dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2021/01/16 14:36:02
(permalink)
Seems the enemy to overclocking is the avx stuff. I can sit a 5ghz with healthy voltage and under 100f temps, and
rock n roll in games that demand high core frequency, but as soon as i put in avx workload, like R20, boom, blue screen
If i run 5Ghz, would it hurt if i ran avx at a 20 offset? to have it do strock loads?
|
glocked89
Superclocked Member
- Total Posts : 187
- Reward points : 0
- Joined: 2019/01/16 23:48:00
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2021/01/16 17:25:22
(permalink)
dmann304
Seems the enemy to overclocking is the avx stuff. I can sit a 5ghz with healthy voltage and under 100f temps, and
rock n roll in games that demand high core frequency, but as soon as i put in avx workload, like R20, boom, blue screen
If i run 5Ghz, would it hurt if i ran avx at a 20 offset? to have it do strock loads?
oh who cares about being able to do high avx clocks? i just set avx2 to -5 offset and avx3 to -15 offset. none of the games i play use avx instructions anyways
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2021/01/16 18:49:46
(permalink)
That makes so much more sense, seeing avx 2 is for tile renders, and 15 would just heat stuff up.
will make 5ghz more reachable?
|
ty_ger07
Insert Custom Title Here
- Total Posts : 21173
- Reward points : 0
- Joined: 2008/04/10 23:48:15
- Location: traveler
- Status: offline
- Ribbons : 270
Re: Concensus on safe daily mesh voltage
2021/01/16 20:06:06
(permalink)
dmann304
That makes so much more sense, seeing avx 2 is for tile renders, and 15 would just heat stuff up.
will make 5ghz more reachable?
I tweaked my Intel short duration power limit, long duration power limit, and power plane amperage until I found where my system typically runs with a typical load that I care about. That way, with AVX stuff, and whatever I don't expect, it's kind of automatically limited to the power usage and heat production I am expecting. I basically started higher than I expected, and then lowered until I started to see boost drop, and that's how I knew what my typical use was. AVX offset would do the same, but there is more than one way to do things. AVX offset seems like a more limited approach. Adding power limits would restrict all future unexpected unwanted conditions.
post edited by ty_ger07 - 2021/01/17 15:10:01
ASRock Z77 • Intel Core i7 3770K • EVGA GTX 1080 • Samsung 850 Pro • Seasonic PRIME 600W Titanium
My EVGA Score: 1546 • Zero Associates Points • I don't shill
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2021/01/17 04:13:30
(permalink)
Like people said, avx stuff is a low priority, seeing most games don't use it. My latencies are pretty good, in the 50's
on par with a 10900k according to Aida 64, and in games, i don't get temps much more then 55c load, so its the avx
loads that bring it to 90c, i don't use a power plan but ultimate, cuz i prefer the snappy experience in windows 10. and
thats what 5ghz can do, besides scores in benchmarks.
|
glocked89
Superclocked Member
- Total Posts : 187
- Reward points : 0
- Joined: 2019/01/16 23:48:00
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2021/01/17 12:56:30
(permalink)
dmann304
Like people said, avx stuff is a low priority, seeing most games don't use it. My latencies are pretty good, in the 50's
on par with a 10900k according to Aida 64, and in games, i don't get temps much more then 55c load, so its the avx
loads that bring it to 90c, i don't use a power plan but ultimate, cuz i prefer the snappy experience in windows 10. and
thats what 5ghz can do, besides scores in benchmarks.
Keep in mind though Aida example benchmarks only show VERY stock setups. If you tune ram on a 10900k or any of the intel ringbus cache cpus, you'll get under 40ns easy. But yeah don't worry about avx, just play your games
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2021/01/17 14:06:23
(permalink)
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2021/01/17 16:11:02
(permalink)
ONe thing i have done, is per core on voltage, and setting ratio. Seems to give me more speed, and less
voltage, seeing i run the world threads 1,2, at 5, and rest at 48, and seems to fly so far. not sure what cores
use avx stuff, but according to hwinfo, seens the middle ones heat up.
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2021/01/19 04:24:29
(permalink)
With alot of testing, and changes, i left auto to tj max, 48 on all cores, xmp to auto, mesh at 30, 1.01v
vin at 1.88v and cpu at 1.30v, stable, temps are in 70's on load, but package hits 105c, then goes back
to 32c, idle, normal? passes all tests including intense tile renders like bmw, and classroom in blender.
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2021/01/20 13:11:20
(permalink)
From alot of threads on the web, i seen that alot of folks who do run Mesh at 30 or higher, they do a bump in voltage
on VCCIO, and Mesh Vcore, or they run them at the same voltage, like example mine, I do at 30 seeing that most
stable. Vccio 1.005v and Mesh at 1.005v. Seems those 2 effect each other? and has noting to do with ram xmp or
speeds.
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: Concensus on safe daily mesh voltage
2021/01/20 17:11:06
(permalink)
dmann304
From alot of threads on the web, i seen that alot of folks who do run Mesh at 30 or higher, they do a bump in voltage
on VCCIO, and Mesh Vcore, or they run them at the same voltage, like example mine, I do at 30 seeing that most
stable. Vccio 1.005v and Mesh at 1.005v. Seems those 2 effect each other? and has noting to do with ram xmp or
speeds.
FWIW I'm running in 'per core' mode with mesh at auto (which ends up 30) and all voltages left at auto (because I'm not pushing CPU multipliers beyond factory values) and I've passed many hours of stability tests without any issues.
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2021/01/20 20:55:26
(permalink)
Wow, of course its gonna pass, if your not pushing beyond factory, So the MESH will run at 30? with per core setting? Also there a couple pro's on here that suggest to not go above 1.116v on Mesh on Cascade Lake X, and keep the VCCIO under 1.115 also.
post edited by dmann304 - 2021/01/20 22:24:21
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2023/12/20 12:42:17
(permalink)
Is 1.125v on VCCIO, to low? For all core oc of 4800? and seems when lowering mesh v to 1.25, i get red marks on corsair IQUE
in memory tab. strange.
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: Concensus on safe daily mesh voltage
2023/12/20 13:18:26
(permalink)
☄ Helpfulby Cool GTX 2024/01/17 09:55:16
dmann304Wow, of course its gonna pass, if your not pushing beyond factory, So the MESH will run at 30? with per core setting? 30x mesh is not factory, factory is 24x. I use 'per core', have set all of my cores to 43x (all cores 43x is not factory either for 10920X), AVX2 offset 3, AVX3 offset 5, and use default voltages. I can pass non-AVX workloads without problems at multiplier higher than 43x but moment there is an AVX2/3 workload I end up with issues. I might be able to address them by raising offsets (moment AVX2 exceeds 40x / AVX3 exceeds 38x temperatures start becoming too much for what my cooling can handle and my taste) but then I am making one step back performance wise so I left it at this as sweet spot, otherwise it is too much pain for too little gain. With these settings, mesh at 30x, and Vmesh at default voltage I am able to pass tests for hours on the end but then I have intermittent NVMe boot issue I mentioned in other thread. Besides, have you ever seen proof that mesh overclocking results in real world gains that are worth the effort? I haven't.
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: Concensus on safe daily mesh voltage
2023/12/20 13:21:37
(permalink)
dmann304Is 1.125v on VCCIO, to low? For all core oc of 4800? and seems when lowering mesh v to 1.25, i get red marks on corsair IQUE in memory tab. Have you tried leaving voltages at 'auto' and setting all core 48X and mesh to 30x? You might want to try that as those multipliers still seem to be in Intel's tables.
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2023/12/21 16:40:45
(permalink)
I have that chip, and seems that it likes 1.310 to pass cinabench runs, and even had to lower cpu speed to 4800 from 5ghz,
cuz of weird stuff going on which i thought was mesh, or memory. It seems happy with 1.25 on mesh too, and 3800 on memroy
at cl16, so i might have lost a bit of life on the cpu, cuz it was stable at 5 ghz for 2 years with 1.4v.
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: Concensus on safe daily mesh voltage
2023/12/21 17:52:35
(permalink)
dmann304
I have that chip, and seems that it likes 1.310 to pass cinabench runs, and even had to lower cpu speed to 4800 from 5ghz,
cuz of weird stuff going on which i thought was mesh, or memory. It seems happy with 1.25 on mesh too, and 3800 on memroy
at cl16, so i might have lost a bit of life on the cpu, cuz it was stable at 5 ghz for 2 years with 1.4v.
It is possible you lost a little over period of time due to "silicone deterioration". It is also possible something else is causing it, like ghost / temperature throttling that wasn't there before (paste is drier after 2 years), changes in underlying code that are now using different path, or any number of other reasons, one or more at the time. That is why I personally don't go for highest possible figures at the beginning of life, instead I back off and settle for those that will keep rocking solid at the end of the life too. So, are you aware of any real world proof overclocking of mesh is worth the headache? BTW, last time I tried CineBench it wasn't stressing my system as much as various video encoders did. Also, I like to stress test with OCBase OCCT.
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2024/01/17 09:39:50
(permalink)
Seems alot of the guilt of these OC's , is the IMC, or super tight timings on dominaotr Platium and at 3800, single rank not B-di, but Micron.
Seems its very happy at 4800, and if i push 4900mhz, seems bendhmakrs suffer a bit, and this is on recent paste on Corsair 420 AIO.
And mesh seems happy at 3200 now aswell. Was thinking of grabbing a 10980xe, but they are too costly, for what you can pay for 14th gen.
Right now, things are very stable, and alot of testing, and workloads, seem to be good.
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: Concensus on safe daily mesh voltage
2024/01/18 19:15:18
(permalink)
dmann304Right now, things are very stable, and alot of testing, and workloads, seem to be good. That is great news! It would be very kind of you if you share screenshots of your settings for the benefit of the community, please :)
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2024/01/19 13:28:07
(permalink)
I see folks here, show uefi settings from screenshots, but how is this done?
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: Concensus on safe daily mesh voltage
2024/01/19 14:13:56
(permalink)
dmann304I see folks here, show uefi settings from screenshots, but how is this done? IIRC by pressing F12 when in BIOS screenshot will get saved to FAT32 formatted USB stick (see Can't Get BIOS/UEFI Screen Capture to Work - EVGA Forums). If that isn't working for you you can always take pictures with your cell phone and share that.
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2024/01/30 17:41:08
(permalink)
My UEFI settings are:
all core 49x
avx offset 2
mesh 32
vin 1.815
vcore 1.360
mesh v 1.25
vccio 1.2v
uncore +300
memeory 3800 cl 16
tight secondaries:
32767
587
24
12
8
These are what i have atm and passes all cinebench runs. and good score, but at 5ghz, its slower in
tests for some reason , which could be diminishing returns. on a 10920x but will go to a 10980xe which i
know has same single core speeds, but will do better in all core, hopefully at 48x or 49 on a corsair aio 420
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: Concensus on safe daily mesh voltage
2024/01/30 18:52:53
(permalink)
dmann304These are what i have atm and passes all cinebench runs. and good score, but at 5ghz, its slower in
tests for some reason ... Thank you for sharing! Are you able to pass OCCT tests at these settings?
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2024/01/31 08:29:36
(permalink)
Fairly easily. I do run 5 min variables on the different options like avx, auto cache, memory ext. however as solid
as these settings are, there occasions when running R20, the test will run great from tinme to time to see if
this remain stable, however their times when it will just stop or hang, then run again, and is good, but with
these settings if i bump the cores up a bit, stays stable, but lose points on R20 or geekbench 5 for some reasons
Jay sais you can lose performance past 48 or 49 on cascade lake x.
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: Concensus on safe daily mesh voltage
2024/01/31 11:07:15
(permalink)
dmann304
Fairly easily. I do run 5 min variables on the different options like avx, auto cache, memory ext. however as solid
as these settings are, there occasions when running R20, the test will run great from tinme to time to see if
this remain stable, however their times when it will just stop or hang, then run again, and is good, but with
these settings if i bump the cores up a bit, stays stable, but lose points on R20 or geekbench 5 for some reasons
Jay sais you can lose performance past 48 or 49 on cascade lake x.
Are you having same kind of behavior (performance loss as you increase multiplier past certain point) in other software's too, like Cinebench R23? If yes that might be a sign of your CPU telling you that you are pushing it too hard, there might be some internal throttling, and to back off a bit. What are your CPU temperatures like?
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2024/02/02 15:30:59
(permalink)
When i was at 48x, it did around 84c, and at 49x around 86c which is not a big deal, but single core went up
from 468 in r20 to 495, but all core when from 6850 to 6550 which is strange, like multicore went down some, but
single went up and temps were under 90c, and can do 105c on tjmax. Its like the chip finds its sweet spot at 48, but
\temps are ok I heard stories about this on X299 when mesh is higher then 32, you lose performance.
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: Concensus on safe daily mesh voltage
2024/02/02 17:41:12
(permalink)
dmann304
When i was at 48x, it did around 84c, and at 49x around 86c which is not a big deal, but single core went up
from 468 in r20 to 495, but all core when from 6850 to 6550 which is strange, like multicore went down some, but
single went up and temps were under 90c, and can do 105c on tjmax. Its like the chip finds its sweet spot at 48, but
\temps are ok I heard stories about this on X299 when mesh is higher then 32, you lose performance.
TJMax for 10920X is 94C, not 105C. Due to bug you must manually set it in Dark's BIOS to 94, you should not leave it on Auto. If you are setting it to 105, or leaving it at Auto and it thinks it should be 105, then you are cooking your CPU. If your cores are hitting 90 then you are dangerously close to 94 and it is likely package is hitting above 94 and CPU is internally throttling down which could explain your drop in multi-core scores. It makes sense that 10920X starts dropping off past 48x, 48x is highest multiplier Intel has baked in for that chip. In other words if I were you I would stay on the safe side, leave it at 48x and be done with it. Few percent single-core score in few synthetic benchmarks is not worth the hassle nor it guarantees real world apps will show that gain too.
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2024/02/03 08:15:05
(permalink)
That makes alot of sense. I am getting the 10980xe in a couple of weeks dor a deal, and i heard from Roman Durbour, that 48x is really that
baked in setting most will get clean on these high core count chips. and i did set it to 94c. on tjmax, but when the xe comes i will set it to its
tjmax.
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: Concensus on safe daily mesh voltage
2024/02/03 11:32:22
(permalink)
dmann304I am getting the 10980xe in a couple of weeks for a deal I wish you tons of happiness with your "new toy" :)
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: Concensus on safe daily mesh voltage
2024/02/04 13:38:32
(permalink)
Is the chip still good in 24? I just wanna max out my dark x299 system. And not seeing a reason
to bother with 14gen right now. I heard with these XE's if you get latency down, and good 24 7 oc with mesh, you
got a solid system.
|