Aaron44126

(Not quite sure where to put this topic.) Just saw this article which has NVIDIA "promising" non-trivial performance increases in various DirectX 12 games in their new 522.xx drivers. I wonder if there is any merit to it? Assassin’s Creed Valhalla: up to 25 percent (1080p) Battlefield 2042: up to 7 percent (1080p) Borderlands 3: Up to 8 percent (1080p) Call of Duty: Vanguard: up to 12 percent (4K) Control: up to 6 percent (4K) Cyberpunk 2077: up to 20 percent (1080p) F1 22: up to 17 percent (4K) Far Cry 6: up to 5 percent (1440p) Forza Horizon 5: up to 8 percent (1080P) Horizon Zero Dawn: Complete Edition: up to 8 percent (4k) Red Dead Redemption 2: up to 7 percent (1080p) Shadow of the Tomb Raider: up to 5 percent (1080p) Tom Clancy’s The Division 2: up to 5 percent (1080p) Watch Dogs: Legion: up to 9 percent (1440p) https://www.theverge.com/2022/10/12/23400430/nvidia-directx-12-performance-improvements-cyberpunk-2077-assassins-creed-valhalla https://www.nvidia.com/en-us/geforce/news/geforce-rtx-4090-game-ready-driver/ Driver links (DCH) Notebook (GeForce / game ready) Notebook (Studio) Desktop (GeForce / game ready) Desktop (Studio) Enterprise (Quadro / RTX) notebook + desktop

Not sure which Alienware system you are referring to specifically. None of these tools work with systems 11th gen (Tiger Lake / Rocket Lake) and later. I thought that Dell had released their own fan control tooling for newer Alienware systems, as part of Alienware Control Center? (I haven't ever tried it myself.)

Eh. I'd be pretty interested if you are able to change one of these values and produce a change in fan behavior. But, I don't think that changing these values is going to do anything. (I'd love to be proved wrong.) My understanding after reading around (+ work I have done on Dell fan control in the past) is that these values are stock "simple" ACPI fan control implementation. Dell is not using this ACPI implementation (which relies on some level of OS support), they're doing their own thing with the embedded controller (which works entirely independently of the OS). I think the key to controlling it, if there is any, will be figuring out how to send commands to the embedded controller via the SMBIOS WMI interface to change its behavior. I thought I might point you to another place where there is a pile of BIOS values that can be changed programmatically. https://github.com/dell/libsmbios/blob/master/doc/token_list.csv This is a list of "tokens" that can be used to change BIOS settings in the system via the SMBIOS WMI interface. Many of these settings are also changeable from the BIOS setup menu itself, but not all of them. Many of them don't apply to all systems. I have previously written a simple console application that can read and write any of these values from Windows. On Precision 7X60/7X70, you can do "live" things like manipulate the keyboard backlight (02EA/02EB/02EC/02F6). Anyway, there are fan control settings in there as well. (00BD, 00BE, 02FD, 02FE, 0332-0335, etc.) I was able to successfully manipulate the system fans using these settings on Precision 7510 and Precision 7530, but starting with last year's Precision 7X60 systems, this level of control does not work anymore. Those tokens do appear to be present in the newer systems' BIOS, they just don't actually don't actually control anything anymore. The "token value" in the leftmost column is only 16 bits, so you can just enumerate all possible values and discover that there are "tokens" in the systems that are not documented in this CSV from Dell. (That's how I found the current fan speed sensor values... they are actually readable using tokens from near the end of the 16-bit field space. And that's why Dell Fan Management is the only application that can read the fan speeds on these systems in realtime, even if it can't control anything right now.) Furthermore... This token list isn't the end of what you can do via the SMBIOS WMI interface. I'm also using WMI commands to change the "thermal mode" (optimized / cool / quiet / ultra performance, as mentioned above). I reverse engineered that from Dell Power Manager and the commands that need to be issued don't use these tokens at all. Anyway, all this to show that there are multiple / completely separate variable stores and the controls for the system fans (if there are any) remain yet to be found. I think the way to root it out is to probably disassemble the BIOS code that does the onboard system diagnostics and figuring out what commands they are running during the fan test. I don't really have the skills for that, but if anyone ever figures it out, I believe that I'll be able to set up tooling to have manual fan control work on both Windows and Linux.

Yes, I used a registry trick to disable it and re-enable the old power settings. (You can actually get the power settings back without disabling Modern Standby as well. There's a link in my sig to an article on "Turbo boost toggle" and it is described in there.) I make use of some of those power options. I have two Windows power profiles -- one for "quiet" and one for "performance" -- and I use Process Lasso to automatically switch between them depending on what applications I have open.

There are some things like EUROCOM Sky X7W? https://eurocom.com/ec/configure(2,334,0)SkyX7W Two 2.5" SATA bays + two M.2 + Coffee Lake 9th gen CPUs. (Not really cheap, though... You can lower the default GPU and CPU options to shave some off of the price.)

I feel like I tried this on Precision 7560 when I noticed that the fan control mechanism is different and was not able to get anything from the EC. I just fired it up on the Precision 7770, and when I hit the "EC" button, it just locked up the entire system 😕 I do think the EC is the key to the puzzle here, however, Dell's EC seems to be more sophisticated than most so it's probably a bit more complex than simple memory overwrites that you can do with other manufacturers' ECs. Fan control with prior systems involved sending reverse-engineered commands to the EC to get it to either return data (i.e. fan speeds) or change behaviors. The commands were actually reverse-engineered from an old DOS-based Dell BIOS update tool or diagnostics tool or something like that, and they worked for a long time.

Unfortunately, the BIOS interface for interacting with the system fans does not allow for setting them to a speed that specifically. (I would love it if that were the case...) You can only choose between off / medium / high. The automatic control is more granular. The whole reason I did this "consistency mode" implementation was to be able to "lock" the speed at a lower speed than "medium"... but that will only work if you can get the system to set the fans there with its automatic control first, so that they can be locked.

Yeah, 25325 is in reference to a fan speed sensor if anything (there is a "CPU temp" reference right above, and its in a section that is all about sensors and alerts). 26443 is a text string.

No, you can only pick one of them to use for each trip point. There are only four trip points (passive, active 0, active 1, critical). The "table" is like a drop-down list of options. The values mean nothing other than "here is the value you plug here if you want this temperature to be the trip point". I don't think the fan control in this laptop is implemented with IME so again I think these values are meaningless in the end (for our purposes).

Well. Not a whole lot to look at, there are only seven references to "fan" in the file. Line 25325 — I believe this is referring to a fan sensor so it is not interesting. Lines 25366, 25373, 25379 — Appear to be toggles to enable/disable each of "FAN1", "FAN2", and "FAN3" devices (...probably lifted from desktop BIOS where there could be more fans) Line 26443 — Just a text string which contains the word "fan" The others are settings for "active trip points" 0 and 1, which @MyPC8MyBrain quoted above. They are in the bit from line 25186 through 25328, labeled "Platform Thermal Configuration", which seems to be an interesting part. There are four tables interesting tables labeled "Critical trip point", "Active trip point 0", "Active trip point 1", and "Passive trip point", which have temperatures paired with increasing values. 0xB20DD One Of: Active Trip Point 0, VarStoreInfo (VarOffset/VarName): 0x698, VarStore: 0x1, QuestionId: 0xE84, Size: 1, Min: 0xF, Max 0x7F, Step: 0x0 {05 91 4B 12 4C 12 84 0E 01 00 98 06 10 10 0F 7F 00} 0xB20EE One Of Option: Disabled, Value (8 bit): 0x7F {09 07 04 00 00 00 7F} 0xB20F5 One Of Option: 15 C, Value (8 bit): 0xF {09 07 51 12 00 00 0F} 0xB20FC One Of Option: 23 C, Value (8 bit): 0x17 {09 07 52 12 00 00 17} 0xB2103 One Of Option: 31 C, Value (8 bit): 0x1F {09 07 53 12 00 00 1F} 0xB210A One Of Option: 39 C, Value (8 bit): 0x27 {09 07 54 12 00 00 27} 0xB2111 One Of Option: 47 C, Value (8 bit): 0x2F {09 07 55 12 00 00 2F} 0xB2118 One Of Option: 55 C, Value (8 bit): 0x37 {09 07 56 12 00 00 37} 0xB211F One Of Option: 63 C, Value (8 bit): 0x3F {09 07 57 12 00 00 3F} 0xB2126 One Of Option: 71 C, Value (8 bit): 0x47 (default) {09 07 58 12 30 00 47} 0xB212D One Of Option: 79 C, Value (8 bit): 0x4F {09 07 59 12 00 00 4F} 0xB2134 One Of Option: 87 C, Value (8 bit): 0x57 {09 07 5A 12 00 00 57} 0xB213B One Of Option: 95 C, Value (8 bit): 0x5F {09 07 5B 12 00 00 5F} 0xB2142 One Of Option: 103 C, Value (8 bit): 0x67 {09 07 5D 12 00 00 67} 0xB2149 One Of Option: 111 C, Value (8 bit): 0x6F {09 07 5E 12 00 00 6F} 0xB2150 One Of Option: 119 C (POR), Value (8 bit): 0x77 {09 07 5F 12 00 00 77} 0xB2157 End One Of {29 02} I thought that these might be fan curves, but they are not. It's just a list of options. I Googled around and ran across things like this documentation which clarified things for me a bit. You can select one of these options to use as the value for the various trip points mentioned above. When the system hits one of these trip points then it will set the fan speed to the value configured for that trip point. I.e., when the temperature rises above "Active Trip Point 0", set the fan speed to "Active Trip Point 0 Fan Speed". (This is standard Intel Management Engine stuff and the Dell EC/BIOS is likely completely ignoring it anyway, since it seems that Dell's fan control is a good bit more sophisticated than that.) I didn't see anything else in the dump that looked interesting with regards to fan control, so, dead end ...?

There is definitely a significant change in how Dell systems handle fan control starting with 2021 systems (Tiger Lake). The system-driven control is much more granular than before (adjusting speeds tens of RPM at a time on a curve, rather than hundreds of RPM at a time with just a few discrete values it will ever use) ... and, the established EC interface that allowed for manual control and the less documented "SMBIOS token" control mechanism for manual control have both been removed. They may well have moved the "tables" that define the behaviors to these UEFI NVRAM variables. I'm going to spend some time digging around @MyPC8MyBrain's dump and see if I can make heads or tails of anything.

I don't think Windows active/passive cooling power profile policy has any impact on the thermal/fan behavior in these systems. You can however change Dell's "thermal mode" between Optimized/Cool/Quiet/Ultra Performance, it can be set both from the BIOS and the Dell Power Manager app in Windows, or there are options to change it programmatically / by script (both on Windows and Linux). "Optimized" is the default behavior and the others each prioritize one of low surface temperature, low noise, or high performance (at the expense of the other two). I'd also love a setting to keep the fans at a minimum value, or just to lower the threshold at which they reduce speeds or power off. It looks like there are fan table values that could be mucked with.

Looks like they are gearing up to launch the first new "general" graphics benchmark since Time Spy in 2016. Speed Way will release on Wednesday (October 12, 2022). It's a $4.99 DLC upgrade if you purchased 3DMark on Steam (before October 12). Starting October 12, Speed Way will be included when you buy 3DMark (but they are also hiking the price by $5). https://aws.futuremark.com/news/3dmark-speed-way-launches-on-october-12

See @win32asmguy posts on page 5, there is a photo with the pad heights labeled.

My information must be old. I will have to check this out. Agreed. Everyone should make this tweak, it is clearly one of the easiest ways to get a boost to CPU performance. I was pretty shocked when I saw the number pop out. I'm going to make a post with detailed step-by-step and link it somewhere near the top of this thread for easier visibility. And it's not like this is doing anything "risky" — as I understand it, it is just putting the "default" value for Alder Lake in there which Dell has set incorrectly in error. Next, I will be trying ThrottleStop to raise the PL1 limit and see what happens... Yes, air comes in from the bottom and goes out both the "back" and "front" sides of the fan, so it actually circulates through the chassis. (I'm not sure how much good this actually does, I think it is more to keep the chassis surface temperature down than to cool off the CPU/GPU. But at least it gives some airflow over the main SSD slot?) Also, despite the hinge sort of in the way, the left fan actually shoves air out of the back of the system as well. In addition to @MyPC8MyBrain's recent posts, look at @win32asmguy's posts on page 5 of this thread. People are still discovering things. At some point we will have more of a comprehensive list of things to try with a sort of feel for benefit vs. difficulty for each.

Well, this is fun. Cinebench multi-core 10 minutes: Stock configuration at system launch -- 15029 After repasting -- 15488 (minor improvement, but it also shifted from thermal throttling to power throttling) I just got around to applying @win32asmguy's IA AC/DC loadline "fix" and the score shot up to 18285. And that was with Windows 11 trying to install updates in the background (didn't realize until later but didn't bother to re-run). Single run was 19323. This is without undervolting, without raising power limits, without even touching ThrottleStop, without repadding, and without replacing the heatsink with the "better" Sunon version (all things I am still considering, just need to find the time to deal with them). @win32asmguy posted the EFI commands on page 5. https://notebooktalk.net/topic/632-dell-precision-7670-dell-precision-7770-owners-thread/?do=findComment&comment=12993 I was put off by the loooong directions in the undervolting guide, but looking at it today I realized that I could basically skip to the part about setting up the EFI USB boot drive since the whole business above that with "discovering" the commands was unnecessary. I took a photo, first two lines are reading the existing variables and last two lines are running the commands to change them.

Yeah, since Dell/Intel have limited undervolting capability because of potential security issues, this is what you have to resort to.

You can undervolt using the established technique ... if you disable the E cores (which will result in lower multi-threaded performance). I don't think ThrottleStop has been updated to properly support hybrid architecture undervolting yet. You can also use the same technique to improve performance by just fixing the IA AC/DC loadline value. I used isopropyl alcohol (been doing that for years) and I had no trouble getting the old paste off.

Yeah that is silly, there is no reason Ubuntu and 3080 can’t get along. Dell probably just doesn’t want to support it.

No go ..... It still hangs at the same spot. 😕

He says DC dimming for panel backlight output, and I was wondering if that maybe doesn't apply to the OLED panel (since it doesn't really have a backlight).

Yes, the heatsink is a different (smaller) part in the 7670. And also it is pretty normal for the CPU temperature to ramp up very quickly up if you push it up to (nearly) 5 GHz. The fans in this system are pretty slow to react to a change in the CPU load, it might take them almost a minute to crank up to full speed.

Thank you, I will give this a try later ..... but this is the sort of thing that I would expect Ubuntu to automatically configure. (Granted, Ubuntu 22.10 is still beta for another few weeks, and it's also not unusual for non-LTS releases to come out a bit shaky.) Ubuntu doesn't load the proprietary NVIDIA driver by default but there is an option during the install to add it (which I did). So, it would not have been loaded in the "live CD" environment. I did not have this issue with Ubuntu 22.04 (kernel 5.15) but I see from your links that this popped up in maybe 5.18. It also seems like kernel 5.15 is too old for Alder Lake iGPU graphics support, so I think it was just loading a generic VGA driver. I don't see boot messages blaming the graphics driver like some of these logs posted show. It seems to be blaming dell_smm_hwmon, if anything. I was wondering if there is some breaking change with that module in the newer kernel version, and I thought about trying to blacklist it, but that's also not a module that I'm wanting to run without either. I'm actually pretty familiar with that one, as I spent a lot of time digging around that code when working on my own (Windows-based) fan control solutions. You may recall the time I had you helping me dump "tokens" from the BIOS trying to figure out undervolting, and we stumbled across the fan sensor values. The code for that dump was partly derived from the Dell SMBIOS code in the Linux kernel. If I do end up switching to Linux, I'll want to set up similar tooling to what I have now in Windows on the Linux side, and I'll be relying on making calls to the appropriate kernel modules... [Edit] Actually, after refreshing myself on this stuff a bit... Maybe I can blacklist dell_smm_hwmon. I think that one is obsolete with newer Dell systems (Tiger Lake and up), unless something has changed in there recently, and I really just need Dell's libsmbios for my stuff.

That may be the case, but it doesn’t immediately follow that the same brand of heatsink/fan is in use or causing the same issue. Need more data!

That's about normal for stock 7770, assuming you got the original heatsink (with Delta fans), unfortunately. It thermal throttles very quickly. ———————————————————————— Installed Ubuntu 22.10 with kernel 5.19. Well, tried to. The live CD environment booted fine, but after the first reboot, I got exactly the same error I did when trying to slap that kernel on 22.04. (Maybe I didn't do anything wrong after all... It's just a bug.) This text appears with timestamps to the left: spi-nor spi0.0: BFPT parsing failed. Please consider using SPI_NOR_SKIP_SFDP when declaring the flash dell_smm_hwmon: unable to get SMM Dell signature (...Then it just hangs there.)