|
Tabemono posted:Is this still the prevailing feeling? I'm doing a 7950X3D + X670E build and trying to parse all the AM5 issues, BIOS updates etc. between motherboard manufacturers to figure out who's the good one this gen is making my brain melt. I don't have personal experience, but going by reddit/techtuber sentiment, ASUS (specifically the high end X670E boards) seems to be bad news at the moment for AM5. Lots of reports of dead boards, strange issues, and bad ram compatibility. Gigabyte seems to have the best memory support, temps, boot times, and performance this generation. Which considering where they were a few years ago is a very strange thing to type! According to hardware unboxed though, all of the X670 boards have competent VRMs, so thats less of a concern for X670 specifically. Since you mentioned needing PCIe slots, if you plan on using any slot beyond the first GPU slot, be aware that there are a LOT of asterisks and oddities to be aware of with the different boards. The Asus E-E Strix and ProArt, the MSI Carbon, the Asrock Taichi, and a few others are bifurcating the PCIe 5.0 lanes to give you that second 5.0 PCIe slot. So putting anything in that second 16x slot will force both slots 8x mode. Whether thats good or bad depends on what you wanna do! On boards that aren't bifurcating, most of them are using far less lanes to the PCIe slots than the physical design might lead you to believe. For example the Asrock PG Lightning has 3 PCIe 16x slots on the board. Sounds great! But due to how the daisy chain works, they are electrically a 16x 5.0, a 4x 4.0, and 1x 4.0. If you are actually going to use the slots (note, most people do not ever use the slots), you will want to do your research. Thankfully its becoming more common for block diagrams to be in the manuals, and shout out to Asrock for doing it on every board these days. And no matter what you do, DDR5 platforms in general are pretty finicky at the moment. I don't think EXPO is a guarantee depending on your speeds, settings, and CPU. Some users are reporting that anything over 6000 isn't possible at all with good stability, regardless of what they do.
|
#
?
Mar 20, 2023 21:23
|
|
|
|
| # ? May 18, 2024 05:42 |
|
|
I ended up buying/Amazon renting a 7600 just so I could get my system up and running to test the other components before their return windows close. Board is the Asus b650-creator, which has the least-bad pcie slot layout, with 128GB of micron ECC. 4800mhz/cl40 (stock) is achievable with 64GB but all four sticks defaulted to 3600 and wouldn't work at 4800. I haven't put in the effort to tighten up the timings or push the clocks too hard until I get my 7950x3d since it'll have a different memory controller, but just setting it to 4200 works and is good enough for now. Training after bios changes is ~2 minutes, but without changes it boots extremely quickly. I sure do miss proper HEDT platforms but Intel's stuff would add another $1500 to my costs and I don't trust Threadripper after how AMD screwed over TR4 owners.
|
|
#
?
Mar 20, 2023 21:33
|
|
|
Are all the memory compatibility issues down to the chipset or IMC? Recalling that Zen 1 practically required Samsung B-Die for a long time no matter what the chipset.
|
|
#
?
Mar 21, 2023 07:32
|
|
|
current CPUs' DDR5 memory controllers are pretty bad right now, only slightly better on Intel because they got a second go at DDR5
|
|
#
?
Mar 21, 2023 07:44
|
|
|
Seamonster posted:Are all the memory compatibility issues down to the chipset or IMC? Recalling that Zen 1 practically required Samsung B-Die for a long time no matter what the chipset. Chipset has practically nothing to do with memory anymore, but the bios and topology does. The Zen 4 IMC seems fairly stable and reliable—it's not nearly as fast as Raptor Lake's but it's more predictable. Motherboards on the other hand can be a hot mess. We've seen a few motherboard vendors have to release bios updates to improve memory compatibility, with Asus maybe being the worst off. I think gigabyte has been fairly reliable though. My understanding is that if you have a good gigabyte motherboard on the latest bios, basically any Zen 4 chip should be compatible with the vast majority of DDR5-6000 kits. This is in stark contrast to Gigabyte's Z690 launch, which was a disaster for them in terms of memory support. That too got fixed in time, though.
|
|
#
?
Mar 21, 2023 08:51
|
|
|
Asus hasn't yet succeeded in fixing their poo poo, then?
|
|
#
?
Mar 21, 2023 11:31
|
|
|
I’m helping a friend plan out a new build today. I typically exclusively use ASRock boards, how are they faring in AM5 land? Was sufficient cocaine distributed to their engineers for this go round?
|
|
#
?
Mar 21, 2023 12:30
|
|
|
they were off to a funny start when they put stickers over the dimm slots that wouldn't tear off cleanly and would leave bits of paper in the slots. but i think they stopped doing that and their boards are otherwise fine. they actually have some of the best value boards, with the cheapest e-variant boards available (boards that have pcie 5.0 x16 slots)
|
|
#
?
Mar 21, 2023 12:47
|
|
|
Has anyone run into issues updating the BIOS on their AM5 mobo? I have an Asus B650E-F and I can't get it to update. I've tried using USB flashback as well as the in-BIOS updater. Some people report an issue using USB 3.0 flash drives but I've been using an old 32GB USB 2.0. I even tried making a 1GB partition so that I could format it with 4KB sectors and that didn't work. I know that in general, it's advised to not update the BIOS unless you have a problem, but as this is a new platform, I expect I'll be updating the BIOS for a year or so. I thought I might have a BIOS-related problems (analog speakers making a small pop sound when my computer wakes from sleep), but a reformat and NOT installing the audio drivers from Asus seems to have fixed that. edit: I don't know if it was putting the flash drive in a different port or loading optimized defaults beforehand but flashing worked this time with the in-BIOS utility. 
Josh Lyman fucked around with this message at 15:10 on Mar 22, 2023 |
|
#
?
Mar 22, 2023 09:30
|
|
|
Josh Lyman posted:Has anyone run into issues updating the BIOS on their AM5 mobo? I have an Asus B650E-F and I can't get it to update. I've tried using USB flashback as well as the in-BIOS updater. Some people report an issue using USB 3.0 flash drives but I've been using an old 32GB USB 2.0. I even tried making a 1GB partition so that I could format it with 4KB sectors and that didn't work.
|
|
#
?
Mar 22, 2023 11:51
|
|
|
Hockenheim posted:I’m helping a friend plan out a new build today. I typically exclusively use ASRock boards, how are they faring in AM5 land? Was sufficient cocaine distributed to their engineers for this go round? I'm just waiting on a 7800x3d to finish my upcoming build but after doing a lot of research and reading a lot of reviews, especially if you wait for a sale they seem to have the best bang/buck going. The PG Lightning has gone on sale basically like clockwork at the beginning of every month since december for $260 after MIR on newegg canada. I've been eternally asrock-hesitant because I worked in PC repair in the mid-00s but the sheer volume of "yeah these are the best boards for most people" reviews got me convinced and my x670e PG Lightning actually just arrived today. e: this is of course all theory but like I said, the internet has broadly been saying that they're categorically Good
|
|
#
?
Mar 22, 2023 21:16
|
|
|
For Zen 3 at least, only the "G" CPUs contain the AMF encoder, right? You need the iGPU to get that, and it's not somehow part of the CPU part of the die?
|
|
#
?
Mar 22, 2023 21:26
|
|
|
Cygni posted:I don't have personal experience, but going by reddit/techtuber sentiment, ASUS (specifically the high end X670E boards) seems to be bad news at the moment for AM5. Lots of reports of dead boards, strange issues, and bad ram compatibility.
|
|
#
?
Mar 22, 2023 21:37
|
|
|
Well it looks like Microcenter isn't carrying AM5 ASRock boards. Drat.
|
|
#
?
Mar 22, 2023 22:04
|
|
|
Hockenheim posted:Well it looks like Microcenter isn't carrying AM5 ASRock boards. Drat.
|
|
#
?
Mar 22, 2023 22:12
|
|
|
They don't have them in the faceted search dropdown filters in the motherboard section but this URL should be sufficient to find ASRock AM5 motherboards https://www.microcenter.com/search/search_results.aspx?Ntk=all&sortby=match&N=4294966996+4294817783+4294805803&myStore=true
|
|
#
?
Mar 22, 2023 22:28
|
|
|
Just to brag: Got today an 7950X cpu, with Asus X670E MB and 6000 DDR5 ram. So far so good, POST-ed the first try, no smoke (yet). Idle temps reported by the MB is 50C for the CPU. However, both Linux and Windows report 35C, with the cpu fans spinning at about 600-800 RPM. Compiled a little program of mine (which pulls in and compiles a shitload of libraries), the temps spiked occasionally to 95C. It felt indeed warmer in the room, warmer than with my old Intel. I'm guessing it's fine. I hope so at least. Haven't updated the bios yet, which seems to be needed as per few posts above, will try that next.
|
|
#
?
Mar 24, 2023 06:09
|
|
|
If everything works and your memory is running with their EXPO profiles enabled, then a BIOS update probably isn't necessary. Some motherboard manufacturers have been improving their memory support via BIOS updates and have sometimes even improved their auto timings for memory, so do update if that's a concern. Hitting 95C a few times during heavy loads is pretty normal for the 7950X. The X-sku chips tend to run warm, and the 7950X by default can pull up to 240 or 250W. You can tame it pretty easily by turning on eco mode and setting the TDP down one level from stock and only lose a few percent of performance, if the heat output is a concern.
|
|
#
?
Mar 24, 2023 06:56
|
|
|
Delid and direct die that bitch. I'm going to assume your cooler can handle the 250 watts but its all getting trapped under the IHS.
|
|
#
?
Mar 24, 2023 07:20
|
|
|
No easy way to directly mount the HSF after delidding unfortunately due to AM5's design but if you did decide to delid then there is a easy way to do with dental floss. https://www.youtube.com/watch?v=BQ00B93w8hY Ironing it afterwards to get the rest of the metal TIM off seems a bit risky the floss method is cheap and safe for delidding at least.
|
|
#
?
Mar 24, 2023 11:37
|
|
|
It's not enough to cut through the adhesive with dental floss. You can't delid the CPU using that method without melting the solder TIM, hence the iron (though there are probably safer ways to go about it). Using a delidding tool like derbauer's gets around this by repeatedly moving the IHS back and forth, which physically disrupts the solder joint. It's the safer method because there's no heat involved and the tool seems designed to be incapable of damaging anything else on the substrate. It just takes more effort to do the delidding and cleanup. That said, this is not the solution to moderately high temps one should immediately jump to. Especially when those temps are actually totally normal for this particular chip. Dr. Video Games 0031 fucked around with this message at 11:48 on Mar 24, 2023 |
|
#
?
Mar 24, 2023 11:43
|
|
|
You're right its not something you should immediately go for. But it is a slick and cheap way vs buying or renting one of the more normal delid tools and I always get a kick out of seeing different ways to do this sort've thing.
|
|
#
?
Mar 24, 2023 12:02
|
|
|
wait until you see the direct die mounting solution they came up with  (no, i don't know what the lighter's purpose is)
|
|
#
?
Mar 24, 2023 12:28
|
|
|
Hahaha its pretty ghetto but if it works that is kinda slick. Still I think its been beat. I remember when someone made a water cooler by cutting a 20oz soda bottle top off, gluing it to either a Duron or Athlon (direct die watercooling!) CPU package, and used a 5gal bucket waterbong as the reservoir/heat exchanger with a Eheim pump. Apparently it worked but it looked super half assed! This is a different one but along those lines: 
|
|
#
?
Mar 24, 2023 13:34
|
|
|
I am absolutely dying at "universal LGA bracket" I've seen a bunch of stuff recently about delidding CPUs - what's the deal?
|
|
#
?
Mar 24, 2023 14:33
|
|
|
ChickenWing posted:I am absolutely dying at "universal LGA bracket" My general sense is to not worry about it unless you're going for overclocking records. Josh Lyman fucked around with this message at 14:53 on Mar 24, 2023 |
|
#
?
Mar 24, 2023 14:38
|
|
|
In the case you or someone else reading isn't familiar with the process at all--on top of the CPU die(s) is a metal plate, which is what you're actually looking at if you hold a CPU you bought in the last 15 years. This is the "integrated heat spreader" or IHS. The CPU die is connected do this with some kind of semi-fluid compound; it used to be thermal paste like you'd use yourself, these days it's liquid metal or straight up solder. The issue is that each "layer" adds some thermal resistance, and as Josh says above AM5 in particular has a weirdly compromised IHS design in order to maintain compatibility with AM4 coolers (personally I think this was a bad move in the long run). You can improve thermal conductivity by removing the IHS and its thermal interface material (which is the "de-lidding", you are taking its "lid" off) and then attaching your cooler/coldplate directly to the die(s), but this is much riskier because if you fail to cover the die(s) with your own paste/etc then they can hit thermal runaway much faster than the CPU can throttle and it can cook itself, or you can just mechanically chip or crush a die by mounting your cooler wrong, or slipping with a sharp object, or any number of other dumb and easy ways to instantly destroy your $800 chunk of thinking sand. It's really only something you consider when you both can afford a super high-end CPU that has these issues and you can also afford to just buy a new one if you gently caress up the process.
|
|
#
?
Mar 24, 2023 14:48
|
|
|
ahead of the diablo 4 beta, i just replaced some old nt-h2 paste with some mx-4 that was pretty much impossible to spread with a tool. the viscosity was very weird even though it was a legit seller, but i just ran cinebench which made temps spike to 90c. temps were bad before, but not that bad, as in ~79c for cinebench all-core does thermal paste need to settle or something over time, or is there just something wrong with the thermal paste most likely? bought the paste 3mo ago to replace some 3yo paste, so it shouldn't exactly be crusty at this point here are the temps with a bunch of stuff open with no dedicated stress test after resetting the history:  (fair amount of stuff open) i could also try tightening the cooler even more, but i'll just wait a bit to see if temps settle; "don't overtighten, and don't undertigten", sigh kliras fucked around with this message at 16:01 on Mar 24, 2023 |
|
#
?
Mar 24, 2023 15:55
|
|
|
Pretty sure the paste should work fine right off the bat, so something sounds wrong. I’ll just add that MX-4 had a different consistency from every other paste I’ve used (thicker, almost granular), so the paste itself might be unrelated to your issue.
|
|
#
?
Mar 24, 2023 16:01
|
|
|
was able to tighten the top screw a fair bit. was pretty tight the first time around, but maybe something gave way when i finished setting it up i don't think mx-4 is great for just spludging a paste pattern, at least with the weird viscosity my batch has. maybe i'll find some old expired noctua and hope it performs better with a more elaborate paste pattern that's what i get for trying to branch out from noctua paste. maybe nt-h1 is an easier alternative to the nt-h2. the later mx's haven exactly been perfect sequels either. was gonna try kryonaut, but it looks a little too much like my mx-4 in photos, which is ironic since i wanted a viscous paste that was easy to spread kliras fucked around with this message at 16:36 on Mar 24, 2023 |
|
#
?
Mar 24, 2023 16:31
|
|
|
Just glop a bunch in the center and let the pressure of the mounting mechanism spread it out for you, imo.
|
|
#
?
Mar 24, 2023 16:40
|
|
|
I had no problems spreading Kryonaut, it's less viscous than most pastes I've used but definitely nothing to worry about. They even include small tool to help.
|
|
#
?
Mar 24, 2023 16:45
|
|
|
i'm probably gonna try a five-dot pattern just so i have something to compare my backup nt-h1 with wish there were a simple benchmark that also reports temps. 3dmark's reporting system is a bit of a mess
|
|
#
?
Mar 24, 2023 16:47
|
|
|
Waiting in the D4 login queue, CPU temps (which now im monitoring for a few days at least) are between 45C and 65C. I ran this morning a simple (but long running) program, single threaded, which was consuming 100% of 1 CPU. The temps reported were ... funny, to say the least. Every other CPU (out of the 32 with hyperthreading), was at 70C while the odd numbered ones were are 45C. Only one core was at 100% usage, as expected. I thought it was interesting how 1 die could influence the temps in others. Either the linux sensors are wrong or the reported temps/sensors are broken in this cpu.
|
|
#
?
Mar 24, 2023 18:14
|
|
|
Linux drivers needed to read the thermal sensors in RPL are new, so very well might not be working right, especially depending on the kernel version you're on.
|
|
#
?
Mar 24, 2023 18:50
|
|
|
I bought a new 5900x system with 128GB of ram for server stuff, while trying to get the idle power down I noticed my rx 570 uses 20w. So just for fun I tried booting it without a gpu, and it worked!? How long has this been a thing, I swear a gpu was required before. Motherboard is an Asus Prime X570-P.
Perplx fucked around with this message at 19:47 on Mar 24, 2023 |
|
#
?
Mar 24, 2023 19:43
|
|
|
Perplx posted:I bought a new 5900x system with 128GB of ram for server stuff, while trying to get the idle power down I noticed my rx 570 uses 20w. So just for fun I tried booting it without a gpu, and it worked!? How long has this been a thing, I swear a gpu was required before. Motherboard is an Asus Prime X570-P. I can no longer clearly remember a time when consumer-grade systems WOULD NOT run headless, it's been so long. There definitely was a time that was the case, but it's been irrelevant for so long that I've forgotten. Definitely at least as far back as the HyperTransport days.
|
|
#
?
Mar 24, 2023 20:04
|
|
|
Some consumer specific boards will still whine (or, worse, display a bright-rear end on board error LED forever. ASUUUUSSS!) when running headless. Thankfully, some have a bios setting now where you can disable VGA detection/error reporting, but obvi youll need a GPU for first boot unless the board has a BMC.
|
|
#
?
Mar 24, 2023 21:33
|
|
|
Mendocino trip report: - DIRT Rally (1) at 1080p with everything set to low put out an average of 88 FPS - Project Wingman at 1080p with everything set to low was still below 30 FPS. I could cut resolution scaling to 50% and get to the mid-40s, but then it looked too ugly, so I set it to 75% resolution scale and capped the FPS at 30 and it was mostly able to hold it steady there while still keeping things legible trying to think of what else I could throw at this thing to put it through its paces. Old games like Civ 4 and Daggerfall Unity hold up just fine, as expected. are there any settings or apps I could tinker with to squeeze out a little bit more performance? I'm not exactly looking to run Cyberpunk on this thing (though I could try!) but just for The Science
|
|
#
?
Mar 25, 2023 05:13
|
|
|
|
| # ? May 18, 2024 05:42 |
|
|
There's not much you can do, the iGPU is pretty bad.
|
|
#
?
Mar 25, 2023 06:13
|
|
