometimes it can fail on or before run 16, but not risking it with this chip at these voltages on air. Now ill bench some firestrike and 3dmark and dirt showdown and post some scores.

Nickolp1974 · 22 Aug 2014 at 15:47

I'd say thats good matt, i get around the same with my tridents in with cpu at the same clock, with these ripjaws running dual channel i'm getting around 18,000

LtMatt · 22 Aug 2014 at 15:58

So you get less for quad channel? I just ran it and scored 17,281 exactly. I will have another go and see if it goes up. My gpu score has gone down a few points though, not a lot but a few.

Silent_Scone · 22 Aug 2014 at 16:03

LOL

Stop meowing it

1.48v for bench runs is probably the max you want to push IvyE on warm water, not Sandy! Sandy is good for a few notches more

http://www.3dmark.com/3dm/2680262?

for comparison, same ram as you, but Ivye also has the OPS advantage. I'm not sure what timings I was using for that run, but the VDIMM would have been no more than 1.7v

I'm pretty sure it was a tight c92400

LtMatt · 22 Aug 2014 at 16:05

Lol i like to be careful, it's my first time. With my 2700k i stopped caring after a while. I'm not at that stage yet with this chip.

LtMatt · 22 Aug 2014 at 16:08

Nickolp1974 said:
I'd say thats good matt, i get around the same with my tridents in with cpu at the same clock, with these ripjaws running dual channel i'm getting around 18,000

LtMatt said:
So you get less for quad channel? I just ran it and scored 17,281 exactly. I will have another go and see if it goes up. My gpu score has gone down a few points though, not a lot but a few.

I tweaked a timing and scored 17,341. Will run it once more time and then move on to extreme.

Silent_Scone · 22 Aug 2014 at 16:10

I can't beat my previous Physics score with the replacement Dom Plats. Score around 17,900 now

LtMatt said:
Lol i like to be careful, it's my first time.

Don't you'll give me a semi

LtMatt · 22 Aug 2014 at 16:11

I think i will be looking at around 17,500 maximum, maybe a bit higher if i get a good run @4.9ghz. I do wonder actually if i used all three red slots rather than the black one if i would get a slightly better physics and combined score. That was how it worked on my P8Z77 WS board with a PLX chip. Using the PLX lanes only got me a boost, at the expense of a slightly lower gpu score. Vice versa for using the cpu pci-e lanes.

Silent_Scone · 22 Aug 2014 at 16:15

I'd be happy with what you're getting for the clocks used. Kaaps Physics IIRC is around that at 5ghz. The fact you're coming in close to mine at practically the same clocks goes to show how you've timed up those sticks. The difference between our scores will be the higher OPS on my chip clock for clock

LtMatt · 22 Aug 2014 at 16:18

MjFrosty said:
I'd be happy with what you're getting for the clocks used. Kaaps Physics IIRC is around that at 5ghz. The fact you're coming in close to mine at practically the same clocks goes to show how you've timed up those sticks. The difference between our scores will be the higher OPS on my chip clock for clock

I was incredibly anal about the timings to be fair so it's nice to see some hard work paying off.

Silent_Scone · 22 Aug 2014 at 16:21

Yeah FS is a good test for memory efficiency. Have you checked what bandwidth you're getting?

LtMatt · 22 Aug 2014 at 16:29

Got my score up a bit more with another two tweaks. The timing i mentioned earlier i can't touch or i get instability, well for 4.8 i had to increase that another notch to 6. I forget the name but i mentioned it earlier. Lowered two other ones either side of it a notch this time.

4.8ghz
2400mhz
290's 975/1250

MjFrosty said:
Yeah FS is a good test for memory efficiency. Have you checked what bandwidth you're getting?

Yeah i did, i scored a bit lower than you but not far off. Will rerun and post it later. I beat you in one or two of the cpu cache things just.

Kaapstad · 22 Aug 2014 at 16:43

LtMatt said:
I think i will be looking at around 17,500 maximum, maybe a bit higher if i get a good run @4.9ghz. I do wonder actually if i used all three red slots rather than the black one if i would get a slightly better physics and combined score. That was how it worked on my P8Z77 WS board with a PLX chip. Using the PLX lanes only got me a boost, at the expense of a slightly lower gpu score. Vice versa for using the cpu pci-e lanes.

Possible on the combined but only a tiny fraction

I think the cards would run X16/X8/X16 instead of X16/X8/X8 (I am not 100% sure about this).

Even then the difference would be minute and not worth bothering with.