Leaderboard

Final (ish) analysis of the whole can't boot / memory problems: Motherboards: Asus X870E Crosshair Hero (still waiting for the Dark variant to be sent out) HP Omen 35L Prebuilt B850 tower (8400F, 5060, single Rage 6000 DDR5 stick, 512GB NVME) Asrock Phantom B850i Asrock B850M-C CPUs: 8400f 7500X3D 9800X3D Memory: Kingston Rage (Rage) DDR5 6000 1x16GB SR Teamgroup (TG) 8200 2x24GB Kingbank (KB) 8400 2x24GB Corsair Vengeance (CV) 9200 2x24GB --------------------------------------------------------------------- Rage memory boots and runs in every motherboard with every CPU TG, and CV boots and runs with every MB and CPU except the 7500X3D KB boots and runs in Hero and HP except with 7500X3D . Total no go in both AsRock boards. This is a two fold with conflation: KB known issue with Asrock boards so they're a no go atm....it was conflating an issue with the 7500X3D. 7500X3D has a dud IMC and can't handle 24GB sticks in single or double config. None of the 2x24GB sticks will post on Asrock boards in single or double sticks with the 7500X3D. They will post on HP and Hero but trying to get to desktop they keep getting BSODs and exception errors on the login screens over and over tested from 4800-6000 it doesn't matter. BSOD city on the MBs it can actually boot. It has no problem with a single stick of SR 16GB Rage memory. All MBs are functional. It was the KB issue and 7500X3D IMC being faulty causing all the problems. Returning the B850i Phantom (popping it in the mail tomorrow) and send the 7500X3D to AMD for an RMA or see if MC would allow a replacement but it will work itself out one way or the other. I plan on running the SFF with my 9800X3D now anyhow and put the 9850X3D delidded into the new Hero Dark when it arrives. In a "we're blessed" moment, Imagine not having a decent amount of hardware to swap and test to log data and diagnose problems all because of a flaky/dud IMC on the CPU and hidden but known compatibility issue with KB sticks with some AsRock MBs.....crazy. ------------------------------------------------- Smart play when you want a third decently powerful GPU in play since you run three rigs usually at all times. Odds of scoring a secondary 5090, unlike the 4090 days, for a somewhat sane price (less than $3k) have long flown the coop. $1100 is super close to MSRP. Did you snag a WF OC SFF or normal? Unlike some of the OCN forum members who continue to pick up new 5090s, that's too rich for me. I'm going to ride this Vanguard into the ground as my main rig gem. This. I'm more than content with my Vanguard since it is a 1.125v sample right out of the box, but the 5080 tops out at 1.040 so that is going to be major fun OCing the 5080 for funsies to see where it goes. Just hoping Jensen doesn't shoot us the bird right back and find a way to lock it out in a future driver revision. I can just imagine Jensen screaming, "THE MORE YOU SPEND THE MORE YOU BUY MOTHER%$*$)($S!!!" and smashing his keyboard against a bunch of GPU displays...... 🤣 ---------------------------------------------- Jensen right now watching everyone celebrating having control over our own hardware to push the voltage where we want:

I have had people (some not even noobs) argue with me and say things like "no, that's not the problem because my RAM is only reaching 65-70°C" and they often don't believe me that anything above about 40-45°C is "too hot" and will cause errors.

Confirmed only blown fuses. GPU is totally fine. I bridged the F1 fuse solder pads with a wire for testing and it's good to go. Now I just need the fuses and we'll be back to normal +1 fuse for extra capacity.

I am actually kind of surprised they have not already blown the way I have pushed this GPU for benching. I've run way more than they should have allowed me to, but Steel Nomad is famous for the massive power spike in the first couple of seconds and I think that experimenting with the 2000W vBIOS together with the shunt modding was the straw that broke the camel's back for the fuses. I normally have used a 800W or 1000W BIOS.

Finally overclocking GPU is fun again.

I used Asus Astral 5080 vBIOS before since it has a 450w limit on the MSI 5080 I had last year for a bit but all three fans worked. I'm going to end up flashing it again on the SFF OC like yours. That and Melon should open them up like a beast especially yours on water so a fan not working won't be an issue for you. I wonder if a different version might wake up the non working fan since there are three variants floating around for the Astral 5080 (Black, Noctua, White) The tight knit community is alive and well. Crazy how enhancements and QoL are being implemented Fast and Furious.....

This is with the Astral 5080 XOC vBIOS. It totally exhausts the 450W power limit (pegged to 450W on the WireView Pro). I've ordered a block for it and will shunt mod it. One fan does not run so it is getting too hot and thermal throttling, but the software works great for the tweaks.

The same DEV that created the original Melonvolt just released a separate tool for manually setting/overclocking the XBar without touching MSVDD at all! Just tested on my 5070 Ti and works flawlessly. Just set the offset, hit apply and bam. Amazing work. https://www.overclock.net/posts/29605861/ I guess both Devs are now sorta working together and will implement this Xbar offset tool into the Melonvolt tool so we don't have to use separate tools. Things are happening so fast lately with RTX 50 series. Getting wild. Jensen is having a bad week! Edit: Tool works great because you can add Xbar without needing to waste power or push voltage to get it to scale up. Saves if you're power limited and saves on wasted heat. Real world gaming in Battlefield 6 was about 2% on my 5070 test. But I'll take it. Added 500Mhz to my Xbar. In total about 13% over a stock card. Not bad at all..

HOLY CRAP! here i was, just about to start work in the office when i stumbled upon this, bam 45 min gone LOL. omg cant wait to check this out once i get home.... what a total middle finger to Nvidia, friggin hilarious! show them those melons, guys! 😄

This is awesome. I tested it on both of my 5080 GPUs and it definitely works. I will test it on the 5090 and see if it will allow me to remove the EVC2 mod without losing anything. I hope that Jensen is not happy. He doesn't deserve to have a happy life.

Awesome. Thank you. I have always removed Defender garbage from Windoze so I don't have to worry about Micro$lop's ever-growing database of false positive file flags or bogus blacklists nonsense.

https://www.overclock.net/posts/29605610/ Defender throws up false virus readings. So had to disable to get it. 1150mV limits by default. 1250mv with XOC button. XBar scales with MSVDD. By default Nvidia made it so you can’t scale MSVDD beyond NVVDD default baked in. But now that we can raise NVVDD, you can push MSVDD higher and therefore XBar.

Now with Melon Volt 0.2 you can punch the voltage on that 5080, on both the NVVDD and MSVDD. No longer voltage limited on 5070, 5070 Ti, 5080, or 5090s on default vBIOS. Now just power limits until shunt that is. Overclock.net forum going wild with new version that lets us unleash the cards! Jensen is not going to be happy about this lol.

xx60 class card above $800. Hmmmm. Nope. No thanks. Hell freze first. Madness is the correct word for this. GeForce RTX 5060 Ti 16GB pricing hits $800+ in the US as memory crisis takes hold NVIDIA raises RTX PRO 6000 Blackwell price to $16,000, now 87% above original MSRP

Yep, RAM is very temperature sensitive. The upstairs A/C in my house broke recently and we had to get it repaired. While I personally didn't mind the 10-15°F higher ambient temperatures, that destabilized the memory overclock on my desktop. At first, I thought something was horrifically wrong as the machine would just keep freezing and BSODing after over a year and a half of working great. Nope, it was just the higher ambient temps causing RAM instability. Whew. I wanted my 2.5% overclock though! Grr... performance being left on the table I couldn't access.

https://www.youtube.com/watch?v=2jXNs3Hrgpg Given the price mark up on Broadwell GPUs, you are going to see a lot more 4090s and 5090s get memory upgrades.

https://www.techpowerup.com/vgabios/274990/274990 https://www.techpowerup.com/vgabios/273887/gigabyte-rtx5080-16384-250110-2 This should work. 450w from Aorus, should make all fans work and has a higher default boost. If not maybe Gaming OC, same 450w but lower boost by default. https://www.techpowerup.com/vgabios/273730/gigabyte-rtx5080-16384-250108

Maybe another variant would wake up the non-functional fan. This is the XOC 450W vBIOS I use on the Prime 5080 and all three fans work. The one not working is near the I/O so that is a more critical fan being close to the die and VRM. But, it certainly will not matter when blocked. Absolutely. I have hated having to use the EVC2 for voltage control. It is a cool mod from a hardware geek perspective, but a real pain in the butt to have to implement, and we should not have to. All GPUs should function like a K|NGP|N or HOF GPU. I'd love it if we could get the power limits unlocked (adjustable) or simply removed (increased to something more than the GPU can ever pull, like 5000W) using software to avoid having to shunt mod. The risk involved with that kind of geeky fun (and loss of warranty) loses appeal when GPUs cost two to five times what they are actually worth. If it is done with software, there is no way to prove or enforce a warranty that excludes overclocked hardware that fails for reasons that have nothing to do with overclocking.

Thanks. Nice. All rolled into one now. No need to use a separate tool for crossbar. I love it.

so guys, for easier follow-ups here the author profile on overclock.net as well as his newly set up GitHub page for the magical mVolt tool that lets us unleash our GPUs via software ONLY 🙂 overclock.net profile: https://www.overclock.net/members/b00nza.701858/?filter_threads=1&filter_replies=1#activity GitHub: https://github.com/b00nz/mVolt/ latest version now supports profiles, profile set at logon, V/F curve editor and other UI tweaks. i did a quick check last night, both Xbar and GPU Core Voltage adjustments work flawlessly on the fly, same with Clock offsets for Xbar, Core, vRAM. like....no need for anything else anymore, really 😄

Z890 Apex #2 landed and launched. The crappy Gigabutt Duo X is on its way back to Amazon. Everything is as it should be now. No problem not booting with 40x D2D like issue with the Duo X. The BIOS was from 2024. I flashed 9901 but left the 2024 version of Intel ME. No plans to update it. No need until something proves otherwise. So far nothing has. Temps are great. Runs 8800 no issue, same as the other Apex.

Sweet! Just tested it and it works perfectly. Thanks for alerting us. Jensen stopped caring about what we want and need, so he needs to also stop caring about what we do if he has not already.

Yup, works like a champ on my 5090, too. I'll do some testing and if it does as good or better I will remove the EVC2 mod and just use this instead. Then I won't have to care about the WireView Pro I2C conflicts because there will no longer be one without the EVC2. The MSVDD did goose my crossbar clock a bit on all three GPUs. I think this could end up being the world's most loved utility. I haven't even used it for an hour yet and I love it. It gives us all what we deserved to get from NVIDIA, but they're just too scummy to give us everything they owed us for buying their absurdly overpriced gimped GPU crap.

https://www.3dmark.com/3dm/161882960 Good enough for #13 Bartlett Lake is the CPU that could have been. This chip with 8400-8800MTs would have been elite.

I think I forgot to mention... Given the unsavory trajectory of GPU pricing, I grabbed another 5080 on sale from Microcenter last week for $1100 and snagged the last one available. I put the Prime 5080 in the SFF living room build because the Gigabyte Windforce OC I snagged has a higher core voltage and crossbar clock and I want it in one of my preferred builds. That's still more than a 5080 should cost, but about $400 less than the asinine price most are selling for and it's not going to get better anytime soon. I rationalized spending less on something that I don't need right now versus paying an undetermined amount more later when I do need it. I also think it might be easier to source a waterblock for the Windforce that isn't selling for an idiotic price. I could only find one option for the Prime, and the price was totally unacceptable. Paying current 5090 prices is a fool's errand and I will have no part of it. Utterly insane and absolutely not worth it, even by a long stretch of the imagination. The price-to-performance math and logic simply does not work for 5090 at current prices.

That was my fear with that board, great looking board and features, 2 DIMMs, but the Gigabyte UEFI has been rumored to be trash, especially for D2D tuning. This is why APEX is such king this gen, despite slightly lower memory tuning compared to OCF. If I can't push D2D, or get a reliable stable setup, being able to push slightly higher on memory OC is worthless to me.

I am returning the Gigabyte Z890 AORUS Elite Duo X. It arrived today after a 5-day slow ride from Amazon in Baltimore and I have spent about 8 or 9 hours fighting it. Super nice motherboard, but something is bugged with the hardware or the firmware. When I first installed it, could not get it to POST normally with two RAM sticks. It would POST, restart and then hang with the DRAM debug LED illuminated. So, I swapped in another memory kit. Same thing. Cleared CMOS several times and nothing. I then got it to boot with one RAM stick, but Windows would freeze while loading. I tried to reflash the BIOS using the EZ-Flash button after renaming the BIOS file GIGABYTE.bin but the button would not do anything. Totally unresponsive. I finally got into the BIOS and the file was seen as "invalid ROM" by the BIOS. I moved the memory stick to the other slot and tried again, then it worked and I flashed the BIOS. After that it booted with both RAM sticks. Both memory kits booted fine. I turned it off, unplugged it, then plugged it in again. This time the EZ-Flash button worked and it seemed all good. Easily hit 8800 stable. I thought "OK, that was weird, but everything is fine now." NOT! I began working on tuning the CPU, but any change made to the D2D clock ratio other than leaving it set to "Auto" it would only boot to BIOS Safe Mode with the message that the BIOS settings were unstable. Another issue I encountered, even though it is a 2-DIMM board, the memory slots are spaced several millimeters further apart than the Apex and the Supercool direct touch RAM block would not work with it. I considered getting an ASROCK Taichi OCF, but don't really like the brand or how it looks. (I have hated all three ASROCK mobos I have owned in the past.) Ordered another Z890 Apex for $25 less than the OCF. It was $100 more than the Duo X, but I don't need any more drama. It will be worth the extra to keep my sanity. No more likely to drip on the GPU than the CPU block. I've never had a memory water block that leaked before. This story made the thought of getting the Taichi OCF an even more unattractive option. That, plus the fact that I have never actually cared for any ASROCK motherboard previously, including a Z370 Taichi and X870E Taichi.

pretty cool, couldve gone 50 other ways straight to gpu melting hell instead 😄

Basically a stack shift to shoe horn in a very poor card where a 9060 should go while those former prices have all shifted up and will most likely continue to shift upward. This card is bad for the price. If it was $199.99 it suddenly becomes an ok budget card but $280? And this was the 8GB version..... I can imagine the 4GB version is just going to be extreme trash versus just normal trash on this one. WOW......I mean I'm glad to see the fuses doing their job but damn..... Hopefully this will be a quick and easy fuse swap and upgrade, but yikes.

Well, it seems the 12VHPWR fuses are blown on my 5090. I was launching Steel Nomad with a heavy overclock and the card turned off (red light for 12VHPWR came on). I checked the three fuses for the 12VHPWR with my mullti-meter and all three are open. So, I ordered new fuses from Mouser. I am going to add a 4th fuse to the open solder pads for F1 that Zotac did not populate. That should take it from 720W fuse limit to 960 with the 4th 20A fuse. I checked the output side of the fuses against ground and there is no short. I think the massive power spike lauching Steel Nomad was just too much too sudden. In one way it sucks because it is very inconvenient, but in another way it is awesome that they did exactly what they are supposed to do. This will give me a reason to test my new soldering tweezers. Should be (hopefully) fairly easy to replace them.

You'll want big speakers and a nice sub for this one.

Here it is. Best to use are Clevo's eDP cables, since you need only to solder on motherboard side. Connectors are dissasembled and reassembled without damaging. There is 8560w boardview file in internets, you need to open it with Open board viewer - for examining mobo side, if you're not sure. For 4K cable you need to use 40pin cable and need to connect 4 additional cables - third and fourth pair of eDP_TX positive and negative channels. Hp original LVDS cable also can be modded, but then you need to solder both sides, and check cables using multimeter. LVDS LCD side is 40pin too, if I can remember. Unfortunately I don't have free time lately that I could go on with modding as I had been doing. So I have to end 8570w journey. I still use it sometimes, letting it retire in peace 🙂 You may hear more of me in another thread, maybe someone will use all that experience saved here to use.

Carmina Burana - O Fortuna Best played at a volume guaranteed to piss off your neighbours

It does fantastic with the stock vBIOS but I have not had time to research or test one with higher power limit. If there is one you suggest give me a link to which one (or upload) and I will run some tests on it. I just ordered a Bykski block for it on AliExpress for $200 ($75 less than from Byski.us) with free shipping. Bykski Armored End Port GPU Waterblock Gigabyte GeForce RTX 5080 Gaming OC 16G (N-GV5080GMOC-X-V2) Here is a 12-loop run of Steel Nomad usng the stock vBIOS. About 10 loops in I installed the side panel and the GPU temps dropped by 2-3°C so the max temps would have been that much lower than HWiNFO64 is showing.

How is that 5080 SFF? It's basically the only reasonably priced 5080 left at MC. Can it handle a higher wattage vBIOS?

Sheesh! What a stinking nightmare. Glad to see there will be a resolution. You're definitely blessed in that you had sufficient resources and spare product to allow the issue to be identified, as well as having a technical knowledge base adequate for it. For your average Point-N-Click keyboard jockey that would have been a valid reason for having a meltdown. SFF. The dimensions are basically identical to the Prime. Which means it would have been yesterday's version of a normal sized video card. "Normal" today means unnecessarily large. 3070 FTW3 will be the emergency backup GPU.

This is such a nice CPU. The SP87 one you gave me actually seems to perform almost the same except for the SP98 does it with less voltage. I think on the chiller it's not going to make much difference which one I am using. Without the chiller it will likely run cooler but they both run very reasonable temperatures even when pushed hard since being delidded. The chiller will be mainly useful for the GPU to boost higher and hold boost clocks. I feel blessed to have two really great 270K Plus samples. I think I paid too much for it, but everything worth having is overpriced and a poor value now, so I digress.

I know! I was messing with it all afternoon on all my cards. I need to shunt mod my lowly 5070/5070 Ti now. Usually the lower tier cards don’t get this fun, but now we can have fun with the entire stack. I laughed too hard at that last sentence. 🤪 Finally AI is being used for good! The tools they made being used against them is just too ironic.

They used the space to the right of the DIMM slots for a vertical NVMe slot and heatsink, which I am totally fine with. I like that a lot, same as DIMM.2 and the NVMe slot on the B850MPOWER next to the memory. But, it is goofy that they put just enough distance between the slots to make the Supercool block not work. The Apex has wider spacing than most as well, and the IceMan block is too narrow for the Apex. So, the Elite Duo X has even more space than the Apex. You don't notice it until you try to screw the block onto the DIMM jackets. You have to bend the modules toward one another to get the screws in, so that is not cool. I initially installed it and got to looking closer and did not feel comfortable with the visible amount of distortion. I tried using thicker thermal pads on right DIMM to shift the cold plate toward the left memory slot to balance it out and relieve the tension, but that was not enough to get them straight enough for me to feel comfortable with the installation being "safe" for the memory slots. (That thermal pad trick works fine on the Apex to use an IceMan block because the space is not as great. It's only about 0.75-1.0mm wider than normal spacing on the Apex.) Things would be better if all of the OEMs held to a standard on this.

"New" system coming out for 2026......rerelease of 2021 hardware.... https://www.bestbuy.com/product/acer-aspire-xc-1715_czn-ub13-desktop-amd-ryzen-3-5000-series-2021-8gb-lpddr5-512gb-usb-keyboard-and-mouse-black/JJ8V8HY7FJ/sku/6679698 Regression at its most finest to compensate for the AI gear hoovering atm.... --------------------------------------------- It makes you wonder why a 2-dimm board would introduce extra length to the traces and chance for signal errors at high/tight settings unless the spacing is compatible for a potential 4 dimm version too? Yeah I'm right behind ya with a Newegg RMA. I already got my RMA to send back the Asrock B850i Phantom. Basically shares the same BIOS as the B850M-C and both are trash IMHO that are having the exact same problems that extend beyond the whole KingBank fiasco. Even with a fan with an open case raining down AC cooled air, if my temps in my room rise....which they most certainly will with a 5090 at work.....an on the edge OC on my memory will crash and burn sporadically especially gear 2 8000+ Another reason a CPU/MEM loop has me interested. I've thought about a window AC unit to offset the fact my computer room routinely gets much warmer than the rest of the house to keep it a rock steady ~70 in here vs the inevitable temp creep when I'm gaming and the 5090 is a portable heater. Seriously, just like the 4090 but moreso, in the winter months I shut the door and use it to really heat up my room when I get cold.

When Gigabyte scores a win it is usually pretty solid and respectable. They often do not, and they are masters of cheapness and engineering shortcuts. They are an example of a company (there are many) that doesn't learn from its mistakes. The X870E AORUS Master was a phenomenal motherboard and I mostly loved it. The firmware had some minor but glaring functionality omissions that they didn't care about. The one that I hated most was no ability to disable onboard WiFi and BT modules. For people that use them, not a big deal. For people like me that do not, being able to disable them is super important. I don't want crap that I don't use enabled and I deserve having the ability to disable it. Like many companies, they only care about what normies care about so they can sell them products and if you're not a normie you get to have whatever the normies are willing to put up with. Guess what? The Elite Duo X has no BIOS option to disable WiFi/BT. But, I can disable the wired ethernet device if I use WiFi. Bastards.

Who said it was a huge shortage of vram? China have everything if and when you need it. Yep, China is the global tech Mecca. Ask and you get whatever tech is possible to get. And no one can stop them. None! Yup, because the US chip restrictions for China doesn't work. Nada. Zero... A JOKE. Modified GeForce RTX 4080 32GB cards flood China’s second-hand market RTX 40 memory mods have become a business This follows several similar cases we have covered over the past two years. Modified RTX 4090 and RTX 4090D cards with 48GB memory have been sold in China, while RTX 4080 SUPER 32GB cards appeared in much larger quantities last year. We reported in October...

TM5 is primarily Memory and some CPU. Windows Memory Diagnostics. I just toss it on top of the pile with MemTest when I'm just hitting it different ways but less intense than Karhu. If I can make it past TN5 extreme I'm usually good but many times toss Absolut on it for that extra lock down for at least 90 minutes on either.

Here's a crazy odd problem for ya courtesy of Asrock.... So remember that Asrock board that refused to boot with the $237.50 combo with the 7500X3D from MC? On the last return day to MC, I decided to roll the dice and keep it and see if it or even potentially the 7500X3D was truly defective or some other issue. I considered it a sub $250 fun project plus I didn't want to have to burn another $15-20 in gas.... I had ordered the B850i Phantom ITX from Asrock (<== major clue) and it wouldn't boot either. Swapped three different CPUs (9800X3D, 8400f, 7500X3D) and four different kits of memory. Same no post issues. Odds of two faulty motherboards is slim to none. Common denominator was Asrock. Had that HP Omen 35L HP sent me a few months back on the shelf (8400f, 5060, 16GB Rage memory, 500w PSU....such power) so I decided to open it up to use as a test bed in this "why won't Asrock work?" fiasco and of course everything works in there. 9800X3D? Sure. 8400f? Yep. 7500X3D? You betcha. KingBank memory? Boots right up no problem right along with all my other sticks. Did some googlin' and..... Issue turned out to be my KingBank 2x24GB sticks which I found out are known to have timing issues with Asrock boards even at Jedec. But wait, there's more. Testing those sticks first every time would load in the problematic timings and basically somehow wreck all the subsequent sticks I would test to the point I couldn't even blind flash the bios update on the B850i. Everything had to be completely reset/reseat and you could not start with the KingBank sticks or the same vicious "doesn't matter what sticks you put in here now, I'm not working" cycle would start again....even a single stick wouldn't boot. If I start with a known, working stick of memory and boot but then power down and install the KingBank sticks it won't work but I can at least go back to another set of sticks and they will eventually boot. It is only when I install a new/different CPU and start first with the KingBank sticks that everything goes to poo. This is with both boards running the newest 4.43 bios from Asrock.... Here is one chap with the problem, others who knew about it and Asrock offering a beta bios to fix the issue at least for his Nova board: I'm still testing and retesting to lock it down but either way, I won't be using the KingBank sticks in the Asrock boards. I'm going to swap out my TG 2x24GB 8200 sticks in my main rig and put the KingBanks in there even if they are overall inferior. If I can really lock this down and have total confidence in the MC combo for my C26 portable build, I'll send back the B850i, but we'll see as testing continues.....

Hammock - We Are More Than We Are Hammock - We Were So Young

Only sm can't support 11th, only supports 10th

from Neowin - Microsoft shares Windows 10 LTSC end of support date, and extended security update details

Ideally stable enough so it does not crash in wow with up to 80F ambient temps without any supplemental cooling or having to use max fan speed. This goes for balanced or performance mode as both have a different fan profile. Does TM5 load the GPU? I thought it was CPU+RAM only. I have not heard of WMD.

I'll test the RTX 4090 MXM on the Clevo X170 when I'll get the heatsink for it. The 4090 works great on the Clevo P775TM1 with the custom heatsink. Great temps with my backlight mod v2

Hi everyone, hopefully I'm not too late to contribute to this discussion. It's been quite a while since I installed the P4200 in my Alienware 18. Recently, out of curiosity, I decided to try something different by modifying the INF file and installing the GeForce driver instead of the Quadro driver. After doing that, I noticed my laptop started behaving strangely. The GPU was no longer performing as well as it used to, so I began investigating the issue. Eventually, I came across this thread and realized I was experiencing the exact same problem. My GPU was being power-throttled at 50W. However, I'm certain my P4200 never had this 50W power limit before. Here's what I did to restore normal performance: Used DDU in Safe Mode to completely uninstall both the Intel and NVIDIA drivers. Rebooted and disabled Driver Signature Enforcement. Installed the GPU driver first (I used an INF-modded version: 582.53 Quadro RTX Desktop/Notebook DCH WHQL). Without rebooting, installed the Intel graphics driver (win64_15.40.5171.exe). Rebooted the system. After following these steps, my laptop returned to normal and the GPU was no longer stuck at the 50W power limit. One small trick I still have to use is manually setting the fans to 100% using HWiNFO. Otherwise, the GPU will still power-throttle, even when the temperature is only around 60°C. I suspect this may be due to the VRMs overheating rather than the GPU core itself. I hope this information isn't too late and helps anyone else who runs into the same issue. Attachment is to show the GPU Power Limit and your motherboard/ec power limit.