Jump to content
NotebookTalk

Ionising_Radiation

Member
  • Posts

    132
  • Joined

  • Last visited

Everything posted by Ionising_Radiation

  1. Oh, I don't claim to be an expert on OLED—just that pretty much every OLED display I've seen uses PWM brightness control, ergo my expectations for the 7670 OLED panel. My phone (Sony Xperia 5iii) My family's phones (iPhone 12 Pro, Samsung Galaxy S21, Samsung Galaxy Note 8) Our TV (LG G1 55") All OLED, all have flickering. It just happens to be that PWM is the most straightforward way to control brightness, I can't really say with certainty why they don't use DC brightness dimming.
  2. In case this is not answered by Mano, all buyers of 7X70 can perform their own tests fairly straightforwardly. Start from the lowest display brightness, and wave your hand in front of the display; if you see a stroboscopic effect, then the PWM frequency is low enough to affect eyesight and comfort. Generally the PWM frequency/duty cycle increases with increasing brightness, and most displays are at 100% duty cycle on 100% brightness. By manufacturer, I've found that BOE is the worst offender, and AUO and LG almost never use PWM; Chi Mei does use it sometimes (my old Clevo W230SS with a 13" N133HSE panel had PWM; the current panel in my Precision is also by CMN, but no (detectable) PWM). I am betting that the OLED display almost certainly uses PWM, and the IPS LCD ones may not.
  3. @Aaron44126, are you using ThrottleStop to monitor core temps and throttling behaviour? This might help narrow down why your CPU is not able to reach its maximum performance, whether it be a power, current, or thermal limit...
  4. He did, yes. In fact, he disassembled nearly the entire laptop to get to the display, because he followed the service manual to the letter, but at least he did, which means he's experienced. I once got a tech that straight-up forgot to plug in the cables on my 7530 and I had to fix them up after the fact.
  5. Got some time to run Time Spy and Fire Strike: pretty much what I expected. Trade a bit of CPU performance for a bit of GPU performance. Personally, I'm OK with this: most of the time, the games I play are GPU-bottlenecked. CPU has been undervolted. I do think however that a repaste will help, since like I mentioned, the display was replaced and I didn't monitor how the technician re-pasted the CPU/GPU. Time Spy Fire Strike
  6. More interesting findings. I recently replaced my display because of some structural damage; I was luckily given the same display as I had before: the Chi Mei Innolux N156HCA-GA4 panel (also CMN151A, dell P/N DPVHF). As usual, this has no (detectable) PWM flicker. I was able to overclock the usually 60-Hz display to a whopping 87 Hz (that's a 31% improvement in frame time: 16.67 ms to 11.49 ms). This is easiest when 'Hybrid Graphics' is switched off in the firmware, and the full-fledged NVIDIA Control Panel is available: My aim is to reproduce this when the Intel GPU is driving the display (for some reason, the Intel control panel has a custom resolution option, but doesn't really work at all) as well as on Linux with xrandr.
  7. Wonderful. @PlagueWrath's VBIOS flashed successfully on the on the 7560 A4000. One needs to use this modified NVflash (I downloaded 5.670, as the command below might suggest) with the following command in an administrator CMD.exe: nvflash-5.670.0-Patched.exe --protectoff && nvflash-5.670.0-Patched.exe a4000_7760.rom and follow closely the instructions on-screen, and accept all the scary prompts. After flashing, the 'Maximum Graphics Power' listed in the NVIDIA Control Panel System Information window goes from 110 W to 130 W: Power draw during benchmarks is also ~20 W higher (Fire Strike tops out at ~115 W power draw, which is significantly more than the ~87 W I used to see), and performance is noticeably improved. I probably need to run more detailed benchmarks, which right now I haven't got the time for, but this is nice: free 7760 performance in the 7560. I replicated this in the Pascal Quadro P3200, and again with the Turing RTX 5000 in the Precision 7530. @Aaron44126, you Alder Lake lot in the 7670/7770 thread may want to take note: this hack still works as late as the 7X60 gen. Extrapolating, 7670 users may still be able to get a performance boost, especially since the GPUs are still Ampere; the flash command should still work. @TheQuentincc, not entirely, but I suspected that both cards were almost exactly the same with the only difference being the power limit; I was right. The good thing is that the VBIOSes extracted from the 7760 are signed, so only a board-ID mismatch needs to be fixed (and this is rectified with the patched NVflash above).
  8. I haven't heard of this, needs investigating. That said, power management with Optimus on Linux with Turing and later architectures doesn't need any intervention at all (bumblebee, optimus-manager, etc); the nvidia driver and the Linux kernel together is sufficient, now that runtime D3 power management and (reverse) PRIME rendering offload has been fully implemented. Only caveat is that (I think) Wayland support is still lacking; Xorg, on the other hand, works perfectly.
  9. Has anyone got a 7760 with an A4000 in it? Could you be kind enough to extract the VBIOS with nvflash, using the command: cd path\to\nvflash64.exe nvflash64.exe --save a4000.rom and upload the resulting .rom file? Thanks!
  10. Any distros you're partial to? I foresee the following points of note on Arch Linux at least: fingerprint sensor may/may not work (may require libfprint-tod-git patch) Optimus should work almost out of the box with Xorg, the Intel mode-setting driver, and the proprietary nvidia driver (zero configuration needed, especially with latest Xorg which even handles PRIME/reverse PRIME synchronisation) NFC and smart-card reader may be hit-or-miss All other peripherals should work with minimal/no configuration Backlight adjustment may need a desktop environment package (KDE's powerdevil provides this functionality, for instance) Battery life should (in theory) exceed that on Windows; this is the same behaviour I've seen with my 7530 after the Turing upgrade, and on my 7560 now.
  11. I need more USB-A ports (and possibly more displays) than the 7560 supports, so I am in the market for a Dell dock. I've read through the Dock compatibility guide and I am comparing between the WD22TB4 and the WD19DCS. I am most concerned about the power delivery: will I be able to charge the laptop with just the WD22TB4, and is the CPU/GPU likely to be throttled? I absolutely don't want the hassle of having two gigantic AC adaptors lying on the floor. Furthermore, is the WD19DCS (which is recommended over the WD22TB4 for the Precision 7560) liable for an update soon? As the model name suggests, it's from 2019, and USB-C PD today supports 240 W over a single port, which makes me feel that Dell might release an updated dock with a single cable some time soon.
  12. I saw this and immediately wanted to address these issues, because I'm actually running Arch with KDE Plasma on my notebook right now. The`ibt=off` kernel parameter is a kernel-driver interop bug; it has nothing to do with Dell Precisions specifically and was widely reported several days ago (Reddit, GitHub). Fn-key support requires something like PowerDevil (comes with KDE Plasma) to be installed. Fn-key brightness control, volume controls, display controls, and the keyboard backlight controls all work without intervention on both my old 7530 and my current 7560. The only keys that don't work are the three calculator shortcut keys, but that's because they're basically aliases for other keys on the keyboard (damn it, Dell). On my 7530, they were media control keys, which also worked fine). The hybrid GPU works great in Arch, but there is a modicum of configuration required: the NVIDIA and i915 modules need to be loaded with early KMS. In fact, I'd argue that Optimus works even better on Linux than it does on Windows (despite the warning in the firmware), because the NVIDIA driver has properly implemented runtime D3 power management without any configuration, and Linux programs don't wake the discrete GPU willy-nilly like they tend to do on Windows. As a result, I can achieve a whopping 12-13 hours of productive battery life out of Linux, whereas on Windows it's 8 hours if I'm lucky. Furthermore, the other aspect of Optimus, dynamic source/sink switching (implemented with PRIME) works with zero configuration. There were bugs in X.Org last year (dGPU/iGPU display and frame synchronisation, 1 FPS if only external display is enabled, etc etc), but as of writing most of them have been fixed. Even so, these are bugs with upstream software, rather than any firmware/drivers with the Precision itself. There are other things that don't work on Linux as well as they do on Windows, like precise trackpad gestures, colour management, and the fingerprint sensor. As @Aaron44126 linked earlier, libfprint-tod is a branch of the libfprint and fprint driver which works with modern touch-based sensors like on the Precision 7560; it is available on the AUR as libfprint-tod-git. All in all, if you want to get Linux on your Precision 7670, I am betting that it will be (mostly) smooth-sailing. The largest vendors (Intel, NVIDIA, even Dell) now treat Linux as a first-class citizen, and frequently release support and updates for their devices almost as soon as the devices are in public use.
  13. What's up, lads (and lasses) 🙂 Glad to see more old NBR folks join in. Hopefully we can see the sort of deep-dive discussions we used to see on overclockers.net, NBR, Thaiphoon fora, etc etc...
  14. The original panel was a BOE NV15N4R with absolutely rubbish PWM. The newer one is an AUO B156HANsomething.something (I don't remember exactly), and I can't detect the PWM at all (either it uses DC control, or a very high PWM frequency). My only gripe about the panel is that it's not 16:10, and that it's still 60 Hz. Even the 7670 still has a 60 Hz panel. Grr.
  15. Interesting; this generation of Precisions has gotten much more media attention than before (possibly because of the CAMM). It's very nice that Dell wants to make it a JEDEC standard and I hope it goes through because it looks technically sound. However, it also looks like the boards are huge—they appear to rival MXM 3.0b modules in size.
  16. No worries: I completely understand your position. Now that I read your comment and mine, I'd probably come to the same conclusion too: many of these changes were inevitable, like it or not. I want to respond to some things though... Dell is still the best at serviceability, no doubt (since when has Apple released service manuals?) but there is always the potential for vendor lock-in (e.g. the Alienware Area-51M DGFF was not upgradeable after the first generation). The CAMM module is a fair bit of a problem. At least Dell is offering an interposer, which is a decent stand-in. However, we all know how expensive memory upgrades are from Dell. Even so, I don't see them moving away, since, like you said, their target demographic for Precisions are businesses with very deep pockets and full, next-business-day support contracts that effectively replace the entire machine to avoid any sort of downtime. As for clickpads, I think this is very subjective: I hate the 'Force Touch' trackpads on MacBooks. I know and feel my finger hasn't moved when I click: a motor vibrating the trackpad a certain way isn't a good enough substitute for a set of three buttons. Speaking of those buttons: I use the middle mouse button a LOT (middle click to close, middle click for new tab, middle click for instant scroll, etc). This is a serious regression for me.
  17. Now that I see the full spec sheet, my reactions are mixed. Many of these changes are great—16:10 display, new hinge, slimmer chassis, massive power and cooling improvements, year-on-year spec bump, quad speakers in front so the laptop can be used in clamshell mode. At the same time, Dell has sacrificed and removed things that have been so useful, like mDP (has Dell marketing really surmised that none of their customers uses it?), the physical buttons for the trackpad, 4 SODIMM slots, moving to the CNVi version of the AX210, aka AX211... In many ways, the 7670/7770 chassis are a bit of a step back from the 7530/7730. Definitely the latter were a pain to take apart, but at least they were almost entirely modular, down to the cables for the track pad. I'm thoroughly not fond of the proprietary direction Dell has taken for the memory—the Precision 7000s were the last bastions of not-crazy-thick notebooks with 4 SODIMM slots. I guess this is the XPS effect bleeding over to the Precision 7000s. One wonders what HP has in store for its ZBook Fury G9...
  18. Oh man, these new Precisions look really good. Slight bummer about the (proprietary) CAMM—it looks like they are dropping to 64 GB SODIMMs only because there will only be 2 slots, and the CAMM is needed to go up to 128 GB. I am also betting the DGFF also changes this time around. Even so, I am going to see if my university is nice enough to bump my 7560 up...
  19. Thanks for verbalising in detail what I was thinking, @Aaron44126. Low-level scheduler support for Intel's new P-E architecture is critical to extract both maximum performance and efficiency from the 12th-gen Intel CPUs. For many users, Windows 10 is a stable product with a known workflow, which Windows 11 drastically breaks.
  20. Holy crap, these resources are pure gold, direct from the horse's mouth! I hope they don't close off access to these files; it's very nice to see future details coming through.
×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue. Terms of Use