|
SwissArmyDruid posted:Intel’s 10-core Comet Lake-S CPUs could draw up to 300W I don't disagree, but every single Intel processor for the last 2.5 years has been branded as "intel's FX-9590", you get desensitized to it after a while difference is the 9900K/KS still lead the charts on gaming despite being a fundamentally 4 year old architecture, and AMD is only going to have something fully equal with another year and another generation, let alone surpassing it by any significant margin (say, as much as the 9900KS leads now) which will probably take another generation after that. (although the battle for multithreading/productivity has clearly been lost since Zen2 of course) This one is going to be interesting though because 2 more cores and a clock increase, this may finally be the breaking point where the spec (especially the "thermal velocity boost") becomes fictional because you just can't cool the thing. These really will need to be delidded/direct die cooled to see their full potential. Stock it's probably just going to slam to 95C and then drop to 4.8 GHz or so. Intel may push up tjmax even further... some partners recommended increasing it to 115C even on 9900K (which is kinda unnecessary tbh but) Paul MaudDib fucked around with this message at 16:55 on Jan 9, 2020 |
#
?
Jan 9, 2020 16:47
|
|
|
|
| # ? May 3, 2024 09:27 |
|
|
It's not really an FX-9590 because it is performant still, it's just a drat furnace in getting there. They're replacing it with Rocket Lake, IIRC a 14nm Tiger Lake in 2021. That's an even wider core, and greater amount of cache and likely even more a pain to cool and feed but at least it won't fall behind Zen3.
|
|
#
?
Jan 9, 2020 19:28
|
|
|
Plus it'll overclock.
|
|
#
?
Jan 9, 2020 19:43
|
|
|
GRINDCORE MEGGIDO posted:Plus it'll overclock. I have trouble believing these are going to even hit stock clocks reliably let alone overclock but I guess the fact that they are advertising a stock 5.2 GHz (5.3 with TVB) on something that isn't a hyper-binned KS sku says maybe they got a little more out of 14++++ or the skylake architecture in general?
|
|
#
?
Jan 9, 2020 19:50
|
|
|
It feels like if they keep pushing this old process they'll never be able to catch up with 10nm. I would have expected performance/frequency to plateau at some point so they can sell Tiger Lake (or whatever) as an improvement on the Desktop. Guess AMD won't let them do that. Not that I'm complaining...
|
|
#
?
Jan 9, 2020 21:41
|
|
|
eames posted:It feels like if they keep pushing this old process they'll never be able to catch up with 10nm. they have no other option. 10nm was broken. 10+ is scheduled for late this year. They have to keep 14nm going until 10nm can relieve it
|
|
#
?
Jan 10, 2020 00:43
|
|
|
10nm is likely never coming to the desktop
|
|
#
?
Jan 10, 2020 01:58
|
|
|
10nm for desktop seems restricted to Intels first gen of Xe GPUs, Tigerlake might get better clocks but I think the calculus is that they need the clock+uarch advantage otherwise it'll be Zen3 topping charts as Tigerlake-S struggles to catch it, compared to Rocket Lake which'll be a furnace but uarch+clock should beat Zen3 still.
|
|
#
?
Jan 10, 2020 02:40
|
|
|
i built a 200ge a300 deskmini with 32gb of ddr4-2933 (running at ddr4-2666) ram, all parts gotten for relatively cheap (110 for SODIMM, 50 for 200GE, only the deskmini chassis was a bit overpriced at $140) then just used leftover SATA drives for storage the upgrade path is to be flashed to run a zen 2 renoir (4700G with cTDP set to 45w?) in the future when AM4 is EOL and good chips are on markdown then i will possibly have built the ultimate "budget-sane" tinybox Anime Schoolgirl fucked around with this message at 04:23 on Jan 10, 2020 |
|
#
?
Jan 10, 2020 04:06
|
|
|
I put together a system that has a Ryzen 7 2700x using the stock Wraith Prism cooler/thermal paste, and put it in a Cooler Master Q300L case that includes a 120mm exhaust fan. I added two 140mm intake fans. I've been running smallFTT prime95 for about 20mins now, and Ryzen Master reports the temp is around 82-83c. All fans are properly at 100% speed. The case is pretty small, but I've read it's thermals are ok. Is that load temp safe?
|
|
#
?
Jan 10, 2020 04:51
|
|
|
.
sincx fucked around with this message at 05:50 on Mar 23, 2021 |
|
#
?
Jan 10, 2020 05:13
|
|
|
U know I meant S series goddungit 
|
|
#
?
Jan 10, 2020 05:20
|
|
|
teagone posted:I put together a system that has a Ryzen 7 2700x using the stock Wraith Prism cooler/thermal paste, and put it in a Cooler Master Q300L case that includes a 120mm exhaust fan. I added two 140mm intake fans. a. That's an expected temperature for the stock cooler on an all-core stress test load. The prism is an actually decent cooler but all 8 cores on the 2700X is a lot of watts. same reuslt here b. No, the Q300L is not a good case for thermals. Case fans are super inefficient in it because the vent holes are so small -- there's more metal than hole on all the mounting points. GN gave the full-size version it's disappointment award. The mATX version is a bit better just because the PSU isn't re-exhasting to the inside of the case like the ATX, but it's still not good. The whole thing is covered with holes, so it would probably be ok just passively exchanging air as a low-workload low-noise machine, NAS box, or HTPC case. But it's not a performance case.
|
|
#
?
Jan 10, 2020 05:57
|
|
|
Klyith posted:a. That's an expected temperature for the stock cooler on an all-core stress test load. The prism is an actually decent cooler but all 8 cores on the 2700X is a lot of watts. same reuslt here Ok cool, thanks for the reassurance haha. Yeah, I left Prime 95 smallFFTs on for a little over an hour and the CPU temp stayed dipped to 82C and went up as high as 85C, but mostly stayed around ~83-84C. quote:b. No, the Q300L is not a good case for thermals. Case fans are super inefficient in it because the vent holes are so small -- there's more metal than hole on all the mounting points. GN gave the full-size version it's disappointment award. The mATX version is a bit better just because the PSU isn't re-exhasting to the inside of the case like the ATX, but it's still not good. Makes sense. It was the cheapest case at the time that my brother wanted for this build because it was the smallest size mATX case within his budget, and he also liked the way it looked. I was expecting thermals to be better, but idle temps at like ~35-45C with random spikes into the low-to-mid 50s here and there doesn't seem too bad.
|
|
#
?
Jan 10, 2020 08:47
|
|
|
teagone posted:Makes sense. It was the cheapest case at the time that my brother wanted for this build because it was the smallest size mATX case within his budget, and he also liked the way it looked. I was expecting thermals to be better, but idle temps at like ~35-45C with random spikes into the low-to-mid 50s here and there doesn't seem too bad. Depends if you've got a hot video card in there as well, but with 2 140s you're probably getting enough fresh air that things won't totally choke. It's just kinda a waste because the 140s will have to fight hard. Fans make more noise when they don't have free flow, and you'll have to set the speed higher to get enough air. I think it could also be improved with way less effort than GN put into embiggening all their holes (and way cleaner results). Just cut out a selection of metal around the perimeter of the fans -- the outer end of the fan blades is what does most of the work. Like so:  No need for a step bit, just a dremel or nibbler. The problem is you can't go too nuts on removing metal because those side panels are the structure of the case.
|
|
#
?
Jan 10, 2020 16:12
|
|
|
Upscaled good. Did a video on EUV, if you need a refresher or don't know why Intel's problems are what they are: https://www.youtube.com/watch?v=oIiqVrKDtLc
|
|
#
?
Jan 10, 2020 17:44
|
|
|
Klyith posted:Depends if you've got a hot video card in there as well, but with 2 140s you're probably getting enough fresh air that things won't totally choke. It's just kinda a waste because the 140s will have to fight hard. Fans make more noise when they don't have free flow, and you'll have to set the speed higher to get enough air. I was wondering why the 140mm Noctua fans were being so loud lmao. And yeah, the GPU runs hot: it's an XFX RX 580 GTR Black Edition that gets up to like 87C when gaming. Hmmm, maybe I'll tell my bro to swap cases when his budget allows since I'm not so keen on using a dremel. My bro might be into it though. I'll let him know. Thanks!
|
|
#
?
Jan 10, 2020 19:16
|
|
|
GRINDCORE MEGGIDO posted:Plus it'll overclock. Most people either can't or won't install enough cooling to deal with much over 200W already. For them a 2x 120mm AIO is about the max they'll deal with. Its why most people never even get their 9900K's to 5Ghz+ and instead tend to run them at ~4.7Ghz or so. In practical terms its probably going to end up overclocking about as well as the FX9590 did and might actually run into real problems running at stock settings if the 300W TDP is real.
|
|
#
?
Jan 11, 2020 10:53
|
|
|
SwissArmyDruid posted:Upscaled good. Did a video on EUV, if you need a refresher or don't know why Intel's problems are what they are: https://www.youtube.com/watch?v=oIiqVrKDtLc I am always amazed at the scale of these machines, the size they work at and the physical size needed to work at such a small scale. Kinda weird that the whole hangup is about how to store the masks without contaminating them.
|
|
#
?
Jan 11, 2020 12:34
|
|
|
Filthy human trash should not mix with clean cold machines. Also yay, we've officially graduated from merely laser-forged computer chips to plasma-forged. The future is here. 
|
|
#
?
Jan 11, 2020 12:36
|
|
|
 Looks like there's a batch of 1600s that are actually 1700s floating around.
|
|
#
?
Jan 11, 2020 14:50
|
|
|
That article is over two years old.
|
|
#
?
Jan 11, 2020 14:52
|
|
|
Hahaha holy poo poo that teaches me to pay attention 
|
|
#
?
Jan 11, 2020 14:58
|
|
|
Does anyone have solid numbers on how quickly Zen2 cores throttle up when they transition from having nothing to do to being busy? Googling for p-state switching latencies didn't yield anything.
|
|
#
?
Jan 11, 2020 16:14
|
|
|
ConanTheLibrarian posted:Does anyone have solid numbers on how quickly Zen2 cores throttle up when they transition from having nothing to do to being busy? Googling for p-state switching latencies didn't yield anything. CPU-controlled P-states switch super fast. Probably limited by how fast the VRM can respond to CPU asking for more volts? AFAIK the only thing that takes non-trivial time is coming out of core parking, ie cores completely turned off. Back when Ryzen came out this was a problem because Windows would turn off all but one core in light tasks & one-core loads, but then still does the thing where it rotates a task to different cores. AMD put the Ryzen balanced power plan in the AMD chipset driver package, it sets the minimum processor state to 90%, which means that cores never park and the CPU stays in p-states. But that isn't needed anymore because windows 10 updated to also not park cores on Ryzen if you're using the standard balanced power plan. Official AMD rep on reddit: quote:4a) However, the Ryzen Balanced plan still sets a minimum clockspeed of 90% on a core that is actively under load. This eliminates some small latency penalties that occur when ramping a CPU from low clock to high clock. This will give the Ryzen Balanced plan a small edge in select cases. It's a few percent, and I've only seen it measured in synthetic workloads.
|
|
#
?
Jan 11, 2020 21:42
|
|
|
ConanTheLibrarian posted:Does anyone have solid numbers on how quickly Zen2 cores throttle up when they transition from having nothing to do to being busy? Googling for p-state switching latencies didn't yield anything. Back in the first gen of Intel speedstep (long-rear end time ago) it was measuring p-state latencies in the nano-seconds. c-state transitions were in the microseconds, so still very low. The problems generally were in applications that created bursty workloads that would trigger rapid cycling between p or c states. The embedded power management controllers are considerably smarter than that now and know to leave things ramped up for a little longer to help mitigate against that specific problem.
|
|
#
?
Jan 13, 2020 16:11
|
|
|
AT has apparently confirmed with AMD and their partners that the rumored "TRX80" and "WRX80" chipsets for a future theoretical 8 channel version of threadripper don't exist. If you want 8 memory channels, looks like you gotta buy Epyc. On the flipside, if you were holding out on buying 3960X/3970X/3990X in case 8 channel versions happened, seems safe to go ahead and buy. https://www.anandtech.com/show/15359/trx80-and-wrx80-dont-exist-neither-does-the-intel-lga1159-socket
|
|
#
?
Jan 13, 2020 23:07
|
|
|
No surprise, 8 channel TR just sounds like Epyc anyway. BangersInMyKnickers posted:Back in the first gen of Intel speedstep (long-rear end time ago) it was measuring p-state latencies in the nano-seconds. c-state transitions were in the microseconds, so still very low. The problems generally were in applications that created bursty workloads that would trigger rapid cycling between p or c states. The embedded power management controllers are considerably smarter than that now and know to leave things ramped up for a little longer to help mitigate against that specific problem. Perfect.
|
|
#
?
Jan 14, 2020 00:24
|
|
|
Hey quick question: Between Tdie and Tctl, which should I care about for evaluating processor temperature?
|
|
#
?
Jan 16, 2020 02:45
|
|
|
GunnerJ posted:Hey quick question: Between Tdie and Tctl, which should I care about for evaluating processor temperature? Tdie, if the two are different then tctl is higher because your cpu has a temp offset. Tdie is also whichever core is hottest, ie each core has its own sensor and the reported temp is the highest.
|
|
#
?
Jan 16, 2020 06:04
|
|
|
Hey goons. So I have a MSI Tomahawk non-Max mobo with a AMD 3600 and the bios settings have my CPU fan at 100% no matter the temps. These are not changeable. What software can I download to put in place a proper curve?
|
|
#
?
Jan 16, 2020 09:38
|
|
|
Try Argus monitor
|
|
#
?
Jan 16, 2020 09:54
|
|
|
Do you have the latest bios? It sounds like there was an issue with fan curves on non-MAX a few months ago, but I'm not sure if it's been resolved. I believe MSi's Command Center app will let you set fan curves. You could start with that if you don't want to shell out for Argus.
|
|
#
?
Jan 16, 2020 10:27
|
|
|
OhFunny posted:Hey goons. So I have a MSI Tomahawk non-Max mobo with a AMD 3600 and the bios settings have my CPU fan at 100% no matter the temps. These are not changeable. What software can I download to put in place a proper curve? That 'auto detect' for the fan type didn't work for me (different mainboard though). Try setting it to PWM.
|
|
#
?
Jan 16, 2020 11:09
|
|
|
The fan speeds are also all set to 100%
|
|
#
?
Jan 16, 2020 11:17
|
|
|
Stickman posted:Do you have the latest bios? It sounds like there was an issue with fan curves on non-MAX a few months ago, but I'm not sure if it's been resolved. Yeah it's the latest bios. I'll give Command Center a try. Edit: Downloaded MSI CC and just hit Smart Mode and then default curve and this is much better. Edit 2: Ah this is something that needs to be set up after every boot. I'll have to keep looking for something more permanent. Edit 3: Selecting Restore Defaults in Bios fixed the fan curve there. Thank goodness. Moly B. Denum posted:The fan speeds are also all set to 100% Yes that is the problem. I can only change settings that are in brackets []. OhFunny fucked around with this message at 13:05 on Jan 16, 2020 |
|
#
?
Jan 16, 2020 11:56
|
|
|
Klyith posted:Tdie, if the two are different then tctl is higher because your cpu has a temp offset. That's what I was guessing but wanted to confirm, thanks!
|
|
#
?
Jan 16, 2020 13:01
|
|
|
NewFatMike posted:I run most of my 3D modeling software under a VM. All my 2D stuff works better under Linux, but it really comes down to "If my next update completely befucks my Solidworks install, won't it be nicer to just revert to a snapshot instead of reinstall everything? Again?" Reasonable and thought-through justification, I'll allow it
|
|
#
?
Jan 16, 2020 16:17
|
|
|
First B550 leaks coming out. https://twitter.com/VideoCardz/status/1217865943398998018 Wccf (lol) thinks that B550 and A320 will go into production this quarter. They also reported a "rumor" that Zen3 and the 600 series chipsets will launch in early 2021, so possibly a longer window than expected. Considering AMD is supply constrained anyway at the moment, might make sense.
|
|
#
?
Jan 16, 2020 19:11
|
|
|
|
| # ? May 3, 2024 09:27 |
|
|
Should that be A520? Because A320 already exists 
|
|
#
?
Jan 17, 2020 12:37
|
|
