Mr. Fox

Everyone would be using a simple die frame, like a Rockitcool or der8auer die frame, built to specs provided by Intel, just like the ILM. Should be simple, easy, cost effective, produce superior results and be better for everyone involved. The fact that they haven't already done that is a little bit mysterious. The concept of having the IHS is kind of stupid IMHO. It is an impediment to heat dissipation and using one contributes to the thermal management woes of everyone that owns an Intel or AMD CPU built with an IHS. User-induced damage is not covered by warranty, so it wouldn't (shouldn't) matter to Intel (or AMD) if stupid, careless or inexperienced people crack the dies, etc. It wouldn't be their problem, nor should it be.

As long as they continue with sockets and not move to BGA filth, bare die should be the norm. It is the norm on laptops and there is no reason it can't or shouldn't be for desktops. It would not only work better, but it would reduce the cost of production and eliminate the hassles associated with delidding.

Bummer. I wonder how much AMD and/or NVIDIA paid them to stop competing? We didn't need them to be the best. We needed them to help keep the other two honest and keep the heat turned up on the quest to maintain market share. I was actually considering supporting them with my wallet once they got their driver issues sorted because I wanted them to stay engaged in the process.

I think the bigger issue for people that are keenly interested in overclocking is the fact that nobody makes extreme performance SFF products. That makes sense because it would be counter-intuitive to even pretend otherwise. SFF, by its very nature, is founded on a concept of compromise in one form or another... price, size, performance... all the above. Some people want that. For a person such as myself, the concept is totally incompatible with my interests.

All dressed up again.

What originally drew my attention was the generic Hynix A-die @Splave was selling on HWBOT. The module (not IC) part number was HMCG78AGBUA081N BA. These are DDR5-5600 16GB modules. I have requested he take a photograph of the actual memory IC so we can see the printed part numbers on the modules. If he is kind enough to do that, I will repost the image here for reference. See image of the part number decal in spoiler below. I searched for HMCG78AGBUA081N on eBay. By accident, I ordered module part number HMCG78MEBUA081N, which is DDR5-4800 16GB. It appears these have the same IC as the common good quality SK Hynix modules we are buying under branded names like G.SKILL, TeamGroup, Corsair and others that will overclock to 6800+ depending on silicon quality. So, while it wasn't the A-die that I intended to order, it was still a truly excellent buy at $80 USD per stick. It overclocks extremely well and will be better than paying more for a branded module that needs to have the heat sink (beauty shields) removed and replaced with legit heat sinks or run naked with a cooling fan to control the thermals. Unless you're into in, paying extra for RGB is a total waste and ultimately an unwelcomed physical impediment if you want memory modules that do not overheat with crappy original equipment "heat sinks" trapping the heat. Below are images of the modules and IC close-ups of the memory in my possession for anyone that wants to search for these or the A-die DDR5-5600 variant. 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.

Well... 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. @johnksss This is actually as cool as the stick were 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. $80 each, brand new... what's not to love? And, no need to fart around with buggy nightmare RGB trash software messes just to kill an RGB rainbow puke parade. Next Steps: Back to the water cooling so I can see how far these sticks can go.

Have you done the washer mod? That should help. Or, get one of these. I have one on both systems. Screws only need to be snugged, not tightened. Also makes a nice barrier for liquid metal. https://www.amazon.com/gp/product/B0B8GHWHNP

Quantum Magnitude 1700 should also be good. I have not tested it. But, it would probably be more expensive than an OptimusPC Foundation block. https://www.ekwb.com/shop/ek-quantum-magnitude-d-rgb-1700-nickel-plexi I think even a Corsair Hydro X block would be a big improvement over what you have now. It is made with a cold plate very similar to what you had on your AIO. That would likely be the most inexpensive option. Heatkiller Pro for 1700 may also be a good option. https://www.titanrig.com/watercool-heatkiller-iv-pro-cpu-water-block-intel-lga-1x00-03-20-wc-0192-00-xx.html?color=335

I have eight Koolance QD3 fittings and two Barrow QD fittings on Wraith and three Koolance QD2 (smaller) fittings on Banshee. Banshee has two EK D5 pumps and Wraith has two EK D5 and one Swiftec D5. What kind of QD fitting are you using? Even with 10 fittings on Wraith, I still have over 200 L/h flow rate. They certainly can affect the flow rate, but the impact doesn't seem significant in my situation. If I had to take a best guess, I would suspect the water block is the issue on your system. This is a newly introduced variable after moving from the AIO to the custom loop, and it is probably the weakest link because it is designed for much older CPUs.

I recommend either EKWB Velocity2 or OptimusPC Foundation block for the CPU. XSPC Raystorm was a good block for older CPUs but I noticed it was less effective on 10th Gen, Ryzen 9 and 12th Gen than it was on X299 for some strange reason. The Velocity2 block works very well, but is less than convenient because of the rear mounting mechanism. It is inconvenient having to access the back of the motherboard for block removal and installation. It looks excellent, but results are similar to the OptimusPC Foundation block and the OptimusPC block is far more convenient to manage. The OptimusPC Signature block is more expensive than the Foundation block, but doesn't perform better. The main design and cold plate are identical. The Signature is more expensive only because it is a solid metal CNC milled "unibody" block without the acrylic top. Performance is the same. The advantage to the less expensive Foundation block is you can view any debris or sediment accumulating and have a better awareness of when cleaning is needed.

Yes, the ASROCK tool has never provided any functionality for me other than displaying current memory timings on any system. I do not know if it provides any functionality on an ASROCK board. There are some minor differences between what is displayed in ASUS Mem TweakIt and the ASROCK Timing Configurator. EVGA motherboards can use ELEET X1 or the ASUS Mem TweakIt utility for realtime memory changes. MSI can use DragonBall or the ASUS Mem TweakIt utility. ELEET X1 won't launch on an non-EVGA system. DragonBall will not launch on an EVGA system, but it worked well on MSI. DragonBall also worked on my Strix Z490 mobo and it works on my Strix Z690 board. My preference in terms of software utilities is MSI DragonBall. I wish it worked on the EVGA motherboards. I like the minimalist experience and no-nonsense GUI that it provides. MSI Dragon Ball(1.0.0.10)-intel.zip

Probably only need to tighten up a couple of things. Here are more details on the timings I am using. I can run 6800 32-40-40-28 1T but there is no performance benefit that I can identify because some of the timings need to be loosened for stability which reduces read/write/copy/latency performance. It is terribly unfortunate to see such incompetence. That guy is right. Alienware is a distaster. Dell's mainstream consumer rubbish is status quo for the industry, so there is nothing to talk about in the realm of trashbooks. We expect nothing less than a dumpster fire in that space and it makes no difference who the molester of idiocy is.

Looking good, bro. What are your read/write/copy/latency numbers looking like?

Doing more research, that part number that I purchased may still be OK. I think it is the same IC as what most of us have for SK Hynix modules from G.SKILL, TeamGroup and others already that can run 6800+ stable. Hopefully, they will be a good bin quality so I don't have to resell them. Looks like we should soon see a DDR5 version of Thaiphoon Burner. I communicated with Vitaliy back in July when I had my memory screwed up by the Unify-X mobo and provided him some memory dump info that may have been helpful. https://www.softnology.biz/news.html August 23, 2022 / Thaiphoon Burner 17.0.0.0 build 0823 beta 2 A miracle happened! A few weeks ago, the pathetic JEDEC scoundrels nevertheless published the final version 1.0 of the “JESD400-5 DDR5 Serial Presence Detect (SPD) Contents” document on their website, declassifying it. Therefore, all sorts of intellectually gifted “robing hoods” and fanatical fans of the Thaiphoon Burner, constantly swarming in it like larvae, in order to break, find out something, attach it to their programs or distribute it to the beggars and vagabonds, run at the speed of a marathon runner to the JEDEC website to download the document if you haven't already. It is surprising that only immigrants from Ukraine, Belarus and the Russian Federation still suffer from such pampering. Well, for the further development of Thaiphoon Burner, the release of the final documentation, expected throughout the year, is incredibly crucial. In accordance with it, the program code has already been adjusted. For example, the table of frequencies and timings in the summary report is displayed as accurately as possible. But there is other good news! The algorithm for reading 1024 bytes of SPD from the SPD Hub of DDR5 SDRAM modules has been successfully tested! Therefore, its implementation in the program should not take much time. In the meantime, you can admire two new screenshots. Screenshot of Thaiphoon Burner UDIMM DDR5 SDRAM module Screenshot of Thaiphoon Burner SO-DIMM DDR5 SDRAM module For Brother @Talon or anyone that wants it. You can use Sensor Panel Manager to change the image to something else, edit your GPU and memory branding, etc. Z690 Dark KPE SensorPanel.zip

Here is something useful I just picked up. Command prompt... copy and paste all on the same line. wmic MemoryChip get MemoryType, Capacity, Speed, Configuredclockspeed, DeviceLocator, FormFactor, Manufacturer, Serialnumber, Partnumber

Dang it! I thought I had matched the part number exactly. I must have purchased from the wrong listing by accident. I had identified a listing with the identical part number, so I may have to turn around and resell them if they suck at overclocking. Here is what Splave is selling (or sold) on HWBOT. He has run them as high at 7600 MT/s. https://community.hwbot.org/topic/219454-fs-binned-ddr5-a-die-hynix/

I bought the last two from the seller I selected on eBay. Just be sure you buy sticks with the exact part number and not those labeled as "compatible" (I saw a couple like that on eBay... very iffy.) Here is what I have set in the BIOS... And, here is what HWiNFO64 shows for everything. SA VID (upper left) matches what I have set in the BIOS. The Monarch RAM jackets look nice. I have Byski jackets that cost about half what the EK Monarchs cost. If you soak the memory modules in paint thinner for about 30 minutes the stock heat sinks will come off with minimal effort and no damage. Be sure to protect the part number label. I stood mine on end and only submerged the sticks in thinner deep enough to not submerge the label. You will definitely want to put a fan blowing on the modules. In the screen shot I posted above, those memory temps are with a fan blowing on the stock TeamGroup Delta RGB modules. Without a fan blowing on them, the temps hit ~65°C and throw errors in memory tests. With the fans blowing on them they pass memory tests at the temperatures shown.

I just ordered two brand new 16GB sticks of the nake green PCB Hynix (same part #) like Splave is selling on HWBOT for $80 each. I hope they overclock as nicely. They'll be perfect for my water cooling setup. Then I can either sell the RGB stuff or store them for spare parts. Just a moment and I will check for you. I have many of those set to manual/override in the BIOS and did the same on the Unify-X. Auto is generally too high and the negative offset thing just doesn't make good sense to me when override is foolproof.

A Dark mobo and EVGA BIOS can take some getting used to and some people never learn to love it. Initial impressions are always that you are getting less, when the reality is you're getting more and not being presented with a bunch of distracting garbage in the BIOS. When I stopped wanting to be an ASUS customer it was very difficult at first. But, I quickly learned to appreciate the fact that it was made to do things the way I want to do them and it turns out to be a better fit for me than an ASUS product or ASUS firmware. I now find using ASUS firmware to be very annoying and cumbersome, and the plethora of settings that I have to fart with are no longer impressive to me. I find it annoying to have to max out power, current and amp limits and turbo time values manually in an ASUS environment. I love the fact that EVGA hard codes it maxed out in the BIOS by default and then hides the menus so they do not clutter things up. The entire premise upon which it is built is balls-to-the-wall overclocking. There is no real thought about people that are gamers and the casual style that is often associated with that. They only recently added RGB/ARGB headers and WiFi and I wish they never had added them. Especially the WiFi is just a waste of system resources and PCIe lanes. I could always insert a USB dongle for WiFi if I ever needed to. I generally disable that in the BIOS. I am on the latest BIOS and it is working good for me. Disable the Guardband option and set -75% on vDroop and those two things should fix the high voltage. To unlock the PMIC past 1.435V enable "extreme voltage" and ignore the warning about high voltage. Does this does not increase voltage, it merely allows you to increase it and I always enable the extreme voltage option. This is basically the same as disabling voltage monitoring on ASUS to uncork CPU voltage past 1.500V, but it is just one toggle switch with EVGA to unlock all voltage limiters at once, rather than individually. I also use static voltage for everything I own. If you know what your load voltage was using adaptive with an offset, try setting that as the static voltage target. Also, the "CPU Voltage" in EVGA vernacular is SVID, not Core Voltage. If the user does not know that it is very alarming to think that much voltage is going to the CPU. It's not. It is what the CPU thinks it should ask for, which is always way too much. Many ASUS users disable SVID support in the BIOS so they do not see those scary but meaningless values. After using it for a while, I decided I liked MSI firmware better than ASUS as well. It was likewise difficult to get used to. It seemed lacking in some ways and yet it carried some of the same faults as ASUS firmware that I appreciated not having to put up with on an EVGA board. Gigabutt UEFI was also different than the rest, and it was one that I could never warm up to. It should be "against the law" for OEMs to use different terminology when it comes to BIOS. They should not be allowed to use an opposite approach to things like LLC (e.g. more is more on ASUS, less is less on MSI).

Yes, it's really strange. I didn't have any advice to offer because I've never seen what you're experiencing with HDMI before. I seldom use HDMI on computers anymore. I use almost exclusively DisplayPort. However, I do use HDMI whenever I use my screen capture device for BIOS videos and I haven't run into any issues there. Is your bios set on "Auto" for the setting relating to primary graphics adapter? I usually leave that set to auto (BIOS default) and haven't had any strange things happening using that approach. If that is already set to auto and Adjustable Bar Support is enabled try disabling that because it is not compatible with CSM. The EVGA Z590 Dark is the only product from EVGA I have owned that I did not love. It left a lot to be desired and I think they didn't get too excited about supporting it because 11th Gen Intel was viewed by so many customers as being a joke. It wasn't as much of a disappointment as X570/Ryzen 9, but any number multiplied by zero is zero, LOL. Anyhoo, I didn't stick with that product long enough to have any observations about firmware updates. The issues Z590 had with ACPI and the malfunctioning M.2 PCIe slots because of that ruined everything. But, I can tell you that EVGA has done a truly amazing job of providing updates on all of their other motherboards and graphics cards that I have owned. They provided more updates than I was willing to apply, since my approach is to not fix anything that isn't broken. The only ones that ever catch my attention are the ones that boast some kind of performance enhancement. Sometimes it's not accurate, but if the firmware update is for "security" and there's no mention of anything else, then that is a hard pass for me. Flashing it just because you can and because something new is floating around out there called an "update" is a truly fantastic way to end up with something disappointing.

Windows 7 installation was nearly impossible (I actually gave up) on Z590 until I used a PCIe PS/2 mouse and keyboard and SATA DVD drive and then it became painless. Trying to do it from USB was nothing but frustrating and always ended in disaster. I was able to install drivers for USB and everything else except iGPU and WiFi 6 from the Windows desktop. If you want to try restoring a Macrium Reflect image to your OS drive and then reactivate it with your own product key, let me know. It will be a huge upload/download, but it might take the frustration out of it for you. Awesome. Congrats. I really do like mine. I don't think ASUS has anything worth buying any more. Their stuff is buggy, unreliable, prone to failure and the RMA process unnecessarily challenging. There's essentially no value in their products unless you want to know a CPU SP rating.

Yes, I believe so. The frequency is TBD, but I have another component coming soon that should be fun. Stay tuned. Thank you. Feel free to comment down at the bottom if you feel so inclined. Even though they are limited to 1.435V and 6400 MT/s (due to the crippled PMIC) those Vengeance modules perform very well and for most gamers and casual overclockers that generate moisture in their nether-regions when they are exposed to RGB they are an excellent option that is priced right. Samsung B-die was the IC to have for DDR4. Other Samsung IC versions were not very good. Nicely binned SK Hynix ICs can handle 800+ MT/s overclocks. If you consider that 4800 is JEDEC default, even the comparatively crummy Samsung and Micron, and inferior bin quality SK Hynix ICs overclock a lot. Splave has cherry picked SK Hynix OEMs sticks for sale on HWBOT Marketplace that are ugly naked green PCB generic OEM modules that go way past 7000 MT/s if you want to spend $450 USD on modules that sells for under $200 for a pair of 16GB sticks. Below is an example of what you want to get if you're going the generic route.

Yes it is absolutely possible if the CPU has a strong IMC. 4000C16 is probably getting close to the limit with 4 modules though, or on a 4-DIMM mobo.