Mr. Fox

Looks like the seller I bought from now has two more sticks of SK Hynix M-die for sale Here is more info on A-die. Generic SK Hynix HMCG78AGBUA081N module part number. The memory IC part number for A-die is H5CG48AGBDX018. The JEDEC clock speed for A-die is 5600 versus 4800 for M-die and it seems like the increase in overclock headroom is similar. A-die is not in full production yet and modules using the IC are difficult to find and very expensive. Next generation DDR5 ICs Review: SK hynix 16 Gbit A-Die - 2x 16 GB with overclocking and gaming WWW.IGORSLAB.DE Today I have a glimpse into the near future of RAM technology for you, and as is so often the case in the computer hardware industry, it comes from China, where

All dressed up again...

Just be thankful that humans don't have to eat the afterbirth.

Below are images of the modules and IC close-ups of the memory in my possession for anyone that wants to search for these online for a cheap price as I found them. As you can see from the Thaiphoon Burner data, the generic modules I bought on eBay look to be the same IC as what is on the far more expensive TeamGroup Delta RGB modules, and probably other brands like G.SKILL and Corsair. The generic modules I purchased are Hynix part number HMCG78MEBUA081N. These are M-die IC # H5CG48MEBDX014 DDR5-4800. The Hynix DDR5-5600 A-die module part number is HMCG78AGBUA081N. I have asked @Splave for a photo of the actual memory IC # and if he is kind enough to share that I will put that information here. Having the IC # can be helpful when purchasing memory modules online when the IC is visible in the photos, because the memory module part number is vendor-specific.

Generic naked green SK Hynix DDR5-4800 modules with no XMP profiles for $80 each on eBay brand new that overclock as good, or better, than fancy schmancy name-brand modules that sell for twice as much.

I saw @Splave selling generic Hynix sticks on HWBOT Marketplace for $450 for 2*16GB binned samples. A little too rich for my blood at that price (especially if it gets the SPD corrupted as mentioned above) I found similar modules brand new on eBay and snagged a pair for $80 each. I received them today. Here is what I discovered in my first 30 minutes playing with them. As luck would have it, these naked generic DDR5-4800 SK Hynix modules with no XMP that I purchased brand new on eBay for $80 each overclock better than my TeamGroup RGB sticks, LOL. And, totally naked with nothing but a fan blowing on them they run a good 15-20°C cooler than memory with stock trash heat sinks beautification covers on them. What a wonderful find at an affordable price. Using exactly the same manual settings they perform better. A few moments later... BOOM - there it is... These are actually running as cool as the Delta sticks were when running on water. It seems like the stock heat sinks we are getting on the overpriced brand-name memory are actually trapping heat rather than shedding it. Conclusion: If you are looking for performance rather than aesthetics, then don't waste your hard-earned money on branded sticks with worthless beautification covers inappropriately labeled as "heat sinks" LOL. Next Steps: Back to the water cooling so I can see how far these sticks can go. $80 each, brand new... what's not to love? And, no need to fart around with buggy garbage RGB software messes.

I looked around to see if we already had one and did not find a thread for this, so I decided to start one. With DDR5 being the direction things are heading, it's time to shift our attention away from DDR4 to the nex big deal. This has been quite a weird journey for me. Everything I knew about overclocking Samsung B-die turns out to be mostly worthless now. The new bully on the block is S.K. Hynix. I am by no means an authority on the subject, but I will be glad to share what little I have learned with those just getting started, and eagerly learn from those that have learned more than I have. What I meant by weird journey is I have had three DDR5 kits that have had their SPD corrupted and needed to be replaced. All three kits got screwed up on a Unify-X motherboard and I have seen reports of the same thing happening to ASUS Z690 owners. The kits in question were two TeamGroup Delta RGB 6200 kits like the one reviewed by Brother @Avacado and a Corsair Vengeance (non-RGB) 6400 kits between the two. I am on my third TeamGroup Delta RGB 6200 kit now, crossing my fingers and hoping it doesn't happen again. Fearing I may encounter the same circumstances again, I replaced the Unify-X with an EVGA Z690 Dark K|NGP|N motherboard and plan to sell the Unify-X. We shall see how it goes and I will report on the experience here.

It is, indeed.

That's going to look really sweet. Can hardly wait to see the finished product.

I know I have said it before, but I love how that workspace looks. It is a work of art. The choice of music is also a good fit for the aesthetic.

Battle of the one-cores, LOL. Is the voltage accurate on the T-Bird CPU, or is CPU-Z bugged reading that as well?

Congrats Brother @Avacado. Well deserved and good choice by @ENTERPRISE, too. One of these days I am going to have to chat with you and find out what prompted your choice in user name. I bet there is an interesting story behind that. Coincidentally, with no reference to that other than coincidence, and excluding the funky color of 70's era kitchen appliances, I consider avocados a basic staple. Hardly a day goes by that I do not enjoy one and guacamole (actually almost anything Mexican food, but always guac) is my favorite.

That is so true. Some (many) are just too lazy and others don't even know how to read. They just look at pictures.

Thanks for updating the leaderboard. The thing that is weird about this benchmark, and maybe why so many do not respect it, is that is can be misleading if you do not understand what it represents. The overall percentile only applies to matching hardware and does not mean it is faster or better than different hardware. The idea has merit in the limited framework and can be valueable in the right context. But, for those that engage in overclocking as a sport it is contrary to logic. You could use the individual hardware percentiles rather than the overall system percentile to accomplish what we are used to doing with benchmarking, but it would still only apply to matching hardware. I could see some people that do not know much about computers, and especially not knowing about benching, becoming very confused by this one. They could think that their system with a 10600K and 1060 is faster than a high end CPU and GPU because of it having a better percentile rating. It faced criticism for bias claiming that it falsely makes Intel look better than AMD, but that is baloney because it doesn't compare dissimilar hardware. The error is in the flawed interpretation and misunderstanding of the results. It is also easily manipulated. The overall percentile rating is affected by everything, which makes the overall ranking percentile a lot less meaningful. For example, I can move from 72nd percentile to 93rd percentile by simply using Disk Management to remove drive letters for all of the slow SATA HDDs and SSDs and leaving only the NVMe SSD visible to Windows. It doesn't test the drives it can't see and your percentile score increases dramatically. The performance of the system is the same regardless of the overall score. PCMark benchmarks never got any real traction for similar reasons to the extent that things that don't matter in real life can have a huge effect on the overall score. But, it was useful for comparing dissimlar hardware as long as you didn't have an element that doesn't matter dragging the score down and potentially leading to false conclusions based on the overall score.

Here is one from an older beast that I used to punish... CPU and GPU in 100th Percentile 7980XE @ 5.2GHz (100% OC) and 2080 Ti FTW3 with shunt mod and 2000W power limit, 2325MHz Core, 8300MHz mem (that GPU is now in Banshee) https://www.userbenchmark.com/UserRun/27314580

Forgot about running this... 5.5GHz all P, 4.3GHz all E, 4.6GHz cache, GPU 2265Mhz Core, 11,352Mhz mem (+1600), 1000W KPE XOC vBIOS https://www.userbenchmark.com/UserRun/53908389

The EK Quantux Velocity² is nice looking. It fills the empty space between the VRM heat sinks and memory. It is made specifically for LGA-1700. I got this from Titan Rig open box for about $40 below retail. It works better than the OptimusPC Foundation block and has a MUCH higher liquid flow rate. The OptimusPC block is a lot more restrictive. Temps are close to the same, slightly better (3-5°C lower core max in Cinebench). The only downsides are the back of mobo installation and fixed position of inlet and outlet.

It needs to be delidded and run bare die with a der8auer die frame. I used to bench my 7960X (16C/32T) at 5.3GHz all core and my 7980XE (18C/36T) at 5.2GHz all core in my X299 Dark mobo. Both CPUs used to pull over 1,000W in Cinebench. It was too stinking awesome for words. That was one of several reasons I found owning a 5950X to be such a massive disappointment. I expected the same, but it was not capable of coming close to that, even with chilled water cooling.

On business trip right now but I'll try to remember to run it when I get home Sunday afternoon.

That is an X299 LGA 2066 HEDT processor. https://ark.intel.com/content/www/us/en/ark/products/198019/intel-core-i910900x-xseries-processor-19-25m-cache-3-70-ghz.html

Now that I think about it, R23 may have been introduced to block Windows 7 users from opening a can of whoop-ass on other (newer) Windoze users. R20 runs like banshee on Windows 7. R23 won't launch on Windows 7. One of the challenges the Redmond Mafia has had getting people to move to their newer trash OSes is Windows 7 outperforms everything they have released as a wannabe replacement. They've had 7 years to get Windoz 1x right (10 years if you include Winduhz 8.x) and success seems to be beyond their skill set.

Thanks, Bro! Other than having a "fluffy" higher number scoring mechanism to psychologically influence the kiddos, R20 and R23 do not seem much different from a benchmark perspective.

OK, the damage is done. Delidded with liquid metal on both sides of the IHS and NO CHILLER in this test. 55x all P-cores at 1.410V override. I'd call this a major win. Compared with the temps shown above running with the chiller and KPX thermal paste before delid... yay.

Here is another one. 908.2 single thread/13089 multi thread. 56x all P, 43x all E, 46x cache.

I am definitely going to have to delid this bad boy. The temps are just way too high with the stock solder, even running on the chiller. The voltage needs are nice and low. https://hwbot.org/submission/5029695_ /img/logo.png Mr. Fox`s Cinebench - R23 Multi Core with BenchMate score: 31445 cb with a Core i9 12900KS HWBOT.ORG The Core i9 12900KS @ 5500MHzscores getScoreFormatted in the Cinebench - R23 Multi Core with BenchMate benchmark. Mr. Foxranks #null worldwide and #11 in the hardware class. Find out more at... https://hwbot.org/submission/5029753_ /img/logo.png Mr. Fox`s Cinebench - R23 Multi Core with BenchMate score: 31711 cb with a Core i9 12900KS HWBOT.ORG The Core i9 12900KS @ 5500MHzscores getScoreFormatted in the Cinebench - R23 Multi Core with BenchMate benchmark. Mr. Foxranks #null worldwide and #9 in the hardware class. Find out more at HWBOT. Side note: My delidded golden 12900K is also working nicely in Banshee. I am using it for work, running stable and cool at 53x on P-cores, 42x on E-core, 43x cache with 4000 15-16-16-36 2T G1. It's nice having the luxury of using a second system running in 24/7 beast mode for the day-to-day work of doing my job.