|
Trolling Thunder posted:I assumed both things as well — who knows how long my CPU was operating like that, but thought it had thermal protections too. General checklist of things you've probably already done, but: Pull the board and PSU out of the case and use the box the board it came in as a test bench. Make sure the CPU aux power cable is plugged in and the connector looks good. Plug the PSU into a different power outlet in your house. Try jumping the board without the power button by shorting the pins on the board directly. Try 1 memory stick and move it to different slots. Try booting with nothing connected (not even GPU/USB/SATA/Network) If your board supports it, do BIOS flashback and reflash the bios.
|
#
?
Jan 12, 2022 00:10
|
|
|
|
| # ? Apr 19, 2024 19:53 |
|
|
Cygni posted:General checklist of things you've probably already done, but: I'd also not rule out the PSU as the problem even if another system fired up with it, if that other system was lower draw. I've seen PSUs that were dying but put out enough power to seem to work, but shut down once the draw got too high. If you want to be really exhaustive, pull the motherboard out of the case and see if it will start up, in case there's a short or something.
|
|
#
?
Jan 12, 2022 01:34
|
|
|
Those are all great suggestions and I’ve done all of it as a good goon should. Thanks for putting that together. The symptoms are the same on a replacement mobo with working psu, so time for a CPU upgrade I suppose. Take care, everyone.
|
|
#
?
Jan 12, 2022 01:53
|
|
|
Igorslab writeup about the AM5 socket: https://www.igorslab.de/en/amds-socket-am5-v2-in-detail-what-is-better-with-ryzen-than-with-intels-socket-lga-1700-for-alder-lake/ Most interesting detail to me: when AMD said ""AM4 coolers will interoperate", they may have only been referring to clip-style ones (the default mount).
|
|
#
?
Jan 12, 2022 20:26
|
|
|
mdxi posted:Igorslab writeup about the AM5 socket: https://www.igorslab.de/en/amds-socket-am5-v2-in-detail-what-is-better-with-ryzen-than-with-intels-socket-lga-1700-for-alder-lake/ Looks to me like coolers that mount to a standard AM4 backplate will continue to work with the standard AM5 backplate. The AM5 backplate is a little different than AM4, however, in that it has some extra holes that the AM5 CPU retention hardware screws in to. Thus, if you have an AM4 cooler that ditches the standard backplate for some insane reason, you'll need an AM5-compatible backplate to go with it.
|
|
#
?
Jan 12, 2022 21:22
|
|
|
Helter Skelter posted:Looks to me like coolers that mount to a standard AM4 backplate will continue to work with the standard AM5 backplate. That is a reasonable interpretation, and I hope the correct one. The analysis on igorslab is always great, but the translation is sometimes very tiresome to work through (for me).
|
|
#
?
Jan 12, 2022 22:30
|
|
|
afaik both Intel and AMD have only changed their mounting holes when the CPU height has changed enough to make mounting force a potential problem. some of the 115x series has slightly different heights and contact area, but they kept it the same 'cause stuff would work well enough to not fail even though some older cooler might be kinda meh on a different CPU.Helter Skelter posted:Thus, if you have an AM4 cooler that ditches the standard backplate for some insane reason, you'll need an AM5-compatible backplate to go with it. Pretty sure the only coolers with non-standard AM4 backplates were during the launch of first-gen ryzens, where they had existing coolers designed for mounting to an included backplate for intel, and just did the quick & dirty update. So if you still have one of those it may be worth moving on from anyways. If it's a waterblock it probably doesn't have a great microfin pattern for the modern asymmetric ryzen. And old towers with direct contact heatpipes are kinda meh these days as well.
|
|
#
?
Jan 13, 2022 03:23
|
|
|
Klyith posted:afaik both Intel and AMD have only changed their mounting holes when the CPU height has changed enough to make mounting force a potential problem. some of the 115x series has slightly different heights and contact area, but they kept it the same 'cause stuff would work well enough to not fail even though some older cooler might be kinda meh on a different CPU. It's weird that direct contact heatpipes was a fad that came and went. What was the deal with that?
|
|
#
?
Jan 13, 2022 04:09
|
|
|
I'd theorize that as die sizes get smaller (and get put into areas other than dead center on package like Ryzen), direct contact heatpipes means only a couple pipes are carrying most of the load - the pipes aren't transferring heat to one another. Also alot harder to get the all important flatness from a bunch of pipes soldered together vs a plate of metal.
|
|
#
?
Jan 13, 2022 08:04
|
|
|
Direct contact heatpipes are most definitely still a thing in lower end air coolers. They aren't going to be as flat and you have to worry more about gaps but they're very cost effective so you see it a lot in the ~40 market segment.
|
|
#
?
Jan 13, 2022 15:43
|
|
|
Seamonster posted:I'd theorize that as die sizes get smaller (and get put into areas other than dead center on package like Ryzen), direct contact heatpipes means only a couple pipes are carrying most of the load - the pipes aren't transferring heat to one another. Very much this. In the right conditions direct contact does work better: heatpipes are more conductive than metal, so the less metal between the source and the pipe the better. But modern CPUs aren't those conditions. It also kinda limits how many pipes you can put in the heatsink and still be effective. More than the usual 4 heatpipes on a generic cheap tower and you start having to butt them up against each other. That tends to not work as well because two round things don't naturally have a lot of contact. VorpalFish posted:Direct contact heatpipes are most definitely still a thing in lower end air coolers. In particular they're cheaper because you can press-fit the pipes into the base, instead of needing to solder a top-plate and base-plate together.
|
|
#
?
Jan 13, 2022 16:45
|
|
|
Yeah limitation in the number of heatpipes is another reason - high end coolers will run like 6-7 heatpipes which you can't do unless you spread the energy over the base. But there are definitely some awesome budget coolers that use HDT, if you're not looking to spend nh-d15s money, a $40 HDT tower cooler should have no issues handling a 100w CPU load quietly enough.
|
|
#
?
Jan 13, 2022 17:11
|
|
|
I want to thank whoever shared this memory OC guide: https://github.com/integralfx/MemTestHelper/blob/oc-guide/DDR4%20OC%20Guide.md It has been very helpful. I'm slowly tightening the timings by changing one at a time and running a stress test overnight, so it might take a couple months to be finished with it, but I will eventually get it done. This is my personal machine but I also do my job's work on it (much faster than the laptop they gave me) so I'm taking it slow to minimize potential corruption issues. After I get all the timings dialed in, I'm going to reduce the initial voltages I set as part of the guide's recommendations and see how low I can get them while remaining stable.
|
|
#
?
Jan 13, 2022 19:23
|
|
|
Cygni posted:Is there anyone smarter than me here to tell me whether or not I should be pissed that Pluton is built into the 6000 series APUs? The takes I’ve seen range from “it’s just a TPM” to “it’s a full access hardware privacy nightmare which hands all of your information to Microsoft, of all people” Necroing this to note that an actual subject matter expert doesn't think it's likely to be a bad thing: https://mjg59.dreamwidth.org/58125.html (Matthew J. Garrett is a prominent Linux developer who's done a lot of work on Linux support for Secure Boot including use of TPMs, as well as various other things. So he knows what he's talking about, and would be more than willing to poo poo on Pluton if he thought it could do evil things.)
|
|
#
?
Jan 14, 2022 04:43
|
|
|
DigiTimes is reporting that TSMCs 3D packaging isn’t ready for prime time, with AMD choosing to direct what little vcache production they can get to the enterprise world. Probably not a surprise, and makes a lot more sense than the bad business-speak answers AMD gave. https://www.pcgamer.com/amds-upcoming-ryzen-7-5800x3d-may-end-up-as-a-limited-edition/ The Rumor Boys have also found reference to new 4000 series desktop parts. Renoir (Zen2) dies with deactivated IGPs. If that’s their plan to fight the 12400/12100 and fill the low end until AM5 shows up, gonna leave a lot to be desired.
|
|
#
?
Jan 14, 2022 08:55
|
|
|
Cygni posted:DigiTimes is reporting that TSMCs 3D packaging isn’t ready for prime time, with AMD choosing to direct what little vcache production they can get to the enterprise world. Probably not a surprise, and makes a lot more sense than the bad business-speak answers AMD gave. this was also suppose to be a test platform for 6000 series/am5 right?
|
|
#
?
Jan 14, 2022 16:27
|
|
|
I suppose the implication is that Zen4 won't have insane amounts of cache via die stacking? Or is the process expected to get higher yields soon?
|
|
#
?
Jan 14, 2022 18:49
|
|
|
Are EPYCs (Zen 3, i.e., 7443P) lockable like the TR PROs? Looking around for a used 7443P chip and I see some from HP servers.
|
|
#
?
Jan 14, 2022 18:59
|
|
|
Combat Pretzel posted:I suppose the implication is that Zen4 won't have insane amounts of cache via die stacking? Or is the process expected to get higher yields soon? Lol if 5800x3d ends up being faster than 7800
|
|
#
?
Jan 14, 2022 19:05
|
|
|
movax posted:Are EPYCs (Zen 3, i.e., 7443P) lockable like the TR PROs? Looking around for a used 7443P chip and I see some from HP servers. They can be if the OEM uses that feature, yes. For HPE in particular, they sent a message to STH in late 2020 - quote:... Thus, while we implement a hardware root of trust for our BIOS and BMC firmware, the processors that ship with our servers are not locked to our platforms. I do not know if they are still following the same practice now though.
|
|
#
?
Jan 15, 2022 00:36
|
|
|
When is the AMD CPU refresh coming? I play flight sims like IL-2 and DCS which are not very well multithreaded and perform well on CPUs with very fast single core speed. At the moment I have a Tomahawk Max, a 3700x and a 3080 Ti. I have been thinking about upgrading the 3700x to something like a 5800x or maybe even a 5900x. A lot of people online are saying that it would be a good idea to wait for the refresh. I would probably still buy an AM4 CPU since this will be the last CPU I can buy on an AM4 platform and I'm not keen on buying a completely new motherboard and DDR5 RAM. I think that the 5xxx generation will be cheaper when the new CPUs are released and hopefully will be reasonable value, but I'm curious how long we'll be waiting to see these new CPUs.
|
|
#
?
Jan 20, 2022 16:57
|
|
|
5800X3D refresh is Q1, that's all we know. You should wait for it
|
|
#
?
Jan 20, 2022 18:01
|
|
|
If you really want to eek out the last bit of single thread performance do you have a water cooler? AMD’s precision boost will clock slightly lower for every degree C above something pretty low like 50C or so.
hobbesmaster fucked around with this message at 23:34 on Jan 20, 2022 |
|
#
?
Jan 20, 2022 18:03
|
|
|
Zedsdeadbaby posted:5800X3D refresh is Q1, that's all we know. You should wait for it "Spring" is what AMD said at CES, and who knows what that really means. I'd expect it in late march or april, but that's really just a guess.
|
|
#
?
Jan 20, 2022 22:22
|
|
|
hobbesmaster posted:If you really want to eek out the last bit of single thread performance do you have a water cooler? AMD’s precision boost will go slightly higher for every degree C above something pretty low like 50C or so. every degree C below, or do I misunderstand?
|
|
#
?
Jan 20, 2022 23:27
|
|
|
Yes, I clarified it.
|
|
#
?
Jan 20, 2022 23:34
|
|
|
There are limits to how much AMD CPUs will boost, and it's more than possible to reach that limit using an air cooler. There's both a max boost clock (4.4GHz on the 3700X), and a hard power limit (somewhere around 78W by default for the 3700X, if I recall correctly). You can disable that power limit through ryzen master to allow the CPU to boost higher if it's running into that power limit, but you'll still run into the 4.4GHz max boost limit unless you do overclocking (and the potential for gains there isn't great on Ryzen CPUs). AMD's precision boost algorithms will also boost aggressively at temperatures well above 50C in my experience. I highly doubt a water cooler will suddenly unlock a noticeable boost to performance in video games, especially not single-threaded ones. You will almost always reach your max boost clock in single-threaded applications while using $30 budget coolers. Dr. Video Games 0031 fucked around with this message at 00:00 on Jan 21, 2022 |
|
#
?
Jan 20, 2022 23:54
|
|
|
hobbesmaster posted:If you really want to eek out the last bit of single thread performance do you have a water cooler? AMD’s precision boost will clock slightly lower for every degree C above something pretty low like 50C or so. I have it cooled with an H115i AIO which seems to be doing well at cooling it. Might well try disabling the power limit with Ryzen Master as someone mentioned. I'll wait to see how the 5800X3D, hopefully it will actually be worth upgrading to.
|
|
#
?
Jan 21, 2022 01:22
|
|
|
Dr. Video Games 0031 posted:"Spring" is what AMD said at CES, and who knows what that really means. I'd expect it in late march or april, but that's really just a guess. did they say which hemisphere?
|
|
#
?
Jan 21, 2022 01:33
|
|
|
The only hemisphere that matters.
|
|
#
?
Jan 21, 2022 05:38
|
|
|
Is a 3600X idling at 2.2ghz on all cores normal? Linux, if it matters. Cool n Quiet and global c states are enabled in the bios.
|
|
#
?
Feb 1, 2022 04:33
|
|
|
I just checked and my 3600X is currently going from under 2ghz to over 4ghz momentarily. This is with two web browsers open with like 30 total tabs or so. It does look my idle would probably a little lower than yours averaged out, but not that much lower. I'm looking at all 12 threads in kSysGuard, for what it's worth.
|
|
#
?
Feb 1, 2022 05:09
|
|
|
Hrm, thank you. Mine appear to have an actual floor or 2.2, but I've seen people online saying theirs go down to 400-800mhz. Just kinda odd.
|
|
#
?
Feb 1, 2022 16:45
|
|
|
BlackMK4 posted:Hrm, thank you. Mine appear to have an actual floor or 2.2, but I've seen people online saying theirs go down to 400-800mhz. Just kinda odd. Does Linux have something equivalent to AMD's Ryzen power plan on windows? My 3700X bottoms out at 2.2Ghz idle as well, on windows with the balanced power plan ("5%" CPU minimum). But with the Ryzen power plan it doesn't go below 3.5Ghz. But modern processors are so fast to switch between sleep and running that clockspeed isn't really a useful thing to judge what your minimum / low-power state is really like. When a core is asleep and you ask what the frequency is, the CPU tells you what the last clockspeed was (and that it's sleeping). So AFAIK the power strategy these days is to just shut down cores rather than clocking down to super-low speeds.
|
|
#
?
Feb 2, 2022 00:04
|
|
|
Different apps may report the clock speed in different ways. Some use what's called the "effective" clock rate which is often lower than the true one. Ryzen Master for instance tells me that my cores are clocked quite low right now: Task manager begs to differ, though:  And HWinfo is giving me different results from either:  So yeah, everything reports clock rates differently it seems. I wouldn't sweat it too much. The actual clock speed is likely what HWinfo reports under "core clocks." The 5600X rests at 3.6 GHz. 2.2 seems pretty normal for your CPU.
|
|
#
?
Feb 2, 2022 00:05
|
|
|
ryzen is especially tricky, because idle cores get gated off completely, and most software doesn't really get the notion of "core currently doesn't exist" and just reports the last valid result. i think ryzen master is still the only app that actually reports that bit correctly so if you have a core that was at >4ghz last time it did something, but then got shut off for idle, lol
|
|
#
?
Feb 2, 2022 00:22
|
|
|
Fair enough, I was trying to make sure I didn't have something misconfigured power wise.  Thank youI have an APC UPS connected to the machine so I can kinda pick out changes now. UPS nominal power is listed as 330w, so a load percent of 50% means the UPS is seeing 115w. The drop in average last night was decommissioning three old hard drives and fixing the hdparm sleep on the other four spinning drives.  
BlackMK4 fucked around with this message at 00:32 on Feb 2, 2022 |
|
#
?
Feb 2, 2022 00:29
|
|
|
My 3600 is hovering around 1000-1200MHz according to Ryzen Master as I watch a youtube video and type this, but Windows task manager says it's clocked at ~4GHz. I think the cores will go as low as 800MHz if I close every running program and do nothing Pretty sure most monitoring software just doesn't read the CPU speed correctly (or the CPU doesn't report it correctly to 3rd party stuff).
|
|
#
?
Feb 2, 2022 05:27
|
|
|
FuturePastNow posted:Pretty sure most monitoring software just doesn't read the CPU speed correctly (or the CPU doesn't report it correctly to 3rd party stuff). The CPU reports the same stats to any software, but it's kinda a matter of opinion as to how you compress those stats into a single "Ghz" number. Or even N cores of individual Ghz. Ryzen master may also be polling at higher rate than other software, since the only thing it pokes at is the CPU, so it's picking up the lows that other software misses. (The tendency is to report the highest clock from the polling period.)
|
|
#
?
Feb 2, 2022 17:48
|
|
|
|
| # ? Apr 19, 2024 19:53 |
|
|
movax posted:Are EPYCs (Zen 3, i.e., 7443P) lockable like the TR PROs? Looking around for a used 7443P chip and I see some from HP servers. Very much so. https://www.ebay.com/sch/i.html?_from=R40&_trksid=p2047675.m570.l1313&_nkw=epyc+dell+locked&_sacat=0 The practice started with epyc because it has some marginal use case in "the butt". Epyc processors have built-in PCH so the security stuff has to happen in the processor (more on that later). Since TR Pro is just an overclocked epyc instead of its own part, it has the same built-in PCH epyc does, and got the same misfeature. It's an incredibly dumb idea. Notably, Epyc does not have a few MB of flash for a BIOS, so that's still on board in SPI NAND chips or whatever. It's utterly trivial to allocate 1k or whatever of ROM on flash chips when you're talking dell/hp/lenovo volume, so that's where the key should have lived. Replacing a processor takes a screwdriver and a few minutes time. Replacing a surface-mount BGA takes complete disassembly and a specialized workstation. This isn't for security, it's for killing the secondary market.
|
|
#
?
Feb 6, 2022 00:39
|
|
