|
orcane posted:VRM "tier lists" are dumb in many ways, but the B450 I Aorus Pro Wifi is still listed as a "mid range" option for Ryzen 2000 on the old X470/B450 list and the newer CPUs actually have a lower power consumption so Mid range means running running a highly overclocked 2700x. Yes, it makes no sense  Spiderdrake posted:There is definitely no manufacturer vrm tier list, given they vary from board to board immensely My post was specifically about the available b450 itx boards, should've specified that.
|
#
?
Dec 30, 2019 18:44
|
|
|
|
| # ? Jun 5, 2024 14:54 |
|
|
Lungboy posted:Aye it seems to be specific to the itx fatality board. Presumably some % of ppl using itx formfactor are somewhat on the go
|
|
#
?
Dec 30, 2019 19:25
|
|
|
CrazyLoon posted:Just a quick note on this, if it performs as it should you will literally not even hear it over all the other fans + if you have any kind of decent aircooling from the front, I seriously doubt it's even needed. There were some early BIOS versions that made them annoying, but as of this past month they are utterly beneath notice imo. On the ASRock X570 Phantom Gaming the chipset fan is obnoxious and audible for sure. The southbridge runs up 65+degC just on idle. The fan becomes less obnoxious if you remove the little metal cover over it, though presumably that's not doing anything special with regards to deflecting air on purpose or whatever. It is also heat piped to a metal surround over the I/O area of the board, which I would imagine they wouldn't do this if they didn't think that this thing was going to put out a ton of heat, and would need to draw it away. This fan the size of a bee's dick though just makes noise and isn't particularly effective. Fan off it settled in at 70C, fan on it's at 67C, which tells me it just isn't great at removing heat from that big sink, compounded by running ITX in a SFF enclosure.
|
|
#
?
Dec 31, 2019 00:39
|
|
|
Well. It's not like a more elegant and passive chipset cooling solution exists where a heatpipe is chained between the northbridge to the VRMs like on over a decade old X48 mobos...Oh wait.
|
|
#
?
Dec 31, 2019 01:12
|
|
|
I mean it is what it is, and there's not going to be any great place to put a heat load like that, or NVMe or whatever, just something to keep in mind.
|
|
#
?
Dec 31, 2019 01:54
|
|
|
I still don't understand why there's only one board doing that (some X570 E-ATX monstrosity from Gigabyte) instead of the tiny poo poo fans that don't really do much. They're already making $500 boards, just take one of the 10 boards and make it $510 so you can put $5 worth of heatpipes on there...
orcane fucked around with this message at 01:57 on Dec 31, 2019 |
|
#
?
Dec 31, 2019 01:54
|
|
|
To clarify the board has a heatpipe and a fan, it just... doesn't have a lot of distance from other heat sources and the fan sucks. Yellow is the pipe, which is just stuck in a groove on the shroud without any real bonding or anything other than physical contact.
|
|
#
?
Dec 31, 2019 02:47
|
|
|
Lungboy posted:There seems to be lots of reports on various forums of the ASRock b450i fatality board having huge issues with 3000 series ryzen chips. Has anyone experienced it? I'd rather not fork out the extra for the Asus board but at least it seems to work. I built 4 machines around this board, and it was a rock-solid performer until I started upgrading from 2700s to 3900Xs. It is.. curiously twitchy. I've bored the thread with the details before, so I won't go into it now. But I would not recommend using it with high core-count (and thus high wattage) 3x00 CPUs. I'd expect it to be fine with anything in the 65W range, because it was fine with 2x00 series chips in that power envelope.
|
|
#
?
Dec 31, 2019 04:26
|
|
|
mdxi posted:I built 4 machines around this board, and it was a rock-solid performer until I started upgrading from 2700s to 3900Xs. 3900X is not really any higher-power than a 2700X. And VRM overheating is pretty straightforward: the chip will start to massively, clearly throttle during long-duration runs, it doesn't make a board "curiously twitchy". It sounds like the board has general stability problems with 3000 series chips and you're ascribing that to the power, which isn't necessarily the case. kind of a bummer because iirc this was one of the few itx boards known to support running headless? And X570 is kind of dumb on a itx board, where it is often used in situations with limited airflow and/or a small chipset fan would be noisy and obnoxious. Maybe Asrock will do a B550 that supports running headless... Paul MaudDib fucked around with this message at 04:55 on Dec 31, 2019 |
|
#
?
Dec 31, 2019 04:51
|
|
|
Paul MaudDib posted:And VRM overheating is pretty straightforward: 3600x/3700x will run fine on anything. Even 3900x will run fine on most things unless you go for the $20-$40 motherboard special for your $500 CPU. VRM quality is really overhyped for the most part. People were predicting crazy OC headroom and all this stuff and it turns out zen2 gains next to nothing from it so the over provisioned VRMs were a big waste outside of "I overclock every weekend it's my hobby to place higher on the rankings" types. It's extremely unlikely any VRM is under provisioned enough to cause stability issues with a 3900x or lower CPU. Plenty of people have ran them on A series boards as a joke and they were stable although had worse performance, and those boards have worse VRMs than the board we're talking about now. I agree with Paul that it's probably some other issue with the board. Most likely bios. mdxi posted:I built 4 machines around this board, and it was a rock-solid performer until I started upgrading from 2700s to 3900Xs. Khorne fucked around with this message at 05:17 on Dec 31, 2019 |
|
#
?
Dec 31, 2019 04:55
|
|
|
Khorne posted:The only thing I'd be worried about with certain VRMs, probably not the board we're talking about, is operational temperature effecting lifespan. VRMs are rated for a certain lifespan within a certain temperature range. When people use awful A series or B350 or like the $40 b450 boards for a 3900x or better the VRMs operate at a high enough temperature to reduce VRM lifespan from like a decade to a few years. Whether it's good enough for a 105W CPU like a 2700X/3900X in the first place is of course orthogonal to the discussion of whether switching from 2700X to 3900X will "create problems".  (which it really shouldn't because they're the same TDP/power limits.)I mean it's a cheap lovely board, of course it's not going to last as long as a premium board will. Other stuff very well may break sooner too. By all means don't pick a board with a derp-tier VRM if there's something better in that price range. And if you can arrange the build so there's some airflow over the VRM, go for it. But overall it's not super productive to fret about VRM temps. You're not Buildzoid doing XOC runs, and you're probably not going to have the thing slammed to 100% with prime95 all the time. If it dies in 5 years, well, you're still coming out ahead buying two lovely boards instead of one premium board, probably. This is one reason why I think people jerk off a little too much over the "upgradeability" of AM4. If in 5 years you put a new CPU on your board, are you really expecting that board to run to a full 10 years with no problems? And do you really trust the VRMs to still run a high-TDP CPU in their senescence? It's fine to plan on a hand-me-down retirement for a family member with a 65W chip or something but it's probably not a super great idea to plan on dropping a 4950X onto your B350 board that you bought 4 years earlier for $75. Plenty of computers do run past 5 years of service perfectly fine (and are often repairable if you are dedicated enough to recap the board and stuff like that) but that's about where reliability starts to fall off. And I mean, that can even happen with premium boards. I've had X99 boards die after 3 years or so, it's just a numbers game. If you know you're going to slam it hard on a constant basis for years at a time then by all means aim upwards but even 105W really isn't that big a deal. Paul MaudDib fucked around with this message at 05:57 on Dec 31, 2019 |
|
#
?
Dec 31, 2019 05:14
|
|
|
It totally kills me how much more L3 cache AMDs have compared to Intel. 64MB on the 3900x.. wow.
|
|
#
?
Dec 31, 2019 05:20
|
|
|
Khorne posted:Twitchy how? I'm genuinely curious. It seems like everyone with the board is having problems with zen2, even with 3600s. The 11/28 bios may have helped significantly because it seems like people who bought it recently are having fewer issues. Did you update any to that?
One of the machines is no longer using that motherboard. It just plain died overnight (these things happen; I don't blame the board), and I replaced it with the stupidly expensive Aorus mITX x570 board because I'm tired of having problems with these machines. It worked perfectly out of the box, and it runs a few degrees cooler than the other three, so it looks like the tiny chipset fan is doing at least a little something. Maybe that effect is less noticeable on a larger board, but even before moving to 3900s, I had learned that if you don't have air moving over the heat-spreaders/heatsinks on these tiny boards, the PCB will start acting as a sink and raise your CPU temps  Assuming that I do another round of upgrades when the 4x00 series CPUs come out, I'm ditching mITX and going to uATX. I wanted to build the most compact stack of machines possible, but the added cost and restricted choice in mainboards hasn't been worth it. Paul MaudDib posted:It sounds like the board has general stability problems with 3000 series chips and you're ascribing that to the power, which isn't necessarily the case. Sure there's dozens of other differences, but in my mind, the motherboard shouldn't care about what's on the other side of the socket. Also the weird, intermittent PSU cycling indicates to me that something is fucky with power delivery. But mostly I feel comfortable pointing at the power systems because the weirdness goes away once I underclock/undervolt the CPUs, and returns anytime the BIOS settings get wiped. I am, of course, aware that this is all correlation, and not causation. But it's where I'd start looking if I was going to try to find the root cause. I don't intend to though. I just intend to not use this specific board anymore. mdxi fucked around with this message at 08:50 on Dec 31, 2019 |
|
#
?
Dec 31, 2019 08:39
|
|
|
Partycat posted:To clarify the board has a heatpipe and a fan, it just... doesn't have a lot of distance from other heat sources and the fan sucks.
|
|
#
?
Dec 31, 2019 09:22
|
|
|
TDP is far from a complete descriptor of how something draws power. I would not be surprised at all if it were in fact a power delivery issue caused by Zen2 being in some way more demanding/picky.
|
|
#
?
Dec 31, 2019 09:25
|
|
|
it's probably the same reason why some RTX cards tax power supplies, likely very high power draw for very short bursts despite not using much more power than pascal overall
|
|
#
?
Dec 31, 2019 10:18
|
|
|
The United States posted:Um, what is going on with the mounting holes on that AMD motherboard? That particular ASrock motherboard uses Intel mounting holes for wider heatsink compatability.
|
|
#
?
Dec 31, 2019 10:19
|
|
|
The United States posted:Um, what is going on with the mounting holes on that AMD motherboard? It uses Intel mounting holes, presumably so that the cooler can be rotated to fit it better. There’s not much clearance there no matter which way you turn it. Also meant the bling cooler that came with the processor is useless.
|
|
#
?
Dec 31, 2019 10:23
|
|
|
redeyes posted:It totally kills me how much more L3 cache AMDs have compared to Intel. 64MB on the 3900x.. wow. Isn't this not directly comparable because AMDs L3 is subdivided, with cores having unequal access? Or is that just for the Rome dies?
|
|
#
?
Dec 31, 2019 15:31
|
|
|
Twerk from Home posted:Isn't this not directly comparable because AMDs L3 is subdivided, with cores having unequal access? Or is that just for the Rome dies? they are 1 i/o die and 1 cpu complex so they are pretty equal access i thinks.
|
|
#
?
Dec 31, 2019 15:52
|
|
|
Twerk from Home posted:Isn't this not directly comparable because AMDs L3 is subdivided, with cores having unequal access? Or is that just for the Rome dies? It's 32mb per chiplet of max 8 cores, so yeah not 100% comparable on the 3900 up, still pretty substantial though.
|
|
#
?
Dec 31, 2019 15:57
|
|
|
It was the 4 chiplet processors where it stood out. Since they're rectangular accessing the cache on a chiplet on the far edge has slightly higher latency than the one on the same edge.
|
|
#
?
Dec 31, 2019 16:16
|
|
|
The 32 MB per chiplet is also split into two 16 MB slices, one for each 4-core CCX. Changing that so the full 32 MB is available to all cores on the chiplet is one of the biggest changes confirmed for Zen 3/Ryzen 4000.
|
|
#
?
Dec 31, 2019 16:16
|
|
|
Looks like the 48-core 3980X may exist afterall. Listing for it found in CPU-Z. Now we will see if it ever makes it to market. https://www.techpowerup.com/262509/amd-ryzen-threadripper-3980x-is-a-48-core-monster-for-when-64-cores-are-too-many-32-too-few
|
|
#
?
Dec 31, 2019 18:00
|
|
|
My 3600X maxes out at 66C while gaming, is that okay? What's the max?
|
|
#
?
Jan 1, 2020 03:53
|
|
|
Pittsburgh Fentanyl Cloud posted:My 3600X maxes out at 66C while gaming, is that okay? What's the max? You want to stay below 75C if possible, yours is a fine result as long as it's boosting properly. Paul MaudDib fucked around with this message at 04:40 on Jan 1, 2020 |
|
#
?
Jan 1, 2020 04:00
|
|
|
Boosting to 4400mhz on all cores except two which go to 4375, looks good to me.
|
|
#
?
Jan 1, 2020 04:13
|
|
|
Twerk from Home posted:Isn't this not directly comparable because AMDs L3 is subdivided, with cores having unequal access? Or is that just for the Rome dies? Zen2 doubled the cache so now you have 16MB per CCX (4 cores). Then you also get another 16MB on the other CCX, or up to another 32 MB on the other CCD. So any one core can access the same amount of cache as a 9900K in single-threaded, and you have twice as much cache per core total. It's really a crazy amount of cache and undoubtedly helps smooth the IO die thing and gaming performance in general. Source games (maybe GldSrc and source2, or Respawn Engine?) loving love having assloads of cache available and actually run faster on Zen2 now. Paul MaudDib fucked around with this message at 04:49 on Jan 1, 2020 |
|
#
?
Jan 1, 2020 04:45
|
|
|
Paul MaudDib posted:You want to stay below 75C if possible, yours is a fine result as long as it's boosting properly. 75c sounds very conservative. Throttle temp is 95c and that’s nowhere near anything that would cause damage or degradation.
|
|
#
?
Jan 1, 2020 05:36
|
|
|
GutBomb posted:75c sounds very conservative. Throttle temp is 95c and that’s nowhere near anything that would cause damage or degradation. 95C is where it goes into catastrophic-failure limp-home mode. It starts reducing clocks around 75C or 80C depending on microcode version. This is the part where you insert a drawn-out semantic debate over whether your chip "reducing boost" is really the same thing as "throttling" but whatever you call it, you the user at the end of the day are getting reduced performance because your chip isn't cool enough, at a much much lower number than 95C. It's not a hard wall, it's throttling back very gently, but it's still throttling. Pascal and Vega both do this as well fwiw - Pascal really needs to be kept below 60C for the maximum boost states to be available, there is a soft throttle up to 70C, a moderate throttle through the 70s, then a rapid throttle past 85C. So I guess if you want you can think of me being super conservative to try and coax the absolute best performance out of the chip. It's certainly not unsafe to go higher, it's just not running the boost as hard as it could. Paul MaudDib fucked around with this message at 06:01 on Jan 1, 2020 |
|
#
?
Jan 1, 2020 05:55
|
|
|
I'm getting ready to put together my new build, and I can't remember one small detail: the thermal paste that comes pre-applied to the stock heatsink is adequate, yes? I won't need to go out and buy a syringe unless I decide to replace the stock cooler later on down the line? It's been ten years since I built a system, and I remember mucking around with paste back then because I wasn't sure if I needed extra paste or not.
|
|
#
?
Jan 1, 2020 19:00
|
|
|
Use the stock paste, it'll work fine.
|
|
#
?
Jan 1, 2020 19:11
|
|
|
I imagine the biggest difference between 92% of thermal paste brands is the label on the box Just a guess that there's only really so many factories squeezing it out 
|
|
#
?
Jan 1, 2020 20:10
|
|
|
GRINDCORE MEGGIDO posted:Use the stock paste, it'll work fine.
|
|
#
?
Jan 1, 2020 23:02
|
|
|
I forget which site it was, but they did a giant comparison of pastes / TIMs and included toothpaste and chocolate, among a few others.
|
|
#
?
Jan 1, 2020 23:08
|
|
|
Statutory Ape posted:I imagine the biggest difference between 92% of thermal paste brands is the label on the box I've seen quite a few vids on the subject. Most brands seem to be a few degrees of each other.
|
|
#
?
Jan 1, 2020 23:08
|
|
|
movax posted:I forget which site it was, but they did a giant comparison of pastes / TIMs and included toothpaste and chocolate, among a few others. Much more recently: https://www.hardwaresecrets.com/does-mayonnaise-last-as-a-thermal-compound/
|
|
#
?
Jan 1, 2020 23:35
|
|
|
Buy some Panasonic soft pgs and just reuse it
|
|
#
?
Jan 2, 2020 00:36
|
|
|
Twerk from Home posted:Much more recently: https://www.hardwaresecrets.com/does-mayonnaise-last-as-a-thermal-compound/
|
|
#
?
Jan 2, 2020 03:28
|
|
|
|
| # ? Jun 5, 2024 14:54 |
|
|
Surely there has to be some useful reason for Miracle Whip to exist! By way of an update, my new system is put together and everything seems to be functioning properly so far, at least until I install
|
|
#
?
Jan 2, 2020 05:05
|
|