All Activity

@tps3443 If you're looking to get a known, good IMC on the 9950X3D2 over on OCN $850 USD but ships from Aus: https://www.overclock.net/threads/9950x3d2-great-imc.1819789/?post_id=29597168#post-29597168

I wish memory was not so retarded expensive. It could easily be your IMC is weak, and it could also be you memory kit isn't playing nice with ASUS firmware. It's too expensive to experiment at this point. Even if it is your memory, getting the right SKU for best results will be ridiculously and unjustifiably expensive, and sometimes very difficult to find at any price. If you lived nearby I would drop my G.SKILL 6000 CL26 EXPO kit into your Hero so you could see how it does. ASUS has some kind of goofy love affair going with G.SKILL and all of their AMD firmware development seems to be focused on making two G.SKILL memory kits sing. It's really annoying. My MSI and Gigabyte AMD motherboards have all performed well with other brands of memory and were less finicky because they haven't invested all of their time and energy in making one or two part numbers work best. I have G.SKILL 8000 CL38 XMP memory that worked fantastic in the Z790 Apex/Apex Encore and it doesn't even like to boot correctly. It was very difficult to get it tuned to run stable at any speed (including 6000). I had to jack around with all kinds of ODT and voltage resistance settings that can normally be left on Auto. That is when I gave up and bought the magic G.SKILL 6000 CL26 sticks before RAM prices turned to crap.

Well I’m only running DDR5 6000c26 daily, so I don’t have to push high VDD’s like that. But yeah 7600c38 32k trefi needed like 1.700v when I was messing with it yesterday. 1.650 doesn’t work. This gave me the feeling that IMC is weak.

There is a penalty for the CCDs just as there is for Intel with P/E cores and now their chiplet nonsense. Monolithic processor design is MUCH better. I'm not sure what gave you the impression that you are needing to give it "too much" VDD/VDDQ. Literally everyone that I know one is using VDD and VDDQ of 1.600V or higher if they are pushing high clocks. AMD will not function correctly with low VDDQ like 13th and 14th Gen Intel did. You'll need to have VDD and VDDQ either the same or VDDQ almost the same. You might get away with 50mV less for VDDQ but MSI and Gigabyte AMD motherboards set them equal by default. You have to literally manually make them different values. When you set VDD they automatically sync them. You also need to be using at least 1.475V for VDDIO (MC voltage). You can match all three without any harm. Many AMD overclockers are using 1.550V-1.600 for VDDIO as well. If I leave mine on "Auto" it goes to 1.475V and works fine. If I try to dial back the VDDIO I have instant issues. You should be able to run LOW VSOC, like 1.175V. This is the equivalent of Intel System Agent. I am running 1.600V VDD and 1.550V VDDQ at 8000 and 1.650V VDD and 1.600V VDDQ at 8200 and 1.175V SOC for both of my AMD platforms. Running less makes them unstable. Increasing VSOC does nothing. The hardest part of the adjustment to AMD for me was having to deliberately set aside my knowledge from Intel overclocking because it was really messing things up and making it impossible for me to get any overclock stable. What works for one seldom works for the other. AMD is also funny on CPU core voltage. If you are stingy with it you might not see BSOD and the normal tell-tale signs you do with Intel, it will seem stable but performance will drop like a rock. Cinebench scores will go down. If you go too far you will see the same behavior as Intel with crashing, but AMD is weird and I suspect a lot of people are living with lackluster Ryzen performance from trying to undervolt the CPU too much.

Seeing how well ram tuning posting and stability responds to adding memory voltage, I am definitely pointing to a weak IMC. It’s like 6000c26 boots fine with auto 11K tREFI with 1.435v VDD. EASY EASY. 32K tREFI? It won’t attempt to post, adding 32k tREFI puts a much heavier load on IMC, but adding voltage to VDD. Quick and easy bump up to 1.450v VDD gets it done! 65k tREFI fully tuned timings? Similar situation. I need 1.480v VDD. Probably can get away with 1.460vdd. But it’s all too familiar! And it put me right back in the seat of tuning on Intel with DDR5. Weak IMC’s loves high or too much VDD/ VDDQ. And that’s my system to a T lol. 😂 @Mr. Fox I’m going to swap it out for another. I’d like 2,200FCLK AS WELL. I’ll see how the next one runs. It’s a BEAST in games doing one thing at a time producing a high FPS though. That’s its main benefit. It’s not as fast as my Xeon doing brute work handling multiple things at a time. Like leaving a game minimized and doing other things simultaneously etc. it’s just not as fluid as my 28/56 Xeon 3175x is. That could be partly due to lower bandwidth, maybe higher latency, or merely due to being two CCD’s? Either way. Gaming performance is great! And that’s what I wanted to see.

After buying an x170sm-g and now needing parts for it any not being able to upgrade to an RTX580 or 90, I am done with Cleveo. I reverted back to one of the i7 laptops I have from 2012 that still has replacable parts. I am going to try and hold out on a new performance laptop , I really don't know how they can be called performance laptops with mobile CPU and a max 55w CPU. Tablets will be more powerful than that eventually. Since computers are just money dumps with worse depreciation than a car. I am not going to upgrade till PCI-6.x , DDR6 and 128gb or 256gb so-dimms come out. No more over priced paper weights with BGA. Maybe they will come with 17 or 18 inch OLED 8K 240Hz screens then too. Who knows. So I guess that will be around 2028-2030 so 2-4 years out.

I need a donor X170-SM-G for parts to repair mine, not looking for a working one, just a cheap parts machine. I don't want to pay Eurocom $25 each for 4 sets of screws and $50 shipping to the US and any other parts I might need so I am looking for a dead parts machine if anyone has one sitting around or knows of any. I only saw KM-G's on ebay and alot of those parts are slightly different than the SM-G.

I have heard several people with 9950X3D2 complaining about memory overclocking being limited by a weak IMC. Whether that is relevant I can't say, just mentioning it as a potential common complaint. Your CPU might not be below average for a 9950X3D2.

Had to make a place for the Z890 Apex. X870E Apex is now installed in the O11D Mini V2 with the 4585PX and 5080. Getting it to fit was a HUGE pain in the butt. Never buy a distro block with the pump facing the inside of the chassis. Big mistake. I have the Alphacool reservior shoved tight against the roof of the case and the 5080 still touches it. With the reservoir against the floor the GPU will not fit. Always choose one with the pump that is in the closed in area behind the mobo tray. (That will also allow you to run the system laying on its back. You can't do that if the pump is pointed up toward the ceiling.) Now that it fits, it looks great. I love this case for mATX/ITX. For ATX it's not ideal unless you love having to spend a lot of time figuring out how to make things fit. (I don't enjoy that. I also hate puzzles.) The Z890 Apex is in the O11D XL EVO with the 5090. It was super easy because all of the connects and mobo layout on the Z890 Apex is the same as the X870E Apex. I only had to modify the two lines going to the CPU block because the orientation of the ports on the block was different than the Thermal Grizzy Mycro direct die block. Speaking of waterblock, I was going to use one of my Optimus Foundation blocks. The silver looks perfect with the white and silver on the mobo. I could not use it. CPU would overheat SEVERELY for some reason (109°C doing CPU-Z stress test). I disassembled it and found it spotless on the inside. Not sure why. Had to run to Microcenter this moring and grab an Alphacool Core 1 block. Now it works amazing. Almost don't need to delid now, but still will. Because I can, and it will be much better. Not because it is unavoidable due to stupid temps. It runs as cool as the bare die Ryzen 9 processors. The B850MPOWER is now on the open bench with the 9950X and 3070.

That's more like it! Once you lop it's head off, I think you'll be able to break 50K which is wild. I saw someone post a 270K+ with 52K CB23 over at OCN today. Nova Lake is going to go hard with it being Intel's second gen dielet approach. They've learned quite a bit about it after the Arrow Lake fumble. Intel engineer said yesterday on Reddit that the E cores are reportedly as fast as RPL/ARL P cores.

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.

It's possible to have a weak IMC on any AM5 chip. Remember, these are only "officially" rated to handle 5600 and I had that 9800X3D which barely could do that before returning and capped out at 2000 fclk on the nose. Nothing would push it over. Order another and/or another motherboard to sort out what is holding you back. It could also be your memory too. I went through a few boards and sets of memory to realize that first 9800X3D was the culprit. Second one ( current ) is much better. ---------------- You could always return everything and go 270k too for shiggles and probably save a bit of coin on top of it.

I think my 9950X3D2 has a weak IMC. Different bios doesn’t help. Max 1:2 speed I can boot is DDR5 7600. Either weak IMC, or my 6000c30 A-die kit is just at the limits? I’ve tried like 1.5v VDDIO LOL 😂 1.300v VSOC UP to 1.650v VDD/VDDQ. Nothing helps. 7600 max, and that’s hit or miss, 7400 posting works every time. I tried to use some preset timings. And that did not help either. Is it possible to have a crazy weak IMC on my 9950X3D2?

@dude-137 Based on my personal experience: Using "Balanced" Windows Power Profile when idle - clock set to 3.60 GHz the temp fluctuates around 60 degrees. "High Performance" profile will lead to a gradual overheating of the PCH up to ~80 degrees - clock at 4.68 GHz. CPU core voltage offset around -72 millivolts. In my opinion, the voltage of the power rails of the PCH should be measured. I have never modified the cooling system at all and still I am far from 100 degrees (also many VMs are running synchronously).

Windows 11 installation is no problem. Just use "Rufus" for building the install media and remove the 11 specific requirements in the settings. My DC came from a defective 8570w bought on ebay. Motherboard had failed (shorted chipset). i7 3820, K2000M and the DC screen were still good.

Did not even think of the 25H2 CE (which I am running) being the culprit, wow. Fitting though..... On the HP Omen, it is running the OEM cookie cutter 24H2 Home edition. Scores looking much better! When you lop off its head for DD cooling, that will kick it up a notch nicely.

https://hwbot.org/benchmarks/cinebench_-_r23_multi_core_with_benchmate/submissions/6041469

Hi All, Another 8570W user here, Whilst not running Dreamcolor i have a Win10 build with K1000M (being upgraded to 980M) currently running .71Ver System bios. How difficult was win11 to get onto the machine as i recall TPM scuppers the build/install. Where did you find the Dreamcolour setup?

BINGO! 25H2 Cancer Edition... 100%. Installed LTSC IoT 24H2 and viola... all is well with the score. Fully modded and debloated. But, I still do not know why my effective clocks are lower than the core ratio set. Looking at similar submissions on HWBOT the core ratio and effective clocks were the same on other submissions. But, no throttling or temperature issues. Now I can delete the cancer Macrium image. Side note @electrosoft - no driver issues with 24H2. Smooth sailing. Not too shabby for a chintzy Lian Li 360MM AIO and soldered IHS... Nearly identical Cinebench R23 performance as bare die 9950X (only cooler than a delidded 9950X) with only 22 threads (16 baby girl Atom cores) versus 32 threads on the Ryzen 9. Now I need to order a Thermal Grizzly delid tool and heater and bare die block and get it on a custom loop and chiller to do things right. https://hwbot.org/benchmarks/cinebench_-_r23_multi_core_with_benchmate/submissions/6041409

I bought both part numbers of dmc cards along with a antenna for 17 r1. my m14x r2 already has an antenna so I can test both cards

His bios was over 12MB already and it uses a newer Intel ME. It's not "simple" to switch from DP_D to DP_B in the laptop, if it was it would be done already. Again everyone who only thinks in hardware takes software way too lightly. I'm a hardware person myself, but after these last few months working on bios modding, I see how tedious it is just to implement one thing. For me this is low priority for now. Once I have done my other mods I can take a look, but edid being a simple issue, it's not much of an issue to me. I'm going to make a simple windows program that autofixes and autodetects this in the future, but that's after I get my 4090 heatsink and swap the eDP panel for my 17 R1.

EDID is just a script that executes at each startups. We just want to use back the DP_B, that's the annoying part, that's it literally it. The rest is usable. Could try on the clevo p570wm. It has a big chip, 64MBit and there are some space left.

ASUS did a much better job of PCIe lane allocation on the Z890 Apex than they did with the X870E Apex. It has two usable PCIe 4X slots (instead of just one) and neither one affects the GPU in terms of dropping it down to 8X instead of it running at 16X as it should. Only the bottom 16X slot steals GPU bandwidth. The two Klevv drives are running Gen3 because they are installed in a Gen3 PCIe NVMe card. It sucks that the only slot that does not affect the GPU on the X870E Apex is the one situated above the GPU. I have room for 4 more NVMe but have run out of 1TB and 2TB spares to install in it. Right now it is on a test bench with a spare 3070. Have to decide which chassis to move it to.