win32asmguy

The dips are interesting. Not that the remix raids are particularly challenging as the game mode is pretty broken but its just more enjoyable to have performance consistent. I could pick up a 2x24GB 8800 CL42 Teamgroup CUDIMM kit from Micro Center for $550. I do have a bunch of laptop memory that may fetch more money now. Legion 9i G10 - 275HX at 58W and 5090 at 150W combined load, not really overheating as it can go up 300W combined loads. I need to just give it 100W on the 275HX as WoW could probably use more TDP. Its also severly limited on memory compatibility. The Kingston Fury 2x32GB kit is the only performance memory that works, and only up to 5200 CL38. No custom profiles or XMP profiles can post. That results in AIDA64 latency of around 105-110ns with Gear2, boosted D2D and NGU. Not great but still far better than the X580 could do. The Hydroc G2 can do better with memory latency but its also the smaller screen, and much more loud and much hotter palm rest.

I have the clevo x370 4090 bios if you would like to try it!

I have done remix raid world tours on the 12900KS + 5090 and it can usually hold 90 fps even in the really heavy fights at QHD. I am guessing it is still CPU limited albeit better than the 2495X could do. I never picked up that team group 2x24gb kit back when y'all recommended it so I will have to tune my a-die 2x16gb on the 285k instead. Hopefully can find good starting point timings as it only has XMP at 6400 C32. Hoping it can keep a stable 120fps in the 1% lows as that is the refresh rate on a new display with kvm I am testing. The laptops are all basically a lag fest in the raids at those same settings...

I have seen WHEA error 17 on multiple other Arrow Lake systems including the Eluktronics Hydroc G2 and Lenovo Legion 9i Gen 10. Supposedly setting PCIe Power State Management to disabled in the bios could fix this but I have not tested that.

I was disappointed when the 5090 mobile GPU option was dropped from the Pro Max Plus lineup. It seems like Nvidia may be the one responsible here as the T16g G3 has not been released yet either so it could be a play to boost 4000/5000 Blackwell mobile sales. The Legion happened to be the best choice for a consumer model 18 inch for me. Its warranty and repairability is worse than the Pro Max 18 Plus but I would be selling and upgrading before the warranty expires with either machine.

The battery drain under load was a big issue for me. It cannot actually sustain 200W as it needs to draw down the battery at 20W per hour to do so. I really just wanted an 18 inch version of this years uniwill chassis. As it turns out the legion 9 g10 is much closer to that. It also has even better build quality than the pro max 18 plus along with a 1200p panel mode which works much better for me when docked at my desk.

Well if you are going for a Pro Max 16 Plus it might have less of a battery drain issue under load as it is capped at 170W instead of 200W. Then just repaste the cooling system with PTM7950 and better VRAM putty. I returned the Pro Max 18 Plus and opted for a Legion 9i Gen 10 instead. Its been a great machine and definitely the best 18 inch model I have used. Its a shame they did not offer a 1600p panel option in addition to the dual mode 1200p/2400p as that is what most people would want to use. I can even run 2x32GB at 5200MT CL38 in Gear2 and get close to 100ns memory latency with D2D and NGU overclocks.

I did not have issues with the micron 3500 512gb SSD included with my Pro max, but I also used W10 so maybe that helped with avoiding the SSD issues posted above. I also have a 4tb sn8100 which has worked well in the hydroc g2 but I did not test it in the pro Max.

@Easa nice post! If you happen to run some combined load stress tests could you monitor if it periodically trips the ia: rapl/psyspl2 throttle flag reducing the CPU to around 15w until the power supply is disconnected and reconnected? Oh and for Optimus make sure advanced Optimus is not enabled as it is pretty buggy.

I think they might consider exposing manual fan control a security concern at this point. To be completely honest I was surprised that the Pro Max 18 Plus does not yet have its EFI variables write protected yet like the Zbook and Thinkpad. The fan noise is better than the previous Precision 7780 and also Thinkpad P16 G2. I spent a bunch of time over the weekend trying to find a way to prevent the power adapter performance locking issue. It seems like it will always eventually happen if the system is in performance mode and under any kind of high load scenario where the battery is draining down while plugged in. I have to box it up and send it back as their support said it was "operating as designed".

24GB Hynix M-die CSODIMMs boot at 7200 with tREFI at 16k, 1.2v. You can also tune NGU via Intel XTU, so I have it set to 30x up from the default 26x. Sadly D2D is not accessible so it is still at the stock 21x. Supposedly that one is less important based on some tuning I have seen on 285K setups.

They may as well cap the 18 Plus at that wattage as well because these drain down the battery above 170W from the CPU+GPU. In theory its designed for 175W GPU + 58W CPU, but the closer it gets to 200W (where thermal throttling starts to occur) drain can be up to 20% per hour. It begins recharging the battery while under load if a charge threshold is set, at a rate of 5W, and reduced CPU+GPU TDP until charging stops again. They also periodically trip some kind of throttling condition which can only be reset by unplugging / replugging the power adapter or a cold boot. Until that happens the CPU is stuck at 15W. I am not sure how extensive Notebookcheck tested their system but it occurs multiple times per hour on my system with the current bios and drivers.

The performance is pretty inconsistent with the battery disconnected. The GPU floats between 60W and 120W while the CPU is around 45W. I guess it does not want to get anywhere near the higher performance limits so a transient spike does not trip the power supply. I tried to ask Dell for a way to disable this feature as I prefer to keep my battery charged at 80% reserved for use when needed. Nor do I want the excessive wear and heat as it continually drains and recharges under the supposed 200W sustained load capability. From what I have heard this is working as designed.

There is probably a hidden bios option in there somewhere I have not found yet that flags it to require CSODIMM. It can boot aftermarket CSODIMM modules not sourced from Dell. 24GB Hynix modules work well as single rank timings are better and the modules generate less heat than dual rank. The CAMM2 modules have a heatsink built for them so that is what I would choose if I needed more than 2x24GB (highest single rank capacity). My biggest annoyance is still the battery drain under load that is by design. I have a custom charge range set from 70% start to 80% stop, and if the system is under load it is basically always either discharging down by up to 20W per hour or charging back up at 5W (trickle) - 60W (express). Sometimes it decides to express charge the battery even if you are under load so whatever compute task that is running just gets throttled instead. I have even seen the firmware trip some kind of internal throttling protection where the CPU refuses to go over 20W until you unplug and replug the power adapter.

If you can find 64GB CSODIMM modules they likely would work. But I think the regular SODIMM 64GB modules available from Crucial currently would not work as the system would not post with a SODIMM installed when I tested.