• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

X299 and 7980XE in 2024

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
(I did edit this into the previous post but saw Earthdog liked it in the time it took me to write, so decided I better separate it out in case it gets missed.)

It seems to be the ram, not the mobo.

Slot 1 -4, ram sticks A-D

1A, 2B, 3C, 4D - 48GB detected
1A, 2B - 16GB detected - either the ram or the slots are the problem. Ram C+D assumed good
1C, 2D - 32GB detected - with ram assumed good, this shows these slots are also good. Ram A or B are suspect.
1C, 2D, 3A, 4B - 48GB detected - A or B still suspect.
1C, 2D, 3A - 48GB detected - stick B suspected bad
1C, 2D, 4A - 48GB detected - confirms all slots seem ok
Clean the contacts on module B
1C, 2D, 3B, 4A - 48GB detected - returning B did not add capacity. I'll pull it and write it off for now. Might look at it again another time.
 
Sorry, lol... I get notifications from here and it's always open on the second monitor, lol. Perhaps a bit of ADHD (oh, look at that shiney thing!!) sprinkled on top and you get instant likes and replies! :chair: :clap:

Anyhoo...what a PITA. Glad you have it identified tho.
 
You could try the bad stick in another MB, just for confirmation. I know g.skill has a great return policy on bad memory, are you going to file a RMA? How old is your set?
 
You could try the bad stick in another MB, just for confirmation. I know g.skill has a great return policy on bad memory, are you going to file a RMA? How old is your set?
It's on the to do list. I know I've two systems in this thread and it is easy to mix them up. The ram in this case is Corsair. I just checked their web site and it seems to have limited lifetime warranty so it doesn't matter when I got it. It was July 2019.
 
I had a Weird Issue running 8x sticks on my X299 with an I9-7900X
Adding them 1 at a time and they all worked eventually and if I left it like that they wouldn't fail unless I reset bios or updated bios.
But resetting bios/Updating and booting with all 8 in I would have some sticks fail for some reason.
I always Chalked it up to module issues, but it eventually worked, and I could not make it fail unless I tried to OC past 4.8ghz on all cores.
I Still have this system running in my Guest Room as a spare PC - I've been debating turning it into a Server to replace my old Supermicro Ubuntu / Etc Server.
Mine's under water and it's been de-lidded It runs easily @ 4.6 all core with low heat - 4.8 is a bit more but it will do it stable.
Reason id run it as a server is its super stable - 10c/20T , 64gb mem - modern. everything basically - Most Consumer Priced Raid Controllers are also still PCI-E 3.0 anyways even normal NVME ones.
Its a Aging System - My AMD - 7600X @ 5555Mhz - DDR5 @ 6000 - 30CL - Etc was a really big upgrade for Gaming and Heat output into my Gaming Room.
 
I don’t have any corsair ddr4 as i stopped using them from a BAD ddr3 RMA. I was told by Corsair RMA Service Agent during my request that A : Any speed above it’s rated speed and B : Any Voltage about 1.65 -> Will void it warranty for which I was denied because of OCing it :-(
 
I tried the suspect stick of ram from before in a non-X299 system. It worked enough to let a system boot to Windows. Didn't check beyond that since I want to use the system for other things soon.
 
Finally did more testing on the suspect stick of 16GB ram in another system (B560). It runs fine. I tried both XMP on (45 minutes) and XMP off (35 minutes) and aida64 memory test was fine on that. I'm tempted to try it again in the X299 system and see if it is detected now.

I also tried some testing on the 8GB modules I was running before that. This is more interesting. Even XMP off, it would error quickly if I ran them 2DPC, but they're fine 1DPC. I tried various stick combinations and it doesn't seem to follow any particular stick, so the "problem" is running 2DPC. I have 4 sticks running in the 2nd X299 system. I've not got any ram at all to run 2DPC on that.

I notice the Chinese X99 boards are really cheap now. Debating getting one for a 14 core Xeon I still have lying around unused.
 
My 10980XE on an Asus X299 Prime Edition30 runs flawlessly with 64GB of DDR4 Vengeance Pro filling all 8 dimm slots. But, my 64GB kit (8 sticks) had to be placed in the corresponding dimm slots in sequence by serial number. Otherwise on boot I’d get a memory error code. Just my experience. In my EVGA X299 FTW k motherboard I am useing 8 sticks of 8GB Vengeance with no worries and no idiosyncrasies. Both machines running XMP and overclocked significantly. Hopefully you can see the scores if you zoom in. IMG_1673.jpeg
 
Last edited:
I forgot to update?

I still couldn't get the 4th stick of 16GB to work in my main system. I tried different slots and fewer modules in case there was some weird capacity thing going on. Just wouldn't use it. Again, the weird thing is the SPD is detected so at least that is visible.

I also tried 4x 8GB instead and that wouldn't even run stably at 2133 with ram that has been tested in the other system. Maybe something in CPU/ram/mobo has aged and just drifted a little out of tolerance causing edge cases to pop up like this. At this point I give up trying to fix it. It works with 3 channels and that is sufficient to keep going. I kinda decided I will replace my main system with a next gen system from either side this year.
 
I noticed that UEFI can be a little buggy sometimes, even when everything was going great. Happens when you switch CPU's mostly. Ghost settings get retained somehow. Maybe as a last ditch effort give the board a fresh flash to show you care?
 
I noticed that UEFI can be a little buggy sometimes, even when everything was going great. Happens when you switch CPU's mostly. Ghost settings get retained somehow. Maybe as a last ditch effort give the board a fresh flash to show you care?
I got nothing to lose, other than time I guess. I'll give it a go later.

I did clear/reset the settings in bios when changing CPU so that should have been enough, but doesn't hurt to retry just to make sure.
 
I forgot to update?

I still couldn't get the 4th stick of 16GB to work in my main system. I tried different slots and fewer modules in case there was some weird capacity thing going on. Just wouldn't use it. Again, the weird thing is the SPD is detected so at least that is visible.

I also tried 4x 8GB instead and that wouldn't even run stably at 2133 with ram that has been tested in the other system. Maybe something in CPU/ram/mobo has aged and just drifted a little out of tolerance causing edge cases to pop up like this. At this point I give up trying to fix it. It works with 3 channels and that is sufficient to keep going. I kinda decided I will replace my main system with a next gen system from either side this year.
It's a popular issue with all 4-channel desktop chipsets. The same was in most X79/99/299 mobos ... matter of luck I would say, but switching RAM or CPUs more often doesn't help for sure. For some reason, I haven't heard about problems with some single memory slots on AMD Threadripper. On Intel, it can be a memory slot, CPU socket, RAM incompatibility, or some sticks that work better in some specific slots (like TechStuff mentioned in post #49). I tested/reviewed most X99/X299 mobos. Every new motherboard was working fine with 4-8 sticks. Every motherboard that I was using for longer for benching and other tests had problems with 1 or 2 memory slots. Typically, the first slot that has problems is 2nd on the right side of the CPU socket. The only exception that has worked great since I got it is ASRock X299E-ITX, but it is probably because it uses SODIMM.
At work, I have 2 PCs with MSI X299 mobos. Both have run 24/7 for a couple of years now. Both have 8 memory sticks that were never removed after the first start.
 
Every new motherboard was working fine with 4-8 sticks. Every motherboard that I was using for longer for benching and other tests had problems with 1 or 2 memory slots.
Am I reading correctly, you have had some systems where it would work fine initially but after some use start to have problems? But then you have others that have been fine.
 
Am I reading correctly, you have had some systems where it would work fine initially but after some use start to have problems? But then you have others that have been fine.

Yes. It was almost always like initially there were no issues at all, but after multiple tests, CPU/RAM reseating, and some more things, problems were starting to appear. I had a couple of RMAs that ended with CPU socket replacement or the whole motherboard replacement. Most motherboards were review samples, and after some tests, I was selling them, so they were still fine. Most of those that I kept for longer and had a warranty, eventually went to RMA, and I was selling them after a successful repair or replacement.
I still have EVGA X299 Dark, which works fine, even though, for some unknown reason, 2 or 3 pins are slightly bent (or maybe reflect light a bit differently, but are on the correct height). When I was using it the last time, I had to reinstall RAM or switch sticks in slots until it recognized all of them properly. Once the BIOS sees all sticks right, then it works fine.

There were many stories with one memory module missing or the BIOS showing all memory sticks, but Windows wasn't addressing one or two of them. This was the same on every X series chipset from Intel that I remember.
As I mentioned, for some reason, there are no problems with AMD sockets. There were no repeatable problems with Threadrippers, and there are no problems with the latest generation. At the same time, Intel has new CPU socket issues every generation. Right now, you may lose stability or damage the socket when the cooler pressure is too high. It's not hard to do that, as when you tighten the mounting screws on some popular AIO, then, you may end up with damaged CPU pins. I see that mainly on MSI motherboards, but maybe it's my luck. Still, MSI puts stickers on the CPU socket to not overtighten CPU cooler screws as it may damage the socket. It's the only brand that does that since 400 series chipsets. I was fighting with ASUS support over that and recently had a case with MSI support, too.
It's slightly out of the topic, but here is one weird photo of how the socket may look after properly installing an AIO cooler (only one time, no reseating, no extended pressure, or anything else):

1702148891075.jpg
 
I tried the bios re-update and did the ME again at the same time. In short, no change. I lost wired network connection after I did that test. Restart PC, didn't help. Restart router, didn't help. Restart switch, I'm back. That's not something I've seen before. If it ain't broke, fix it!

As I mentioned, for some reason, there are no problems with AMD sockets. There were no repeatable problems with Threadrippers, and there are no problems with the latest generation. At the same time, Intel has new CPU socket issues every generation. Right now, you may lose stability or damage the socket when the cooler pressure is too high.
Wasn't there something about board flex on Intel? Do the stability brackets eliminate that and related problems?

I'm still stuck between a rock and a hard place in what I want for my next main system. I don't expect either side to give me everything I want at a price I'm willing to pay. Zen 5 might have significant uplift in AVX-512 perf if you can tolerate separate CCDs and the long wait for X3D models, and Arrow Lake should be a big step forwards in general too.
 
Wasn't there something about board flex on Intel? Do the stability brackets eliminate that and related problems?

In theory, it helps, but I find a design flaw in Intel's sockets. Each time I see an issue with more pins, they're in a straight line or on some specific spot (like 4-6 next to each other). Sometimes it's 1-2 in the corner. Some series have pretty thick backplates, so it's hard to bend socket/PCB, and these issues still happen. MSI suggests it's from the cooler's mounting pressure, so as long as the socket isn't for some reason too fragile, then 3rd party brackets, like those from Thermalright or Thermal Grizzly, should help. I can't tell how it is in reality as I wasn't testing it, and these products were not available in the past.

There were 2 Foxconn socket issues in 2 different generations. Lotes sockets were fine, while in Foxconn, some pins could burn under higher currents. There were also issues with power design, but it was related to some motherboard series and manufacturers. No matter how you look at it, there are way too many socket issues in Intel motherboards, and we can't hear about repeatable issues in AMD motherboards.

For me, Intel is good for tests, and I sell most of them once the new generation is out or I no longer need them. In my daily/gaming PCs, I only have AMDs. Maybe the next Intel generation will be better, but the last couple of them were pretty bad in terms of efficiency/TDP/heat.

I'm still stuck between a rock and a hard place in what I want for my next main system. I don't expect either side to give me everything I want at a price I'm willing to pay. Zen 5 might have significant uplift in AVX-512 perf if you can tolerate separate CCDs and the long wait for X3D models, and Arrow Lake should be a big step forwards in general too.

I wasn't really convinced, but I like weird stuff, so I bought 2 Minisforums motherboards when they were on sale + additional promo codes. I say that as these motherboards have soldered Ryzen 7/9 CPUs that support AVX-512. They also support 2x M.2 PCIe 5.0, PCIe x16 slot (with bifurcation), 2.5GbE LAN, Intel WiFi 6E, and a cooler/heatsink is included. The only problem is that ordering it from the Minisforum store takes 1-2 months. On Amazon, there is only one version with an 8-core CPU - https://amzn.eu/d/gNZFc61. It costs £395.99 for the CPU+mobo+CPU heatsink, so it's still not bad. It also requires SODIMM DDR5, but a regular 5200/5600 kit is enough.
Intel mobo has more options and offers RAM OC, but the CPU doesn't support AVX-512.

I'm not going to convince you it's a great motherboard, as it has its limits, but the price for what you get is pretty good. I run mine for 2 months 24/7 without issues. The CPU is low power (TDP can be adjusted up to 200+, but the CPU won't take more than ~120W). I also noticed that the mobile Ryzen 7000 is available up to 16 cores, while the 8000 is only up to 8 cores. The IPC is almost the same for both of them.
One advantage that I see in these mobile chips is that the CPU is low-powered but also keeps max boost on all cores, not only 2-4 like in desktop models. Once you unlock TDP and keep the CPU cool enough, then you will see all cores running at 5.1-5.4GHz (depending on the CPU). The Tj is 100°C and can be set higher, up to 115°C, so it's easier to keep it without throttling if you want a lower fan speed (I guess it's just because these CPUs are mobile series, so they have a higher range).
From bad things, the support is below average. They reply quickly and try to help, but I wouldn't count on BIOS updates or long support as the BIOS team is very small and jumps onto other products every couple of months. The warranty is for 2 years but requires shipping to Hong Kong (as I saw in the product comments).
 
Last edited:
For me, Intel is good for tests, and I sell most of them once the new generation is out or I no longer need them. In my daily/gaming PCs, I only have AMDs. Maybe the next Intel generation will be better, but the last couple of them were pretty bad in terms of efficiency/TDP/heat.
That's in part why I'm still on X299 and Rocket Lake. Nothing from either side has ticked all my boxes, and I keep waiting. I took Rocket Lake over Zen 3, and no regrets there. AVX-512 support got me to bite, and 5800X3D a year later was too little too late. If I had to buy today I'd go 7800X3D for a pure gaming rig, but it doesn't feel enough for me as a main system CPU. I don't want multiple CCDs, and Intel power efficiency is as lacking as AVX-512.

Arrow Lake doesn't address the AVX-512 problem, but for the first time in a very long time Intel will have a process advantage over AMD, in theory at least. It'll either be an epic CPU, or an epic fail. Zen 5 I'm sure will be a great iteration on Zen 4 and the main thing holding my interest is the rumour they have doubled FP execution units in it. If true, we're looking at Skylake-X beating IPC for FP heavy workloads. Zen 4 doesn't attain that, although it does get ahead of Rocket Lake.

I wasn't really convinced, but I like weird stuff, so I bought 2 Minisforums motherboards when they were on sale + additional promo codes. I say that as these motherboards have soldered Ryzen 7/9 CPUs that support AVX-512.
In my past life when I had more disposable income I might have got such just to play with. Now I have to be more careful with my money and I'm not sure that is it. I did try pricing up a 8600G and 14500 desktop build, but it doesn't make sense to me to get a play thing right now.

Edit:
Data I'd like to get some day:
Alder Lake AVX-512 implementation - maybe some day I'll get a used set that still works.
Raptor Lake P + E core testing
Zen 4 C + c core testing
 
Prime number finding software, similar to Prime95. In effective IPC terms vs AVX2, Skylake-X implementation is about 80% faster, Rocket Lake is about 40% faster. Zen 4 is somewhere in between based on other's data since I can't do my own controlled testing on it. If Zen 5 rumours are correct it could smash all before it.
 
Back