*Official* NBR Desktop Overclocker's Lounge [laptop owners welcome, too]

Did you try using Linux terminal commands to fix it as shown in that link I posted (above) already? I was literally going to throw all three away and tried it just for giggles. I didn't expect anything and was shocked when it did.

 

OK... based on things I am reading in HWBOT forums, OC.net and other venues that is definitely the case and nobody running a R9 5950X can run with FCLK above about 1800 max without having WHEA errors, random shuts downs, etc. because AGESA is utter feces and totally runs and otherwise nice piece of kit. It is not an "ASUS problem" it is an AMD problem. Looks like the only remote possibility of having a 5950X that is stable and static (meaning not working today, but not working tomorrow,) is to BIOS flashback to a very early firmware package (because normal flashing will not allow a downgrade that far back in time). Dang. I hate when I buy buggy rubbish. TONS of people are having problems that seem worse that what I am experiencing unless they leave the FCLK on Auto and leave untapped performance on the table.

 

So the 5950X is stuck at 1800 whereas the 5800X/5900X can do above that? I'd have to test that because I have only been testing it at 1800 for the moment.

 

LOL. It is going to be interesting how dedicated Intel OC'ers take to Alder Lake. I wonder if Intel will allow the smaller cores to be disabled.

 

I just dropped back to the first BIOS for 5950X with the oldest available AGESA (1100 patch C dated 11/11/2020) to see how that does. I had to use BIOS flashback to get the motherboard to take it.

Edit 1: Need to test more to confirm, but it already seems better with the oldest available firmware, LOL.
Edit 2: Nope. This introduced some weirdness in Windows.

 

I have three of those PCIe to NVMe cards and they're awesome. Nothing fancy, very cheap, but work flawlessly. It is a great way to handle a crash test dummy OS as long as you do not run out of PCIe lanes and end up with your GPU running at 8x. I really miss that about 7960X and 7980XE. I never ran out of lanes.

 

Nope: This introduced some new weirdness in Windows, like installers hanging right at the end and programs becoming unresponsive. *sigh* crap.

 

Ok something cool about the 3090 Kingpin.

You can start a Port Royal stress test, let your card load up, set an overclock, and start flipping dip switches while running. It’s great! All in real time too.

Like, I always powered the card off to mess with these. But I just did that for safety purposes I guess. Because it registers immediately! It is just like a light switch. This really helps to find the best balance while overclocking.

Just figured I would let anyone know about this. (I thought it was potentially dangerous)


But I have flipped every switch on and off while the card is rendering at full load. And it registers immediately! Not a skip, not a pop. No sparks, no bad happens.

No issues at all. I was the guinea pig.

 

I need one of those crash test dummy OSes you speak of. Instead, I’m the dummy and I use my daily OS for these things lol.

 

I found the early revision BIOS versions gave me better numbers on my 5800x but had so many quirks and bugs and obviously the USB issues that the updates were worth it. I know there were rumors the reason the later Agesa updates produced lower numbers and
much tapered PBO negative values and PBO boost range (neutered down to +200) was because chips were dying. I just ran across a bunch of anecdotal evidence of chips crapping out but nothing concrete or statistically significant.

 

My "crash test dummy" OSes are those I install and let the Redmond Retards have their way with updates and filth added. I call them that because I don't care if they work right or perform well, I am just wanting to see how bad they get screwed up by Micro$lop and use it as the daily driver. Kind of a social experiment. The OSes I care about a tweaked a bunch and have updates blocked so they don't screw it all up for me.

The way mine shuts off I could see that rumor having some truth to it. I have given up on trying to match FCLK. It's too unstable even though it improves memory latency a lot. I did find in the BIOS where to set UCLK and FCLK differently and that has greatly improved my stability. No more WHEA errors showing up in HWINFO.

 

Improving. I don't know if I can get it cold enough for 5.0GHz, but I am not done tweaking. The liquid metal helped a great deal.

https://hwbot.org/submission/4808841_
https://hwbot.org/submission/4808842_

https://hwbot.org/submission/4808844_

 

And always nice when people keep their mouth closed when their #dear beauty# is taken with their panths down. But bro @Mr. Fox, same as me won't bother with this. Same as he do for Intel. Performance first. Always! But AMD software/firmware combined with bugs doesn't help on the mode forced to use some specific or newer versions of software, drivers and firmware. Working USB ports is a bare minimum you should expect from your pc setup. Can't get that with older software/bios/software from AMD.

Mitigations require software re-coding
http://forum.notebookreview.com/thr...atches-and-more.812424/page-136#post-11116100

Edit.
See also the comment section https://forums.tomshardware.com/thr...ity-for-amd-zen-and-zen-2-processors.3721626/

 

I agree 100% but I do want users of both companies to know that each have their idiosyncrasies. I try to be as neutral as possible because I enjoy products from both companies. I root for both. When they both do well, we get hyper competition....well I don't know about the future.

 

This is usually the first sign of pushing fclk too hard when the whea errors start to show up in hwinfo. I remember when some were arguing that it was hwinfo that was the culprit causing the whea errors (lol) when I was dialing in my timings and trying to jettison TM5 and hwinfo whea errors. TM5 would pass but hwinfo was still spitting out erratic whea errors. In the end, I ended up tightening up my primary and terts at 1:1 3733 as 3866 was just too finicky even on the last BIOS revision with looser timings whether I ran 2x8 or 4x8. It produced the snappiest system and best WoW performance for me with zero whea.

 

Yes, same experience. No TM5 errors, but running TM5 or AIDA64 memory benchmark would produce CPU interlink (I think that is what it was called) WHEA errors visible in HWiNFO64 even though TM5 and AIDA64 run fine. That fits with the Infinity Fabric being pushed beyond stability. Where the problems start for me with the shutting off is at higher than 4.7GHz or severe load, like Cinebench R20 with no dynamic AVX adjustment. Not to mention the " it boots fine now, but wait until tomorrow or the next day and it won't" problem. That is insanely frustrating. It hasn't happened once since I backed off the FCLK, UCLK and memory clock being locked at 2000 across the board. High memory latency isn't wonderful, but it's better than an unstable system. Above 4.7GHz on all cores, running AIDA64 memory benchmark was also causing my system to turn off with the FCLK matched to the memory clock. I am running it at 1900 now instead of 2000 and UCLK at 1000 instead of FCLK and UCLK being tethered to one another. Seems much better overall in everything except memory latency. Unless AMD does something pretty major, their memory performance is just going to suck compared to Intel. And, I am not sure that there is anything they can do because it seems to be an architectural Achilles' Heel/flaw with Ryzen. But, it looks like Intel is copying some of what they do, so they'll probably both end up with high latency issues after all is said and done.

By the way, I took screenshots so I could find the settings again later. Maybe these will be useful to others to minimize their struggles. They are really buried. If my memory clock set to 4000 and I have the BIOS setting for FCLK set to "Auto" (at least my BIOS) is sets FCLK at 1800 and UCLK at 1000. Very stable. If I try to manually set FCLK that makes UCLK the same, which is totally unstable. Bumping the UCLK makes my system totally unstable. The "UCLK DIV1 MODE" option set to UCLK==MEMCLK/2 will keep UCLK at 1000 when my memory is set to 4000. FCLK can be manually set to another value without affecting UCLK.

I knew I had seen this once before and had a hard time finding it. I have also experimented with 5 or 6 different BIOS versions and the knuckleheads at ASUS seem to have a really bad habit of removing things, renaming and/or moving them on newer BIOS versions. That really sucks. For example, in this YouTube video Roman shows a feature (Dynamic OC Switcher) that was removed in newer BIOS versions for some reason. Or, they moved it and buried it another place. I can't find it and it is not on the menu where he shows it. That bugs the crap out of me.

​

You have to go to the Advanced menu, AMD CBS menu, then drill down three more levels to find this under XFR Enhancement. With UCLK being held at 1000 I manually bumped FCLK from 1800 to 1900 with no stability problems. I will see if I can go as high as 2000 on FCLK without losing stability with the UCLK being held in place at 1000.

​

 

Nope. 1900 is as far as it can go without incurring errors. Set to 1966 from 1900, I get this doing nothing but web browsing.

I received it. Nice product. I like it a lot.

 

Don't forget the feedback. And the cat may interest bro @Spartan@HIDevolution Or is it a dog? A small one? Some of they are tougher than a train. But big and muscular is still KING:

 

It is a kitten less than 6 months old.

I have used the keyboard about 30 minutes so feedback will follow a bit later on. It looks and feels great. Not super clacky, but solid tactile feel.

 

This thing is awesome. It's a portable all in one desktop with multiple screens. He even made his own power bank to power it away from a power outlet, and claims it lasts him 7 hours while gaming.

Now this guy needs to make something like a Slabtop.

 

Man that's just cool. Actually really enjoyed watching that. He's got a great channel as well.

Sent from my SM-G970U using Tapatalk

 

Finally achieving good results with the 3090 KP Hydro Copper block.

The boosting of the card is wrong due to an old profile in Evga PX1 being applied, from the standard KP XOC bios (Which uses the standard 3090 boosting frequency) I had to reinstall PX1 to fix the issue.



Here are the temps on 520 KP HC bios. Excuse the yelling children in the back ground.

 

Nice work, excited to see your OC results and temps.

Sent from my SM-G970U using Tapatalk

 

EVGA explains how Amazon's MMO bricked 24 GeForce RTX 3090s pcworld.com | TODAY

A bad batch of early cards and a unique power load from Amazon's New World MMO are to blame for two dozen dead RTX 3090 GPUs.

An analysis of dead EVGA GeForce RTX 3090 cards that failed while playing Amazon’s New World game indicate a rare soldering issue limited to a small batch of cards is responsible, a company spokesman told PCWorld.

As it turns out, it was a ‘rare soldering issue’ which was discovered by EVGA after X-Ray analysis of the affected cards. Only cards produced in 2020 were reportedly affected by this issue.

And how many cards have soldering issues that may show up later? Can't be only the two digit numbered cards (2 Dozen) that died. How many cards is normally produced in one batch? 100-500?

----------------------------------------------------

Btw. Miners will always have a chance to get graphics cards for their #work-flow#. The biggest boys in this disgusting business have near 500.000 cards at their hands.

"It is truly disgraceful behavior by distributors to deliver such a large number to companies specializing in building mining hardware when gamers are struggling to find cards in stock"

http://forum.notebookreview.com/threads/crypto-news.836558/page-2#post-11116326

 

I'm on the i7-11800H GE76 with the Crucial Ballistix 3200Mhz CL16 (64GB). I was able to switch it to Gear1 and I ran Windows memory checker which returned no errors. But any other software like AIDA64 or Memtest86 return thousands of errors. Did you do any stability checks at Gear1? I'm assuming I should go back to Gear2 and check to make sure that resolves issues as I just put them in and the first tests I ran were in Gear 1.

Sent from my SM-G970U using Tapatalk

 

My stock memory could not handle CL16 after further testing.

However I’m using my Vegeance dual rank 3000mhz CL16 stuff now. It’s rated for that speed at just 1.2v which makes it really good for MSI books that can access XMP and custom memory tuning. Otherwise default SPD is JEDEC at much slower 2666 CL18.

I’m not having any issues at 3200Mhz CL16. However I can’t drop my TFRC as low at 3200mhz 16 Gear 1. I think I’m stuck at 500 TFRC now. 400 was crashing games or logging whea.

 

Ok, these sticks run at 1.35v stock. I think I can bump up the VCCIO and VCCSA voltage to try to stabilize, does that sound right? I know nothing about memory overclocking. I'm just wondering if that's why I'm getting errors in Gear 1 because the memory controller isn't getting enough voltage maybe?

Sent from my SM-G970U using Tapatalk

 

Great information. Fortunately for me, my 3090KP has been ok for the most part with even really high power consumption.


Very reliable. The good new is that for anyone who had an Evga card, they can easily and quickly get a replacement.

 

I have 2 hours to order a 3090 ftw3 ultra hybrid. If someone wants one I will give it to you cost plus shipping to you. Just want to help someone out and not pass up the evga queue.

 

Those are pretty neat cards last I saw

 

If you do any overclocking or run the memory above 3600 you will not find stability with 1.350V. You should run at least 1.400V if you want to avoid errors and stability problems. Many XMP profiles are unstable because the default voltage is too low. And, you can probably stay with default for VCC10 and VCCSA until you push beyond 4000. If you start going above 4000 and tighten up timings, you can bump VCCIO, and if necessary VCCSA. I usually increase VCCIO first to see where that gets me.

Speaking of memory, I think this is as good as I am going to get with these G.SKILL Ripjaws V modules. They are excellent and ran at 4600 CL16 on the Z490 Dark, but this platform doesn't have capabilities that are even remotely close to comparing with the 10900K and Z490 Dark. They are also not Ryzen-centric on the SPD. The SPD on these modules is optimized for Intel, not AMD. It seems Intel 11th Gen is also a downgrade in memory overclocking for Intel.

Even though the results are decent (especially latency) I am going to see if a Crucial Ballistix MAX 4400 2*16GB CL19 kit (should arrive tomorrow) is better suited to 5950X than the Ripjaws with Samsung B-die. If not, I will return them to Amazon for a refund. I hear rumors that Ballistix MAX is the best memory for Ryzen 9, so I guess I will see if it is true or just a load of crap from Ballistix fanboys. On Intel, Samsung B-die is the best.

Anyhoo... I accomplished this. Totally stable with a multitude of testing with memory clock, FCLK and UCLK all set to 1900. The improvement is in stability and latency. Benchmark scores seem unaffected (not enhanced). I expected better results from 3800 CL14, but achieving total stability with those timings at 3800 on an AMD platform is an achievement even if the results in benchmarks did not show any meaningful gains. (I did not test gaming results because that isn't particularly important or interesting to me. I benched CPU-Z, wPrime32M and Cinebench.)

I am not sure what the deal is with W11's horrid L3 cache performance. I will retest to see if that is consistently poor or if it was a random anomaly. W7 and W10 LTSC are basically the same.

 

Well you're not the only one experiencing strange L3 results on Win 11, that's for sure.

https://windowsreport.com/windows-11-benchmark-tests/

Sent from my SM-G970U using Tapatalk

 

Thank you for the link I'll have to check that out. I hadn't noticed that other people were experiencing similar weirdness. I don't think the Redmond Retards could do anything right if their lives depended on it. I can't say that I'm surprised because Windows 11 is a pretty botched up mess overall, not to mention stinking ugly.

 

When you overclock the memory to gear 1 from gear 2 you will notice the System Agent VID goes up automatically, but I do think there are ways to force higher VCCIO Mem OC and VCCSA. I'll poke around in the BIOS and see if I can't find it for you. If you're getting errors I would start with loosening up the TRFC first. That is one of the fastest ways to destabilize the overclock.

Are you also on the latest BIOS from MSI for that laptop? I would update to the latest BIOS if not already on it. Latest patch notes mention improved system stability.

I've gone back and began running memory stress tests to check for the issues you describe but so far I am not getting a single error or instability detected at 3200Mhz CL16 18 38 with 500 TFRC and Gear 1. Latency is around 63ns in AIDA64. All of my games and general OS has been rock stable. It's entirely possible it just doesn't like that memory kit you're trying to use.

Trying to get the latency down on my 11900K setup. My 11900K has a terrible IMC and can't run 3866 at reasonable latency. Looking forward to Alder Lake, if for nothing other than to mess around with a different platform. Seems like 1.55vdimm is going to be my 24/7 daily. Good thing these bad boys are rated for 1.55v out of box.

https://imgur.com/a/EuhVJGC

 

If this is about correct, the 8 small added phone cores and the "K" moniker really jack up the prices for 12th gen Intel chips. Probably to match AMD's 16 cores 5950X (about same performance hence around same prices).

But the new power efficient small cores will likely reduce your power bill year over when they do small tasks on your pc in idle/small loads as etc update your OS in the background. Yep, worth its weight in gold to reduce your investment costs for coming Alder lake platform

Possible Intel 12th Gen Core Pricing Leaked, i9-12900K Costs 741€ techpowerup.com - Today, 10:00

The top Intel 12th Gen Core "Alder Lake-S" 8+8 (P+E) cores processor will cost 741€ ($878.50 United States Dollars) including taxes, according to a leaked document revealing retail channel prices of various upcoming 12th Gen Core desktop processors. It also speaks of the i9-12900KF, the "almost-flagship" part that comes with a disabled iGPU, going for up to 708€ (incl taxes).

The Core i7-12700K, the 8+4 (P+E) cores part that lacks Thermal Velocity Boost, will go for up to 524€ (incl taxes). Its "KF" twin will be about 20€ cheaper. The mid-tier Core i5-12600K processor (6 P-cores and 4 E-cores), is going for up to 365€. Not long ago, this was the roughly the price of Intel's top mainstream-desktop processors (such as the i7-7700K). The iGPU-devoid i5-12600KF will go for 333€. Intel is expected to debut its 12th Gen Core desktop processors and compatible Socket LGA1700 motherboards in Q4-2021, along the sidelines of the Windows 11 launch. The first wave of processors are expected to only be unlocked K or KF parts, with locked ones only arriving in early Q1-2022

 

Yes I'm on latest BIOS. This is the 11800H in a laptop so I'm not sure if that factors in as well since 1.2v is default and these sticks are running at 1.35v. I believe the TRFC on the sticks is set at 560 so I could try changing that to 500 as you suggested. I found VCCIO and VCCSA voltage override in BIOS but I'm a little uncomfortable as I don't know how much to overclock those numbers. Thanks for your help.

Sent from my SM-G970U using Tapatalk

 

I never knew that 5950x was hot. I only had luck with Zen2 BGAs which ran much cooler than Intel BGA at stock paste. deltaT between Intel BGA & AMD BGA was 10-15C w/o undervolt on 35C ambient temps. With Intel you always have to tweak undervolt values after uCode update, with TS the occurrence of black screen due to insufficient voltage is minimized since it will revert to default state w/o user intervention with a pop up notifying the user.
I never got a chance to test rocket lake BGA. I was furious with 10th gen i3 on dell that overheated out of the box after my friend asked me why is the laptop shutting off every 2 mins when its not even 2 days old.

 

Let me clarify that so it does not seem I am slamming AMD on this being a problem. The problem is me and how I use my systems. It's not crazy hot running stock. Pretty normal in that scenario. Hot, yes... very hot. But, it is 16 cores being pushed to the edge of functionality by a nut like me. I don't have any interest in running stock clocks. If I can't overclock the snot out of it, then I don't want it. It runs very hot when heavily overclocked, but not abnormally hot or remarkably different than my experience with Intel CPUs that have not been delidded. The unfortunate part is there seem to be a few more architectural and firmware limitations to overclocking on AMD and running delidded and bare die is not possible. All of my overclocking fun on Intel for the past few years is predicated on a delid and bare die setup.

 

I'm glad I went with a Comet Lake based system and didn't wait for Alder Lake. These prices are ridiculously high. And here I thought Comet Lake was priced absurdly.

You can be sure of one thing, I am going to buy even less stuff in the future and will make all my stuff last even longer.

My parents instilled good spending habits in me.

 

And by 2022 Q4 Raptor Lake will be coming to compete against Zen 4 AM5 platform. It seems really like a botched test platform with short life, I hate Intel more on how they ruined Rocket Lake IMC and Clock scaling and Cores. Really frustrated to pick a damn CPU because of the shenanigans of the IMC lottery (even bare min 3733 Gear 1 is hard now for 11th gen on a mass scale, what a joke), Heat of 11900K with less cores and PCIe4.0 vs the better 10900K PCIe 3.0, if we use an NVMe on either of these RKL will have PCIe 4.0x16 class speed on GPU lanes and CML will have 3.0x8 on the GPU link, another NVMe on RKL will make it 4.0x8 = 3.0x16. Basically RKL gets one NVMe slot without effecting any PCIe lanes from mobo. For high FPS through a fast card like 3080 or 6800XT at 1080P the lane speed does matter.

With Alder Lake IMC gear comedy will continue, along with other whole can of worms. New pricing, I think it's just that DDR5/Gen5.0 on GPU lanes add up the cost, as for the performance from those leaked benches, it barely beats a Ryzen CPU in SMT and matches in ST so they are pricing like AMD Ryzen 9 series, but the way how Cinebench calculates or any bench vs Real life gaming workload performance and Productive tests should decide the fate of this Alder Lake platform as a whole. Are you going to pay for the PCIe5.0 and DDR5 cost is all I see from this launch tbh. Can't wait to see this launch, 2 months more and by Oct end the embargo will possible lift I guess.

 

Yep, the older is still the best +reps

I think a lot become destroyed because Intel was forced to push Rocket lake on older 14nm and saved all 10nm silicon for the Jokebooks so they could be able to compete with Ryzen mobile. Scared to death.... Apple jumped the ship. Same also it seems for Google with their Chrome-jokes. And Microsoft prepare for a good bye as well with own chips for their Surface-books. Intel have to come up with something new/better or all the DIY users will also jump the ship as well. Maybe probably saved by the chips shortage? Yep. The OEMs need every chips they can find. And I'm not a AMD fanboy.

 

I haven't owned a personal desktop ever (not by choice), but when you say DIY users will jump ship, where do they jump ship too? Do you guys just build all your stuff yourselves and piece together what you want? I feel like, at the end of the day, there is only so much you can do to "jump ship" because you are still stuck with a few companies who make processors. You can get lots of different motherboards, accessories, etc. but what choice do you have if you are trying to get away from a certain processor?

 

Intel have lost many of their loyal customers. With no chips shortage it would be much worse. Remember AMD could sell more chips if they had enough access to (silicon) chips from TSMC. Why should people pay more or the same for 25% less cores if they have an another options? Or less perfomance vs $$$? If the competitor offer more for less some have to lose. Apple is gone, soon Google. Maybe Microsoft as well. The big boy in chips industry as Intel can't make many more fails before it hit them hard. But they still float well on own chips factories (they sell everything they can produce and have good capacity).

Edit. If I want more points on the bot or want to reach the top of 3dm leadebord for etc 3d benchmarks I wouldn't jump on Intel. At least not at the moment. Remember points on the bot is not eternal. Tomorrow you can be kicked to 2nd or better say last You have to live in the moment, not the past or future. My middle son have told me that more than one time Sometimes the younger can see it more clearly than you. But not always. But always learn from it. None is perfect, at least not before they say hello to to big lord in the great and big heaven.

 

Sounds like you have wise children. I have two daughters and it's amazing how often they remind of the perspectives I had when I was younger. I agree, there is no guarantee for tomorrow and the past is gone. Best to find peace where you are at the moment. As far as the tech industry, I see globalization and monopolization as the biggest threat to choice, not to mention price fixing. For that matter, those two threats are dangerous for all industries. Say what you will about capitalism, I'll take healthy competition and freedom of choice anyday over the socialist alternative of a One World Order.

Sent from my SM-G970U using Tapatalk

 

The latency isn't great, but it probably never will be with Ryzen. This RAM is definitely more stable on Ryzen than Samsung B-die. The modules are super heavy and the heat sinks are pretty serious stuff. I think I am going to keep them and move the amazing G.SKILL sticks that don't work so great on Ryzen to my other desktop.

I still get CPU Bus interconnect WHEA errors with IF at 2000 and that is also an architectural problem for Ryzen that may not get fixed. But, improvement is always a good thing, and this certainly is an improvement. I will experiment with higher than 1.500V to see if I can get it to boot with tRFC set lower, but it doesn't want to boot with lower tRFC than 631 with 1.500V.

 

Does anyone have RealTek onboard ALC4080 experience ? I was thinking on the AMD side equation for the Crosshair VIII Extreme, the board has that ALC4082 audio IC, it's a USB audio instead of the ALC1220 usual fare of the HDA type found on all the AMD and Intel boards. And upon checking I couldn't find any information on the drivers part of these new USB Audio devices from Realtek. Worse is ALC4080 has a ton of issues with EMI problems, ASUS posted a firmware fix for the 4080 on it's mobos, but folks still post they have issues. And ALC1220 has Windows 7 HDA RealTek driver but not the 4080 (?). The 4080 is also used on ASUS Maximus XIII Apex. Z490 / Z590 DARK both use ALC1220 platform.

Also upon checking the Windows 7 drivers for this ALC4082 I found ASUS forums, @Falkentyne mentions problems associated with the drivers, the worst one can imagine with the 408x. The boards with these have top notch ESS DACs and Nichicon Capacitors with Impedance sensing circuitry but guess what ? without the ASUS Sonic bloatware (UWP / Appx) the drivers will have lag during .Wav. and there's no fix apparently except installing the OEM bloatware pack. Damn so much of these expensive parts don't they actually do a QC on these, and none of these have .inf drivers for simplicity ?

Crosshair VIII Extreme also has Marvell AQtion AQC133CS 10Gbe LAN which is the latest PCIe based Ethernet controller and apparently unlike AQC107 10Gbe and other series based on 3.0 this is 4.0 and doesn't have a Windows 7 compat. On top that board has Thunderbolt 4 ports too, I don't think that will either work on the Windows 7. Only reason I'm checking this board is because it's one of the ASUS boards which is either equal or above the Dark Hero, and I'm guessing it's X570S revision of the I/O chipset on AMD platforms, since all the new 2021 AMD boards on X570 from MSI/GB/ASUS got refreshed with fanless design citing X570S, no reports of the users on USB issues regarding this refresh so far as they are brand spanking new. And also no information on the B2 stepping revision of the AMD Ryzen 5000 Vermeer (do a Ctrl+F and type USB, folks also ponder the same), hoping it also fixes the USB issues at HW level ? In 2017 AMD did a B2 refresh of their Summit Ridge / Zen silicon addressing some issues which couldn't be fixed by AGESA, this time they do not mention anything, my guess is silent refresh addressing the HW issues. Which is on 2 fronts, B2 on CPU and X570S on the Chipset.

On the same HW parts quality and drivers I found that, idk if It was mentioned here already, Intel 225-V (runs fine on Win7) which is the most famous 2.5G LAN controller on all boards across the PC DIY has Hardware issues all over the place, it has been in supply since 2019 and all boards from Intel to AMD budget and premium use it. Fix is cap to 1G LAN speed or HW revision 3, which is found on Maximus XIII Hero so guessing all the new boards will have the latest HW chip.

 

I have everything working with Windows 7 on the Crosshair VIII Hero (non-WiFi), including the USB Type-C ports. I don't know that Thunderbolt functionality is enabled on them, but they are blazing fast with my USB SATA and NVMe drives. I don't care if it has Thunderbolt functionality because I don't use that for anything. I do not need WiFi, and I can put in a USB WiFi module if some weird reason warrants it. I do not believe there are working WiFi 6 drivers for Windows 7. I could not get the AX WiFi on the Z490 Dark working on Windows 7 because I could not find (or mod my own) working drivers. I disabled it and the second NIC in the BIOS.

I prefer the Realtek PCIe 2.5GbE over the Intel I225-V network adapter on the Crosshair VIII so I disabled the Intel NIC in the BIOS.

I have had no issues so far with the Realtek audio on the Crosshair VIII.

 

It looks like this BIOS might actually be the best one. Using the same custom memory timings that I posted earlier, is looks like read speed it faster and latency improves by about 3-4ns. If that turns out to be accurate I will post more info.

https://community.hwbot.org/topic/1...e-ln2-oc-guide/?do=findComment&comment=596497

 

So I switched this Ballistix 3200 back to the default Gear 2, assuming that I was getting Memtest errors because I had changed it to Gear 1 right after I put it in without testing. Unfortunately, this didn't change a thing. I tried upping the voltage to 1.4 but still the same result. So I set it to the default JEDEC profile with 2666Mhz and 1.2v, Gear 2, re-ran Memtest and no errors at all. So I just have a few questions. I don't know what offset to put for my VCCIO and VCCSA as I have no idea how to tell what they are running at currently. Also, does anyone know what the "Dynamic Memory Timing" option is, its currently disabled? Since I have yet to experience any actual issues while using the computer at the 3200Mhz 1.35v, no BSOD's or random shutdowns or anything at all, is it really that bad to use the memory while I know it "could" run into errors? Assuming I'm doing nothing mission critical and all my files are backed up, is it just bad practice to run memory when you know it's unstable or is there some catastrophic potential consequences like killing the RAM itself?

 

I had a feeling it has nothing to do with you selecting gear 1. Are you undervolting your CPU/Cache at all? Have you tried reducing this? Did you log the errors with the stock ram while in gear 1?

Dynamic Memory Tuning should be real time memory timings, meaning you can change the memory timings while in Windows with something like Intel XTU, but I haven't tried this myself. I personally would want to ensure I wasn't logging any errors even for gaming use. Random crashes or reboots will likely occur otherwise and eventually corrupt OS.