dude123

Hi, I also experienced the 1.18 firmware issue. What helped me was connecting the AC power supply in addition to the WD19 dockstation i'm working with. Generally, not restoring the former performances, just making it more usable. Also noticed, fan won't kick in to full speed no matter what, and also gaming is the worst. The Nvidia is not raising above 10% and games stutter. in TS, PL2 of core is active even on idle

I also get sound stuttering here and there. Fortunately, it's not my job use case. I feel sorry for you anyway. Dell is a complete dead horse. Me also, never going to buy any of their systems, and I also regularly recommend our IT to stop the mandatory purchases from Dell.

So, I've just repasted using liquid metal.. now getting 24k at CB on single runs, repeatedly (compared to ~21k with the former paste [mx4] and with a much moderate undervolt).

Next on my list when I reach to a job-load break..

Believe me, that's exactly what I did, you can see some lines are using exactly your initial values. That excel is just an attempt to figure out if there are any parameters that can make a performance bump to the levels you reported. Moreover, I really tried to make the system run as long as possible before the PL1 limit breaks in, as I find it the most performance limiting issue. Couldn't make it happen. Of course I can use a very humble uv values, but that's not going to show me a significant improvement. And just to remind what I've said earlier, the crash can come after 1-2 days, sometimes causing me to loose ongoing work. I'm now facing issues with some king of "half" idles. Running a long task, the system suddenly throttles an application i'm using to 50% cpu utilization when the screen goes off... but as I said I'm not sure if that's the application or the OS or the Dell.

here are my attempts. TS_tracking_table.xlsx

Yes I followed your results, thanks. Thing is, even though undervolting did give me a performance bump in CB23 results, a custom benchmark (with operations and programs I use for my work, uses only CPU, and with different scenarios taken from my everyday heavy load jobs) has showed no significant improvement. never mind. I'm going to stick to stability. undervolting gives me random BOSD after different durations of operation. Decreasing the uv to a "safe" level will decrease the performances, and what do I get from that. I had a ~3 day run, PC bsod after 48 hrs. urrrrr. Also have issues with no uv. PC is "napping", programs crash. Of course this could be application related, but adding the uv layer means more uncertainty. It's going to be a desktop next time 🤣 cheers.

nm just thought it can be another reference for my current system.

Can you share the script ?

Could be. I'll get my liquid metal and try again. Ok so now I feel I can load on some TS questions 🤣 Do you make the increments to both cpu-core and cpu-p-cache at the same time - i.e. step by step side by side, or your moving one of them to the limit, move back, and incrementing the other? Are you hitting PL1 at the stable state? can you share your TPL configuration for this system What do you mean by locking? just changing the FIVR Turbo Groups values ? My multipliers never reach 47 (FIDs - correct?). TS Bench puts them around 42, then about 38-39 after some time. CB load brings them to about 44 at the very beginning, then scales down to about 33. Thanks

wOW thanks for collecting all this.. to be honest, i'm very skeptical that the power plans have any significant effect, as in my case, the medium speed fans and the high load "steady- state" reduce the effective power x10 times. But I'll give it a try. How would you estimate the possible improvement with CB23 results. Lets say i'm on 19.5k.. As of TrottleStop. I find it very unsteady. Sometimes it looks like its doing something, sometimes not. For example, I've had a very good setup, in terms of performances (relative to my PC). After sometime it gave me a BOSD during low load. Also changing the turbo groups ratio seems to have no effect sometimes (randomly). Anyway, I tried to go back to the same parameters, but this time the performance are not that good. Very frustrating.

Couldn't make Ultimate power plan appear. Used the CMD commands suggested online but it just does appear in powercfg /L nor in windows power plan list. I can see the line in the the registry though. Searched for setting a csEnabled flag in the reg to 1 but the csEnabled key doesn't exist anywhere. About s3 state - those keys doesn't exist: [HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Power] "PlatformAoAcOverride"=dword:00000000 "CsEnabled"=dword:00000000 nor under ControlSet001. The powercfg /a shows: think this needs to be unlocked through usb boot C:\Windows\System32>powercfg /a The following sleep states are available on this system: Standby (S0 Low Power Idle) Network Connected Hibernate The following sleep states are not available on this system: Standby (S1) The system firmware does not support this standby state. This standby state is disabled when S0 low power idle is supported. Standby (S2) The system firmware does not support this standby state. This standby state is disabled when S0 low power idle is supported. Standby (S3) The system firmware does not support this standby state. This standby state is disabled when S0 low power idle is supported. Hybrid Sleep Standby (S3) is not available. Fast Startup This action is disabled in the current system policy. I'm near 20k. That's sad. If this was personal machine I would have thrown it back already. Can't understand how Dell is delivering those. That's more of my corporate's problem as they insist buying from this vendor. Eventually they get non productive employees. Thanks 🙂

All right - many thanks for the virtualization tip! So my first attempts starting with mpc8mb recommended values + few changes bring be a bit below 20k on CB23. That's a good improvement. But not near the top. I'll dig in to work my way up. Yet the most annoying thing is that "steady state" level of cpu utilization. So PKG power is about 78W and fans are medium. Which mechanism governs this?

Hi, Just found this thread after a pretty frustrating sessions with my (work) 7670, 12950HX equipped. So initially I have a system with the latest BIOS (1.8) and the better fans (S-something brand). First results from CB23 are 17k, which is better than others were reporting. I noticed that strange thermal throttling behavior reported on earlier posts - CB multi processor session is started, with what seems like a full-load - PKG power peaks to 157W and stays about 120-130W. During this time the fans kick-in. The first CB pass is finished and shows a predicted score of ~ 20+k. It takes about ~ 1 min in this state and then the machine drops performances and stays at about 75W PKG power. Later CB passes are slower and the final score is 16-17k depending on (I don't now). I used to have a 9880H before (7650 ?), and from what I remembered during heavy load the fans were always working, loud and fast. This isn't the case with the 7670. It looks like the system decides it doesn't want to work hard. Fans become medium speed and as I said, power remains at 75W. Trying to break through that barrier, I've done couple of things as recommend along the thread. First was repasting, and then enabling CFG and overclock using the methods in OP. Set ultimate performances plan in the BIOS. So the situation now is I'm trying to use ThrottleStop to make it work harder. Problems are: TS doesn't seem to change anything. At least nothing noticeable. I've used @MyPC8MyBrain recommendations for TS as initial values. It looks like the OC multipliers (47 for P cores) are ignored ( do I check it correctly: watching the FID values). The CPU core undervolt is -145mV. Increased from 125 - I don't see any difference. The most dominant performances drop is clearly because of the "1 min drop". Tried to change PL values under TPR, but this also has no clear effect. I will be glad for any recommendations with this. Thank you