Mr. Fox

By all means check with the liquid metal to confirm the improvement, but I would strongly recommend getting the Ryzen Mycro Pro direct die block from Thermal Grizzly at some point anyway when it can be worked into your budget. It is ideal and designed exclusively for direct die application. It's near perfect and it was engineered to be. The same applies to the IceMan Cooling direct die blocks. They're not engineered with the same level of precision and attention to detail as the Intel Mycro Pro (I have both) but still work extremely well (essentially identical to the Mycro Pro). Either option on my LGA-1700 systems is far and away better than anything else.

Same here. I have no issues whatsoever being happy with SATA SSD speeds for OS drives or frequenty-access storage, or SATA HDD for occassional-access file storage, such as drivers, software installers, OS ISO files, Macrium drive image backups, etc. NVMe SSD speeds (even Gen 2) are nice to have but anything above Gen 3 is not what I would consider necessary and not worth paying extra for. The only thing I see as a drawback with SATA is the extra cables and the added physical space it occupies inside of the computer. What I care about most is having storage that is the least expensive option and reliable. Average mainstream speeds are totally sufficient. This 6-second snippet was my favorite part of the video.

Very informative video that cuts through all of the silly nonsense and deceptive marketing hype associated with storage. PCIe 5.0 is really a waste that steals resources from motherboards and attached components that could be put to a much better use (extra M.2 slots, SATA ports and PCIe slots).

💯old = reliable and stable / new = inconsistent and compromised

I wish MSI would get their firmware sorted. They've confirmed they know "per CCD" overclocking is broken, have released numerous beta BIOS updates, but have yet to fix it. The only way I can set the weaker CCD at a max stable clock and max out the stronger CCD is by using Windows software. I really hate that, and Ryzen Master is a crappy piece of rubbish. This function worked on the Taichi, but memory overclocking on the Taichi really sucked. The Safe Boot feature on the Carbon is also broken. Safe Boot will work on one BIOS version, then they'll break it on the next, then fix it again. As much as I hate A$$zeus I may have have to ignore what a sorry outfit they are and order a Gene from China and send the Carbon back to NewEgg before my holiday return window closes. I really like the Carbon, but don't know why they can't figure things out on the firmware. (These are things the PBO gamerboy sheeple would probably never know were broken.) https://hwbot.org/submission/5725591_

Intel mesh would likely see solild gains in the same way FCLK (Infinity fabric) overclocking helps on Ryzen. There is not a lot of information to be found online because overclocking is being slowly killed by retarded architectural trends and engineering defects. I always thought it was difficult to find good information for Intel non-HEDT overclocking, but it is far more difficult with AMD to find any useful info. There is less on HEDT to be found most likely because of the comparatively astronomical cost of the motherboards, CPUs and extra memory modules making the purchases of those parts cost prohibitive for overclocking enthusiasts. The only way to get into it without it costing an arm, leg, one eye and the soul of your first-born is to wait until the parts are obsolete/EOL products. EVGA Dark and Classified motherboardd included an option in the BIOS to disable the Intel ME and I always disabled it because it is totally worthless and unnecessary. It should be a standard BIOS feature on all motherboards, all price ranges. A side-benefit that some people care about is computers with Intel ME disabled are more secure. EVGA did, and ASUS does, provide an option to select micro-code versions in the BIOS on their enthusiast-tier consumer motherboards. I would hope the same should be applied on HEDT boards. If not, there should be. I do not remember seeing options to choose micro-code version in the BIOS prior to LGA-1700.

I am excited to see how much it drops your temps as well. It always has been unnecessary and a bad idea. ASUS is the worst about bundling it in the BIOS cancer updates rather than letting computer owners decide if they want the new ME cancer. It's stupefying and baffling that there are people that go out deliberately looking for ME updates to manually upgrade to the newest feces to slow down their computers.

I don't use it anyhow. No need for silly bloat like that IMHO. I felt the same way about GeFarts Experience and do not install it. There is nothing I can identify to suggest a good reason for me to install the new NVIDIA App or GeFarts Experience. I am very sad with the thought that the no-nonsense legacy NVIDIA Control Panel will be going away. It's going to suck having to install the App trash

Yes, they are totally legit. But, they are primarily a local storefront. Many things they sell are not available for shipping, and must be picked up in person. Much like MC only a whole lot smaller. One thing to be aware of, if you were not already... if your motherboard does not have Resizable Bar Support the Intel GPU may not work correctly. Intel's GPUs must have that functionality to work correctly. I do not know if your HEDT motherboard has that or not. Totally accurate. They are not a 24/7 e-tailer focused on internet sales, but I have found them pleasant to work with and I had no issues at all with the trash bin 14900KS I returned to them because it was rubbish.

Had Core Ultra included hyperthreading it is somewhat unlikely I would have purchased the 9950X. The lack of it was the nail in the coffin for 285K. It made the CPU a eunuch on face value. No testing necessary. Game over. It was as if Intel deliberately forfeited the match and simply handed AMD the win. Bastards. So, speaking of 9950X, @Raiderman CPU delid is done and it's ready to send it back home again. MSI released BIOS v1.A1G for the X870E Carbon. This platform was very sketchy using 24GB memory modules. I can now run my TeamGroup Xtreem 8200 48GB kit on it with the memory tuned at 8000 manually. It will boot 8400 and run 8200 with some minor errors that I could likely tune away if I wanted to waste an inordinate amount of time trying to. There is not a lot of difference between 8000 and 8200 or 8400 and 8600 on Intel. There is even less difference in those steps on AMD. It's kind of freaky how 8200 and 8000 are nearly identical on the 9950X. Read/write/copy and latency don't really change or improve with the higher clock, so I am probably not going to waste time trying to tune 8200.

That's odd. I have not experienced either of those issues and I use Zoom at work and it works fine. Did you get the out of GPU memory playing games? I don't play games a whole lot, but haven't run into that with COD or BF, Crysis titles, etc.

I mentioned a couple of times the PCIe quad NVMe card that I use in the bottom PCIe slot on the Carbon with no effect on the GPU, but could NOT use on the Taichi without the GPU falling to x8. So, Brother @cylix even if the Taichi were not an E-ATX, being able to use all 4 NVMe slots at the expense of forfeiting the viability of using the bottom x16 slot makes the Taichi is a horrible trade-off and that makes it an instant loser. (Nevermind the fact that it sucks at memory overclocking and has horrid memory latency compared to the Carbon.) The Carbon has 1xM.2 Gen5 above the GPU slot and 1xM.2 Gen 5 below the GPU. It has two more Gen4 M.2 slots, so four in total. If you use the M.2 slot under the GPU slot it will cut the GPU slot to x8. So, I leave that M.2 slot empty and use the bottom PCIe x16 slot with the Sabrent card (Quad NVMe) with no effect on the GPU. I give up one M.2 slot for the luxury of adding three move NVMe drives. So, I have seven NVMe drives in use on the Carbon and the GPU is still x16. This is what I use. I have one in the Z790 Apex as well. SABRENT 4-Drive NVMe M.2 SSD to PCIe 3.0 x4 Adapter Card [EC-P3X4]

I am finding X870E Carbon 1A1B Beta (from svet as MSI forum) is working as well as 1A14 or maybe a little better, but only if I leave PPT set to "Auto" instead of disabled. Disabling PPT seems to adversely affect memory training on AMD. It reduces latency on Intel. https://hwbot.org/submission/5717708_

Fanboys of any sort are hard to respect. I have a lot more respect for people that have no issues calling balls and strikes and are not afraid to tell someone their baby is ugly, as long as they give credit when it is due regardless of brand. There is a list of pros and cons for everything.

https://browser.geekbench.com/ai/v1/155172 | https://hwbot.org/submission/5716615_ 3DMark11 Physics Only (CPU) https://hwbot.org/submission/5716585_

That's interesting. I wonder why? Production issues? Or a NewEgg issue? Edit: It looks like a production (stock availability) issue for both Ryzen 9 CPUs and the 9800X3D. AMD has only the undesirable Ryzen 7 models and below available on their web store. The good CPU supply is gone. https://wccftech.com/amd-ryzen-7-9800x3d-cpu-wider-availability-next-quarter/

Always measure results, not marketing information. I am interested if the consistency (viscosity) is better and anticipate the cooling will be what it always is for liquid metal. Darn it, you caught me. I need to update the logo.

I agree. Silly to pay a lot for something like this. I recently ordered these. Looks like they're essentially the same with a different brand name on the outside. They are tightly spun and less fuzzy than regular cotton swabs. https://www.amazon.com/gp/product/B09WDH7TW3 I will post something when the time comes. They offer a 5 gram tube but it is over $40. That is a ton of liquid metal and a reasonable price considering how much, but I ordered the 1.5g for a whole lot less money just to see how it works before I order a ton of it. I pay less attention to the W/mK ratings than I used to because they are often not an accurate indicator of how the product works. I think some of the companies in China just make stuff up and put it in their product information because they know they'll never need to provide proof to support what they print. I tried some thermal paste recently that had a very high W/mK rating and it was rubbish. I got it at no cost and will use it for temporary application purposes on something I know will be taken apart again.

That's pretty sweet. Loving that Cinebench score. Show us some photos when you can. Try setting the XMP profile and increase the voltage to like 1.500V for the memory and see if that works. I find XMP profiles often need more voltage than what they set by default.

I have not seen this liquid metal before. Looks very interesting--especially the higher viscosity part--so I ordered some to try the next time I need to take something apart. It seems like over the years that liquid metal formulas have changed and have a lower viscosity than they used to, and I think that makes it more likely to drip or run. (The original Liquid Ultra was much thicker than Conductonaut, then they changed it and made it extra watery.) It looks like they use more indium and silver in their formula. https://www.amazon.com/gp/product/B0CJ9QGDFV They have some other products on their web site.

Happy to do it, brother. So, here is an interesting story. The Z790 Edge system that I use for work has been randomly having an issue where it would hang during POST and just get stuck with the white LED on. After several attempts to boot it up it would always fail the same way, but then it would always suddenly decide to function normally right before I was about to give up. This has been going on randomly for like 90 days. Really has been annoying. So, Friday evening and multple times Saturday the Carbon starts doing the exact same thing. I'm like "what the heck is going on" and so I flipped the KVM over to the Edge and it's doing the same thing, at the same time. Turns out it was my Roccat Vulkan 120 keyboard. I use the same keyboard, mouse and monitors for those two systems and switch back and forth using the KVM. I tested with a cheap old keyboard I hang onto for stuff like this and it worked fine with both machines. Hooked up the Roccat again and instantly both machines start doing the same hanging during POST again. So, I grabbed a Corsair K70 Core at Walmart and I'm back in business. I got it because it was cheap ($65) and better than any of the off-brand Chinese trash on Amazon for the same price. Turns out the K70 Core is a really nice entry-level mechanical keyboard. And, one nice thing about it is once I program my bright white backlighting and save it directly to the keyboard I am done. No need for crappy software bloat to manage the rainbow puke. The keyboard has three onboard profiles and I made them all the same thing so that it doesn't matter what profile is loaded... the outcome will always be white LED at max brightness. I like it better than my older and more expensive Corsair keyboards that I used for years until they finally died. The old Corsair K's had a habit of forgetting my settings and reverting back to the spastic animated rainbow puke. The Core also has a nice tactile response without being overly clicky. In fact, it doesn't click at all. It mostly has a dull "thud" sound. I like it. (It has Corsair MLX Red linear switches.) I guess I'll throw the Roccat Vulkan in the trash. That's going to make some people happy that hated how clicky and loud it was. It was quite annoying to some folks during teleconferences.

It gave me ample time to wonder whether I was making a terrible mistake. I am glad it was not, LOL. 🤣

That's absolutely awesome, bro. I am excited that you are getting into that. I look forward to seeing what kind of cool creations you come up with. LOL, not really but I am willing to help a friend (any friend) when the opportunity exists. If I do too many I might develop an repetitive stress injury. To get the IHS off of mine I had to shift the IHS back and forth like 75 or 80 times. I can see why Roman uses a cordless drill with an allen bit now. I am not going to do that, but the thought did cross my mind when I was tightening and loosening the screws that many times. A one point I wondered if it was going to actually turn loose. Then it did. It moves such a small distance each direction that it is not hard to understand why it was tedious. It didn't make any unsettling noises like I sometimes hear when shoving an Intel IHS off the die in one uninterrupted motion. It is absolutely worth doing though. The thermal improvement is no less dramatic than it is with an Intel CPU. Maybe even a little bit more since the CCD chiplets are so small. It's too bad that was not true with the 5950X delid. Part of it is likely the Mycro Pro direct die block. That thing is amazing.

I obviously cannot speak to the W790 platform based on actual experience, but I think the logic would remain constant, in that 1DPC will always be better than more per channel in terms of overclocking and latency simply because of the extra traces, longer traces and noise the extra channels add that impair overclocking and add latency. The higher the memory clock goes, the more harmful the added DIMMs per channel become. The mere existence of the extra slots is what is detrimental... moreso than whether or not the slots are populated with memory sticks.