Is my 5950x a dud?

ChocAndGray · 12 Feb 2022 at 08:25

wookiee87 said:
i upgraded to the asus maximus extreme x570 and now have a all core 4.7ghz core clock on my 5950x, i still run into problems if i try and push more, the all core clock of the cpu anyway is 4.9ghz with single core boost upto 5ghz, to get a all core 5ghz you need some massive cooling running, the cpu stock will drop clocks anyway to around 4.1-4.2ghz under full load.

dont feel too bad if your cpu is stuck at 4.7ghz, a all core oc at that still nets huge performance vs stock, the 5ghz is nice but that is only 1 core maybe 2 if you get a good chip, and if temps are too high it will drop back.

I wish it was an all core. The biggest all core I can get is 4.5

I was talking about single core boosts. Previously every core could hit 4.9ghz now they really struggle

wookiee87 · 12 Feb 2022 at 08:36

Jay343434 said:
I wish it was an all core. The biggest all core I can get is 4.5

I was talking about single core boosts. Previously every core could hit 4.9ghz now they really struggle

how long have you had the cpu and at what voltages have you been running it at?, just for context when i first gpt my cpu i dropped back the volts from auto to around 1.25-1.3v and after some tweaking i managed 4.7ghz at 1.272v on my current motherboard, i set llc to 5 as well to maintain stability, my cpu has a good imc too and i have oc my ram to 3733mhz at cl 14,15,15,36, (from 3600mhz at stock speed) but every cpu will be diffrent, as long as you can stay below 80 degrees in stress tets you will be ok for 24/7 use.

Dallies · 12 Feb 2022 at 10:08

If I remember rightly I set the profile to docp standard and set core performance boost to enable. Everything else was on auto, apart from ram speed. The ai tuner did the rest. I think the core performance boost controls amd precision overdrive. But as I said I wouldn't recommend it. And as already mentioned you need good cooling! Otherwise the cpu will keep throttling.

My cpu was rated at 105w but with all core boost it was close to 200w power draw!

wookiee87 · 12 Feb 2022 at 10:32

Dallies said:
My cpu was rated at 105w but with all core boost it was close to 200w power draw!

yeah my all core 4.7ghz in games can pull anywhere from 145-175w for the cpu alone, good job i have 2*420mm rads keeping things nice and cool, temps range from mid 40's upto around 75 peak depending which games i play, if i run any kind of benchmark i see easy 200 watts for the cpu which is very high, a rule of thumb for amd overclocking never double the rated tdp of the cpu, its not smart and will degrade the cpu over time.

i did my benchmark tests and called it a day, now i just use my pc for games and the oc on my cpu is totally fine

Grim5 · 12 Feb 2022 at 10:47

Dallies said:
If I remember rightly I set the profile to docp standard and set core performance boost to enable. Everything else was on auto, apart from ram speed. The ai tuner did the rest. I think the core performance boost controls amd precision overdrive. But as I said I wouldn't recommend it. And as already mentioned you need good cooling! Otherwise the cpu will keep throttling.

My cpu was rated at 105w but with all core boost it was close to 200w power draw!

What does core performance boost even do? I tried it and it runs exactly same on and off and Cinebench r23 score is exactly the same too

LeMson · 12 Feb 2022 at 10:49

What a carry on this thread is lol

Dallies · 12 Feb 2022 at 12:17

Grim5 said:
What does core performance boost even do? I tried it and it runs exactly same on and off and Cinebench r23 score is exactly the same too

According to my old manual core performance boost and precision boost overdrive need to be enabled for turbo boost to work. It's amd, complicated! With my Intel rig I just enable multi core performance boost and it boosts to the max on all cores.

Curlyriff · 12 Feb 2022 at 18:14

Guest2 said:
Hard to say without someone testing!
If 1.265V is stable at 4.7 all core, 1.43 should take it quite a bit higher if you can get rid if the heat.
Someone with good cooling and deep pockets should test for us in the name of science. Sure someone will eventually and youtube it

My 5950x chip only gets to 4.6Ghz at 1.43v with full WC loop so you doing very well tbh with that as far as I can tell. If I try to push anything beyond 4.625Ghz it is insta crash.

Grim5 · 12 Feb 2022 at 18:31

Curlyriff said:
My 5950x chip only gets to 4.6Ghz at 1.43v with full WC loop so you doing very well tbh with that as far as I can tell. If I try to push anything beyond 4.625Ghz it is insta crash.

You need to decouple the two ccx's. On the 5950x the first 8 cores are binned higher than the second 8, by quite a bit actually.

You can probably do 4.8/4.9 on the first 8 cores and 4.6 on the second. That was the case for me, I set mine to 1.3v and then it also started crashing above 4.6 so I decoupled the ccx's and set the first one to 4.8 and left the 2nd on 4.6 and it was stable

Cyber-Mav · 12 Feb 2022 at 19:53

Heofz said:
hmm yeah see, here's a screenshot from @bongo78 on his thread:

all of his cores are able to boost past 4.9 Ghz no problem it seems :/

wont be a simultaneous high clock on all cores. maybe each core took turns to hit that high point. or workload must be mega light otherwise he gonna hit the tdp limit.

Guest2 · 13 Feb 2022 at 09:47

Curlyriff said:
My 5950x chip only gets to 4.6Ghz at 1.43v with full WC loop so you doing very well tbh with that as far as I can tell. If I try to push anything beyond 4.625Ghz it is insta crash.

I gave up with clocking my 5800X and left it at default PBO settings. It just wasnt worth clocking it.

RAM is all I have overclocked now. 3600MHz C16 1800IF @3800MHz C14 1900IF. 3080FE and 5800X at stock.

I push GPU a bit higher for Timespy but thats only benching with max GPU and case fans - https://www.3dmark.com/spy/25895217
Similar to running CPU on the edge for benching. Everyday, not worth it

ChocAndGray · 13 Feb 2022 at 09:59

Guest2 said:
I gave up with clocking my 5800X and left it at default PBO settings. It just wasnt worth clocking it.

RAM is all I have overclocked now. 3600MHz C16 1800IF @3800MHz C14 1900IF. 3080FE and 5800X at stock.

Yeah I'm rapidly coming to that same conclusion. I honestly can't notice any tangible differences with how I use my PC gaming and mining.

It seems the gains you could get on hardware when overclocking are nearly long gone.

It's a psychological thing for me though, I like seeing high round numbers lol.

Kirby Wurm · 13 Feb 2022 at 10:35

Jay343434 said:
It seems the gains you could get on hardware when overclocking are nearly long gone.

Pretty much. I've set my 5900X to stock again as OC just wasn't worth it for the under 5% CB23 improvement and literally none in games. My GPU can add another 8% performance with an undervolt and OC, but nothing like the 10-20% OCs I've seen in past.

My RAM is tuned pretty hard though, and that makes a tangible difference (3800MHz CL14 with 1900 IF).

ChocAndGray · 13 Feb 2022 at 10:38

Kirby Wurm said:
Pretty much. I've set my 5900X to stock again as OC just wasn't worth it for the under 5% CB24 improvement and literally none in games. My GPU can add another 8% performance with an undervolt and OC, bit nothing like the 10-20% OCs I've seen in past.

My RAM is tuned pretty hard though, and that makes a tangible difference (3800MHz CL14 with 1900 IF).

Interesting that you mention the ram. I'm running at stock xmp/docp timings as they're pretty tight b-die anyway but I need to do some testing in game benchmarks and see what difference it makes.

Curlyriff · 13 Feb 2022 at 11:47

Guest2 said:
I gave up with clocking my 5800X and left it at default PBO settings. It just wasnt worth clocking it.

RAM is all I have overclocked now. 3600MHz C16 1800IF @3800MHz C14 1900IF. 3080FE and 5800X at stock.

I push GPU a bit higher for Timespy but thats only benching with max GPU and case fans - https://www.3dmark.com/spy/25895217
Similar to running CPU on the edge for benching. Everyday, not worth it

Oh yeah that's what I have ended up doing in leaving PBO on. It seems to be best balance tbh for workload and gaming anyways. The benching was just to see/for fun really.

Curlyriff · 13 Feb 2022 at 11:49

Grim5 said:
You need to decouple the two ccx's. On the 5950x the first 8 cores are binned higher than the second 8, by quite a bit actually.

You can probably do 4.8/4.9 on the first 8 cores and 4.6 on the second. That was the case for me, I set mine to 1.3v and then it also started crashing above 4.6 so I decoupled the ccx's and set the first one to 4.8 and left the 2nd on 4.6 and it was stable

Interesting, will that a go to see how it does with that kinda setup later. Cheers.

Kirby Wurm · 13 Feb 2022 at 12:22

Jay343434 said:
Interesting that you mention the ram. I'm running at stock xmp/docp timings as they're pretty tight b-die anyway but I need to do some testing in game benchmarks and see what difference it makes.

I found minimum frame rates improved noticeably in some games and my TimeSpy also improved substantially. It won't be substantial across all games but well worth it for some that are CPU bound and memory intensive.

ChocAndGray · 13 Feb 2022 at 12:29

Kirby Wurm said:
I found minimum frame rates improved noticeably in some games and my TimeSpy also improved substantially. It won't be substantial across all games but well worth it for some that are CPU bound and memory intensive.

Which do you think makes more difference higher frequency or tighter timings?

Kirby Wurm · 13 Feb 2022 at 12:42

Jay343434 said:
Which do you think makes more difference higher frequency or tighter timings?

A bit of both but timings seem to have the biggest impact. Going from loose CL16 to tighter CL16 gave a 10% improvement in some benchmarks, and then a further improvement when I managed to get to CL14. I'm no expert on this and relied on copying others' settings but it seems a worthwhile exercise. Again, it very much depends on the game and whether you're GPU bound. This article shows that well:

https://www.techspot.com/article/2140-ryzen-5000-memory-performance/#DDR4_Frequency_Scaling

ChocAndGray · 13 Feb 2022 at 12:43

Kirby Wurm said:
A bit of both but timings seem to have the biggest impact. Going from loose CL16 to tighter CL16 gave a 10% improvement in some benchmarks, and then a further improvement when I managed to get to CL14. I'm no expert on any of this but it seems a worthwhile exercise. Again, it very much depends on the game and whether you're GPU bound. This article shows that well:

https://www.techspot.com/article/2140-ryzen-5000-memory-performance/#DDR4_Frequency_Scaling

Thanks I'll have a read