|
Lube banjo posted:Though I have been having some weird temp issues lately, replaced cpu fan a few times because I can't figure out why I am getting up to 69c just watching a twitch video Is that maximum or a sustained average? If it's max then just ignore it, it doesn't matter. Ryzen 2k & 3k have a behavior where they will momentarily boost single-core to max clock using max volts, despite the system nominally being idle and the only 'work' being done is for example moving the mouse around. Even having a monitoring program open will generate micro-spikes (because the CPU boosts to fill the requests for sensor readings ASAP). It's weird and dumb, but that's how the AMD boost algorithm works. The thing is, this boost can spike the temp quite high no matter what cooler it has. The limits of heat conduction out of the individual core is literally what makes it so hot, not your heatsink or fan. So you will get momentary readings of 69C that vanish as soon as the next reading happens.
|
#
?
Mar 7, 2020 20:39
|
|
|
|
| # ? Apr 29, 2024 14:26 |
|
|
Klyith posted:Is that maximum or a sustained average? If it's max then just ignore it, it doesn't matter. Unfortunately 60c+ is sustained. I'm gonna repaste again... Precision boost off, and a -.03125 offset Setset fucked around with this message at 20:50 on Mar 7, 2020 |
|
#
?
Mar 7, 2020 20:47
|
|
|
orcane posted:As for fans... I guess some people have amazingly bad hearing or live next to a construction site or their GPU fans are so loud they don't care about the wannabe delta fans on their mainboards. In an article on AMD giving up blowers for future reference video cards, today I was reading people saying their Navi reference blowers were almost silent I am also convinced most people have major hearing damage based on what they consider "quiet". You're 3000 RPM blower cooler isn't barely audible and neither are your 2200 RPM case fans.
|
|
#
?
Mar 7, 2020 20:59
|
|
|
You should probably considerung not undervolting Ryzen 3000: https://old.reddit.com/r/Amd/comments/ceakbs/if_you_want_to_save_powerreduce_thermals_reduce/ 
|
|
#
?
Mar 7, 2020 21:01
|
|
|
Lube banjo posted:Unfortunately 60c+ is sustained. I'm gonna repaste again... Precision boost off, and a -.03125 offset Ooof, yeah something is wrong between the CPU and heatsink. Repaste, but also look real close at the mounting to see if it was off-axis or had asymmetric pressure. The mount is much easier to mess up than paste. If your case has an access hole behind the mobo so that the heatsink can be changed without taking it out of the case, it's still very good to at least set the case on its side while putting the heatsink on. Getting the mount right while the weight of the sink is pulling it down on one side is difficult. Best to have the sink sitting straight on the CPU on its own.
|
|
#
?
Mar 7, 2020 21:12
|
|
|
orcane posted:You should probably considerung not undervolting Ryzen 3000: I set mine to 1.3 volts in Bios (offsets I tried were unstable), I tested performance with Cinebench and lost expected amount of performance. I don't care though as it reduces the CPU temp spikes dramatically so the fans won't go up and down all the time when idling or doing light stuff which is super important to me. My asus mobo has fan smoothing up/down time option but it does nothing. I'll see this thing about PPT sometime soon though. E: Maybe even right now. E2: Uuh, would this even help with the spikes during idle use? I don't care about power use and thermals during high loads, my cooler handles that really well at acceptable noise levels. quote:This does not affect idle temperature. It's a power limit. Even with the current poor idle behavior, idle power is much less than any reasonable power limit setting. Meh, I'll just leave it as is then. Sininu fucked around with this message at 21:26 on Mar 7, 2020 |
|
#
?
Mar 7, 2020 21:13
|
|
|
Klyith posted:Ooof, yeah something is wrong between the CPU and heatsink. Repaste, but also look real close at the mounting to see if it was off-axis or had asymmetric pressure. The mount is much easier to mess up than paste. Aaand solved. I did the stupid thing that dumb people do when getting a new CPU cooler. I won't mention it, but it was dumb... orcane posted:You should probably considerung not undervolting Ryzen 3000: Those seem like extreme undervolts. -.03125 isn't even on the chart so its probably effectively pointless
|
|
#
?
Mar 7, 2020 21:26
|
|
|
60°C for heat transfer through plastic is pretty good? 
|
|
#
?
Mar 7, 2020 21:29
|
|
|
orcane posted:60°C for heat transfer through plastic is pretty good? I knew there was an issue when I ran timespy and hit 80c 
|
|
#
?
Mar 7, 2020 21:44
|
|
|
orcane posted:60°C for heat transfer through plastic is pretty good? Don’t forget mayonnaise and toothpaste...
|
|
#
?
Mar 8, 2020 11:02
|
|
|
EmpyreanFlux posted:Also all Apple, Huawei, Qualcomm, etc are moving to 7nmEUV+ and 5nm, which leaves AMD a considerable amount of wafers for all their projects as long as they remain on a tweaked 7nm node. Volume is more important than chasing marginal performance gains on 7nmEUV. HalloKitty posted:How old is your system? There's basically no better time than now to upgrade to AM4, because if Zen 3 turns out to be amazing, you can always upgrade with one of them in place, without changing anything else. Didn’t realise AM4 was eol this year, all I needed to know.
|
|
#
?
Mar 8, 2020 12:13
|
|
|
https://www.pcworld.com/article/3531309/ryzen-4000-mobile-laptops-could-hit-18-hours-of-battery-life-amd-says.htmlquote:Calling AMD’s upcoming 7nm Ryzen laptop CPUs a “watershed” moment for the company, AMD VP of computing and graphics Rick Bergman hinted that at least one model of laptop could hit “18 hours” of runtime. /doubt
|
|
#
?
Mar 10, 2020 02:19
|
|
|
SwissArmyDruid posted:https://www.pcworld.com/article/3531309/ryzen-4000-mobile-laptops-could-hit-18-hours-of-battery-life-amd-says.html isnt stuff in financial briefings supposed to be factual? like they can get sued for lying to investors or something right?
|
|
#
?
Mar 10, 2020 02:52
|
|
|
VostokProgram posted:isnt stuff in financial briefings supposed to be factual? like they can get sued for lying to investors or something right? "Obviously what isn’t known is which laptop Bergman meant, how large the battery was, and what test was used to derive the 18 hours." They can be factual while still using extremely favorable settings and a laptop with a gigantic battery.
|
|
#
?
Mar 10, 2020 03:16
|
|
|
The new Dell XPS 13 2020 model claims "19 hours" of battery life on Mobile Mark with it's Intel Ice Lake processor, so these claims aren't that out there, they're just based on synthetic benchmarks and things like running the screen at 30% brightness.
|
|
#
?
Mar 10, 2020 03:19
|
|
|
The United States posted:The new Dell XPS 13 2020 model claims "19 hours" of battery life on Mobile Mark with it's Intel Ice Lake processor, so these claims aren't that out there, they're just based on synthetic benchmarks and things like running the screen at 30% brightness. Hell, it even just says 18 hours of runtime. The screen might not even be on.
|
|
#
?
Mar 10, 2020 03:41
|
|
|
And it's very likely just quoting the OEM claim so Dell or HP is more on the hook. Not lying when the box says it* *18 hour battery life was obtained by leaving word and Excel open, 4 hours of streaming video. disclaimer pulled from my butt.
|
|
#
?
Mar 10, 2020 12:22
|
|
|
The United States posted:The new Dell XPS 13 2020 model claims "19 hours" of battery life on Mobile Mark with it's Intel Ice Lake processor, so these claims aren't that out there, they're just based on synthetic benchmarks and things like running the screen at 30% brightness. My previous work laptop was a 2016 or 2017 XPS 13. Whichever year they introduced the 4K screen, because I picked the 1080p model to get better battery life. My use case was "working remotely using Chrome (with a buncha tabs), Terminal (also with a buncha tabs), and Slack (sitting in a million channels), with the screen set to 'whatever i need to see comfortably in this cafe'". I could always work a full day, plus a bit of hanging out, and only needed to plug in at the end of the day. So back then I was getting 10-12 hours of always-on network terminal usage. Being handed a 2018 Macbook Pro with it's poo poo-rear end battery that only lasted about 5 hours sucked rear end after that. But my point is that you absolutely could get double-digit runtimes from laptops 3 years ago. It's just going to be tuned for low power consumption, and if your use case is "endless kernel compiles while playing FPS games" then you will be disappointed in multiple ways. mdxi fucked around with this message at 15:17 on Mar 10, 2020 |
|
#
?
Mar 10, 2020 15:15
|
|
|
Eh, my server rack is 6 feet behind me and half the time I'm in a datacenter. Guess it just doesn't bother me. 
|
|
#
?
Mar 11, 2020 05:02
|
|
|
.
sincx fucked around with this message at 05:50 on Mar 23, 2021 |
|
#
?
Mar 11, 2020 05:54
|
|
|
sincx posted:Wow Microcenter has the 3900X for $400. That's down entirely to what you're doing and your needs, but I'd go with a 3900X now at that price and then simply move to the 4900/4920/4950 etc later if your workloads need that unified L3 or rely more heavily on AVX.
|
|
#
?
Mar 11, 2020 11:32
|
|
|
EmpyreanFlux posted:... move to the 4900/4920/4950 etc later if your workloads need that unified L3 or rely more heavily on AVX. Why, has there been any specific claims about AVX changes on 4000 series?
|
|
#
?
Mar 11, 2020 12:08
|
|
|
peepsalot posted:Why, has there been any specific claims about AVX changes on 4000 series? I swear there was, but I can't seem to find it on a cursory google search. I think I'd still leverage better AVX performance since Zen 3 looks to be wider and fatter cores compared to Zen 2.
|
|
#
?
Mar 11, 2020 13:15
|
|
|
EmpyreanFlux posted:I swear there was, but I can't seem to find it on a cursory google search. I think I'd still leverage better AVX performance since Zen 3 looks to be wider and fatter cores compared to Zen 2. I remember a slide being linked in the thread at some point, my google on Ryzen says 1000 and 2000 series used 128bit and 3000 moved to 256bit, maybe that's what we're thinking of? Not seeing anything to support a move to 512 for 4000.
|
|
#
?
Mar 11, 2020 13:51
|
|
|
There was a rumour floated by the usual places late last year that Zen 3 is improving fpu performance by 50%
|
|
#
?
Mar 11, 2020 14:04
|
|
|
I like how people imply that those tiny x570 chipset fans must be super loud or else you are deaf or something. I had to open up my case stick my head near that fan and poke it with a finger to stop it and it indeed makes hardly any sound at all. The loudest things in my case on idle are either my PSU fan or the GPU fans but I can't really tell with those at the moment.
|
|
#
?
Mar 11, 2020 16:46
|
|
|
TheCoach posted:I like how people imply that those tiny x570 chipset fans must be super loud or else you are deaf or something. Not all the motherboards are the same.
|
|
#
?
Mar 11, 2020 17:22
|
|
|
TheCoach posted:I like how people imply that those tiny x570 chipset fans must be super loud or else you are deaf or something. My chipset fan is louder then my CPU fan at 900rpm. It is garbage.
|
|
#
?
Mar 11, 2020 17:34
|
|
|
Cygni posted:Not all the motherboards are the same. Exactly, and I agree with der8auer that these fans were not necessary. It sucks that some of them are loud but not all of them are like this and thus making blanked statements like "most people have hearing damage" about this is really dumb. EDIT: This board is MSI MPG x570 Gaming Plus, the VRM is supposed to be bad but I guess if you just run a stock 3700X there are no issues. TheCoach fucked around with this message at 17:41 on Mar 11, 2020 |
|
#
?
Mar 11, 2020 17:38
|
|
|
TheCoach posted:Exactly, and I agree with der8auer that these fans were not necessary. It sucks that some of them are loud but not all of them are like this and thus making blanked statements like "most people have hearing damage" about this is really dumb. I have this board and have never heard the fan, I am pretty sure it never goes on because fan logging has always shown it at 0 RPM.
|
|
#
?
Mar 11, 2020 19:30
|
|
|
If they're off or spinning slowly enough so you can't hear them they're not actually moving air and any $1 passive cooler could have the same or better effect. It's a simple fact that a 40mm fan needs to spin very fast to do anything, which is why the boards that shipped with more "proactive" fan curves are actually not quiet at all. Also, they're another part that can fail. The complaints about the fans are not just about their noise.
|
|
#
?
Mar 11, 2020 19:46
|
|
|
the VRM chart going around is really just for if you intend to use higher core count CPUs pretty much Just built this 3900x system last week and I found out that Windows Update demanding every CPU cycle can still bring the entire system to its knees 
|
|
#
?
Mar 11, 2020 20:33
|
|
|
orcane posted:If they're off or spinning slowly enough so you can't hear them they're not actually moving air and any $1 passive cooler could have the same or better effect. Until you imagine scenario where someone has SLI video cards with big bulky shrouds that block ambient circulation from getting to the heatsink, installed in a bad case that puts RGB intake fans directly behind an unventilated glass panel.  That guy needs a chipset fan. edit: gently caress the :2f2f: smilie is gone? drat! Klyith fucked around with this message at 20:39 on Mar 11, 2020 |
|
#
?
Mar 11, 2020 20:36
|
|
|
Plastic still on the glass smh
|
|
#
?
Mar 11, 2020 20:40
|
|
|
pixaal posted:I remember a slide being linked in the thread at some point, my google on Ryzen says 1000 and 2000 series used 128bit and 3000 moved to 256bit, maybe that's what we're thinking of? Not seeing anything to support a move to 512 for 4000. IF they have any plans for adding AVX512 in the next gen, that would be pretty sick, but that's a big ol' "citation needed".
|
|
#
?
Mar 11, 2020 20:47
|
|
|
peepsalot posted:I don't know about 1000 series, but my 2700 already supports AVX2, which is the 256bit version. Zen1/Zen+ run 256-bit operations on 128-bit units taking two cycles. Zen2 has 256-bit units and runs 256-bit operations in one cycle. So nominally it can do twice as much AVX work per cycle (obviously a program is not 100% comprised of AVX work so this is less in practice). That's why Zen2 is about 37% faster than Zen+ core for core and 61% faster than Zen1 in Handbrake x264 1080p. This is just one of the many things where Zen2 is waaaayyyyy faster than the earlier variants, it is a way way better core. 1600 AF (and 3000G) still own for dirt cheap builds but for a "normal" all around computer you really should just pony up the extra dough for Zen2, it is a much faster and much more consistent performer. I would 100% take a 3600 over a 2700X for a normal all-around desktop. x264 doesn't make huge use of AVX-512 as I recall but if they can run the 512-bit units as 2x256b units like Skylake-X can, then that would probably be another 30%-ish real-world boost. The numbers will be higher for x265 since that does use AVX-512 heavily, I'd even believe more than 50% speedup on that. Paul MaudDib fucked around with this message at 22:02 on Mar 11, 2020 |
|
#
?
Mar 11, 2020 20:52
|
|
|
Klyith posted:Until you imagine scenario where someone has SLI video cards with big bulky shrouds that block ambient circulation from getting to the heatsink, installed in a bad case that puts RGB intake fans directly behind an unventilated glass panel. 
|
|
#
?
Mar 11, 2020 20:57
|
|
|
I'm sitting on the sidelines because I refuse to go back to the bad old days of chipset fans. I'd buy an x470 -- it's not PCIE 4.0 that's stopping me, but I really want a USB-C header for a front USB-C port. That doesn't seem to exist, especially in an ITX format. Maybe I'm just misinformed.
|
|
#
?
Mar 11, 2020 21:21
|
|
|
Was the arrangement to use the old GloFo process for the chipset just temporary? I have to believe the x670 or whatever won't consume as much power
|
|
#
?
Mar 11, 2020 21:26
|
|
|
|
| # ? Apr 29, 2024 14:26 |
|
|
KS posted:I'm sitting on the sidelines because I refuse to go back to the bad old days of chipset fans. I'd buy an x470 -- it's not PCIE 4.0 that's stopping me, but I really want a USB-C header for a front USB-C port. That doesn't seem to exist, especially in an ITX format. Maybe I'm just misinformed. Asrock X470 Taichi has it, it’s ATX though.
|
|
#
?
Mar 11, 2020 21:30
|
|
