|
Natron posted:I'm on an x570 a pro, and I also get the custom BIOS splash screen, and I also get the lovely load time on boot. I think it must be something odd about the BIOS, because it takes almost twice as long to boot as my old i7 4770k on the exact same drive. Once it's booted the performance is better, of course, but it's a bit weird that my boot time is noticeably slower with a much faster 3900x in there. You can make it faster by manually setting all RAM timings and enabling various fast boot options, but it's still kinda slow. Hopefully a future bios update speeds it up. It has nothing to do with SSD or loading anything. The custom logo can be disabled but it doesn't change boot time at all.
|
#
?
Sep 1, 2019 13:07
|
|
|
|
| # ? Jun 3, 2024 21:42 |
|
|
Natron posted:I'm on an x570 a pro, and I also get the custom BIOS splash screen, and I also get the lovely load time on boot. I think it must be something odd about the BIOS, because it takes almost twice as long to boot as my old i7 4770k on the exact same drive. Once it's booted the performance is better, of course, but it's a bit weird that my boot time is noticeably slower with a much faster 3900x in there. DDR4 takes a lot longer to boot than DDR3 at the BIOS level because the DDR4 standard requires memory training on every boot. My old 4790K system spends far less at the BIOS booting than any DDR4 system I have seen firsthand.
|
|
#
?
Sep 1, 2019 16:38
|
|
|
uhhhhahhhhohahhh posted:Yep, everything on the front screen except BGS, which I can't find in my BIOS For some reason mine just really doesn't like Power Down mode/Gear Down mode disabled.
|
|
#
?
Sep 1, 2019 16:50
|
|
|
Golluk posted:For some reason mine just really doesn't like Power Down mode/Gear Down mode disabled. I don't know why mine is working now with Power Down disabled. Wonder if it's something to do with temperature? I was previously stable on 3466 with the old Fast timings, and power down disabled would cause instant memory errors. Doesn't make much sense for what's just a power saving feature. I haven't tried Geardown disabled again though.
|
|
#
?
Sep 1, 2019 17:51
|
|
|
I understand it correctly, Gear down isn't just power saving, it switches from t1 to t2 on the fly if the memory cant keep up with t1.
|
|
#
?
Sep 1, 2019 22:46
|
|
|
pixaal posted:I understand it correctly, Gear down isn't just power saving, it switches from t1 to t2 on the fly if the memory cant keep up with t1. PowerDown is the power saving feature.
|
|
#
?
Sep 1, 2019 23:16
|
|
|
Renoir (Zen 2 APU) rumors and leaks are starting to crop up: https://www.tomshardware.com/news/amd-renoir-apus-lpddr4x-4266-memory,40284.htm LPDDR4X 4266MHz support sounds pretty good for memory bandwidth, and I really hope Zen 2's fat chunks of cache will help with APU performance as well. Fingers crossed there'll be a 6C/12T APU with a good chunk of graphics and a thin mITX mobo to throw it into for science experiments.
|
|
#
?
Sep 1, 2019 23:44
|
|
|
NewFatMike posted:Renoir (Zen 2 APU) rumors and leaks are starting to crop up: If it's 6C/12T, it's also 8C/16T. I think that's going to be the way forward, and Komachi(sp?) also said it's 13-15CU, Vega/Navi hybrid (uses DCN 2.1). AMD isn't going to drop more CU on a die for no performance gain, so I suspect this hybrid has better bandwidth utilization or some kind of SRAM/L4 on the 12nm die. 13-15CU seems weird, but Raven was 11 so what do I know. There was also a Vega GPU found recently with 16CU used standalone with an i7 3930, so here is my guess. The upcoming Renoir uses a 7nm Zen2 die and a I/O die that's really a GPU, on 12nm. The Zen2 dies are optimized for low power usage as well. The Vega I/O die has 16CU and some features from Navi, and possible SRAM on it so 16CU is actually usable. It'll get RIS support as well. BUT, the Vega die will pull double duty. It'll feature on the CPU package, but is also completely capable of being a standalone GPU, and thus it'll fill in on the low end of the RX 5000 release, below the RX 5600 and RX 5500. I have no idea if it could be used as the B550 or A520 chipset to potentially enable display output, but that'd be interesting as all hell. It's also possible it'll get a different package between desktop and mobile, maybe mobile gets the space for HBM2? That'd allow AMD to sell those at a better premium since there'd be a noticeable performance uplift, and it's way more difficult to sell that on desktop. There is also another APU coming, apparently monolithic Dali, intended to be a low power solution. Maybe this uses a stripped down Zen2 core and RDNA2? Maybe even K12?
|
|
#
?
Sep 2, 2019 00:40
|
|
|
I really wish you hadn't told me what you wanted because now I really want that
|
|
#
?
Sep 2, 2019 01:31
|
|
|
pixaal posted:I understand it correctly, Gear down isn't just power saving, it switches from t1 to t2 on the fly if the memory cant keep up with t1. I turned it off to get CL15, but is that a performance loss since I'm at t2? Not really sure if I was at t1 before. Memory overclocking makes zero sense to me, and the overclocking calculator gives much worse timings than my stick's XMP profile.
|
|
#
?
Sep 2, 2019 03:30
|
|
|
Please please PLEASE let this finally be the SoC that matches a loving MX150.
|
|
#
?
Sep 2, 2019 04:40
|
|
|
?? Vega11 already matches MX150 provided you throw enough memory bandwidth at it. Granted, that was so hideously expensive when DDR4 prices were hosed you were better off just buying a gpu anyhow.
|
|
#
?
Sep 2, 2019 05:01
|
|
|
Seamonster posted:?? Vega11 already matches MX150 provided you throw enough memory bandwidth at it. Granted, that was so hideously expensive when DDR4 prices were hosed you were better off just buying a gpu anyhow. I mean, no? Vega 10 never matched the MX150 (well, the not lovely variant anyway, NVIDIA!  ) and it never helped that OEMs like to be lovely and do single channel memory constantly.A properly fed Vega 16 would blow the gently caress out of an MX250 and probably compete with an RX 560/1050 though, and you've finally reached 1080p low at that point in modern titles, 1080p high in anything esports.
|
|
#
?
Sep 2, 2019 14:12
|
|
|
Emailed Gigabyte support about identical settings being in 2 different places in the BIOS on the X570 Elite and they sent me version F4L 
|
|
#
?
Sep 2, 2019 17:40
|
|
|
Great deep dive on what boost speeds you should expect with Zen 2 by taking user surveys: https://www.youtube.com/watch?v=DgSoZAdk_E8
|
|
#
?
Sep 2, 2019 17:58
|
|
|
I'm jizzin' for Ryzen!
|
|
#
?
Sep 2, 2019 22:43
|
|
|
E - I now have two ryzen machines, and realised I should have just bought threadripper.
GRINDCORE MEGGIDO fucked around with this message at 00:41 on Sep 3, 2019 |
|
#
?
Sep 2, 2019 23:21
|
|
|
Pittsburgh Fentanyl Cloud posted:I'm jizzin' for Ryzen!
|
|
#
?
Sep 3, 2019 00:25
|
|
|
peepsalot posted:You said that already. It can't be stopped!!!!
|
|
#
?
Sep 3, 2019 00:31
|
|
|
Pittsburgh Fentanyl Cloud posted:It can't be stopped!!!! Can it at least rhyme?
|
|
#
?
Sep 3, 2019 01:25
|
|
|
ufarn posted:Great deep dive on what boost speeds you should expect with Zen 2 by taking user surveys: The takeaway : the 3600 is an even better buy than we thought, because you will absolutely be able to get some class action money from AMD at some point if you buy one.
|
|
#
?
Sep 3, 2019 02:56
|
|
|
K8.0 posted:The takeaway : the 3600 is an even better buy than we thought, because you will absolutely be able to get some class action money from AMD at some point if you buy one. what'd AMD do to make that a possibility?
|
|
#
?
Sep 3, 2019 05:41
|
|
|
MagusDraco posted:what'd AMD do to make that a possibility? I think it's questionable, given they perform slightly better than claimed, but they did lie about boost clocks. Especially if you look at peak effective speeds of single cores. Khorne fucked around with this message at 06:17 on Sep 3, 2019 |
|
#
?
Sep 3, 2019 06:11
|
|
|
(extremely FX defendant voice) yes but what is the definition of a
|
|
#
?
Sep 3, 2019 06:13
|
|
|
Pittsburgh Fentanyl Cloud posted:I'm jizzin' for Ryzen!
|
|
#
?
Sep 3, 2019 08:49
|
|
|
This is what happens when you don't take full advantage of your GAMECACHE.
|
|
#
?
Sep 3, 2019 12:23
|
|
|
AMD statement on the boost issue, they're working on it. https://twitter.com/AMDRyzen/status/1168901636162539536 edit: also https://twitter.com/1usmus/status/1168911500586749959 Paul MaudDib fucked around with this message at 17:47 on Sep 3, 2019 |
|
#
?
Sep 3, 2019 17:29
|
|
|
In order of people who actually get this fix Gigabyte ASUS ASRock ... ... Biostar ... ... ... MSI
|
|
#
?
Sep 3, 2019 17:58
|
|
|
EmpyreanFlux posted:In order of people who actually get this fix Maybe MSI will release this one quickly.
|
|
#
?
Sep 3, 2019 17:59
|
|
|
I can't even imagine what 3950 availability is going to look like. It took six weeks for the 3900 to actually be in stock at Microcenter, Newegg, and Amazon. And that lasted for just over a week. They've been out of stock everywhere, again, since about Thursday.
|
|
#
?
Sep 3, 2019 18:45
|
|
|
I'm more interested in seeing how many people sell their "placeholder" 3900s to chase those 8 extra threads.
|
|
#
?
Sep 3, 2019 19:07
|
|
|
Khorne posted:Didn't MSI have their bios guy(s) quit a few months ago? It makes sense for them to be slow during a transition period. I know Asus did, no idea about MSI.
|
|
#
?
Sep 3, 2019 20:11
|
|
|
Seamonster posted:I'm more interested in seeing how many people sell their "placeholder" 3900s to chase those 8 extra threads. I'm sticking with the 3900x until next gen.
|
|
#
?
Sep 3, 2019 20:16
|
|
|
My 2700 is going to keep me happy until the 4700X, I'm actually not greatly interested in Zen2 unless it's really the last gen for AM4, which okay I guess I upgrade to a 3900X in 2020.
|
|
#
?
Sep 3, 2019 21:51
|
|
|
Seamonster posted:I'm more interested in seeing how many people sell their "placeholder" 3900s to chase those 8 extra threads. Khorne posted:I'm sticking with the 3900x until next gen.  The 3950 is too power-hungry and too expensive for me. Four of them is $3000. My hobbyist scientific computing budget couldn't manage that even if I hadn't already upgraded 3 nodes to 3900s. Also, the enormous L3 cache is a huge part of what has made the 3900 so much more performant than the 2700 on my specific workloads. Adding another 8 threads drops the cache/thread to 2MB from the 3900's 2.666. Some of the workunits my machines handle want 4MB each. I'm aware that my concerns in this area are pretty niche though.
|
|
#
?
Sep 3, 2019 22:18
|
|
|
mdxi posted:
Out of interest, what do you use them for. My interests have always been gaming but I like reading up on what other uses people do with their PC's.
|
|
#
?
Sep 3, 2019 22:46
|
|
|
If you need high cache per core and more active corea make it slower, cant you just turn some cores off in BIOS?
|
|
#
?
Sep 3, 2019 22:52
|
|
|
EmpyreanFlux posted:My 2700 is going to keep me happy until the 4700X, I'm actually not greatly interested in Zen2 unless it's really the last gen for AM4, which okay I guess I upgrade to a 3900X in 2020. 
|
|
#
?
Sep 3, 2019 23:02
|
|
|
Not sure if my 3900X based system is idling a bit high (140 w +) because of firmware, motherboard (ASUS X570-Ace WS), PSU, or GPU (1080 Ti SC Hybrid). Only running 3 case fans and 2 fans in an AIO with a Toshiba XG3 NVMe SSD. Numbers aren’t really adding up for me unless the motherboard takes like 25+ watts. Also, not sure wtf is causing some full system lock for several seconds stuttering on Windows 10 sitting on the desktop with or without Game Mode on. Definitely haven’t heard that happening to anyone else and I’m on a fresh install so I’m trying to figure out if my RAM timings or HPET could possibly have an impact.
|
|
#
?
Sep 3, 2019 23:13
|
|
|
|
| # ? Jun 3, 2024 21:42 |
|
|
pixaal posted:If you need high cache per core and more active corea make it slower, cant you just turn some cores off in BIOS? iirc there is a gotcha here because Ryzen is not really 1x72 MB cache, it is 4x18 MB cache. Turn off the cores and you lose the cache attached to that CCX. You would have to turn off one core from each CCX just like the 3900X, and that may not be possible to do in BIOS. And even if you could, that probably wouldn't be worth another $250 for an extra 83 KB per thread, the 3900X is the optimum for him there anyway. They're right that the 3900X is an absurd amount of L3 for the money, the 3950X will only offer an extra 2MB at 50% higher cost. Paul MaudDib fucked around with this message at 23:20 on Sep 3, 2019 |
|
#
?
Sep 3, 2019 23:14
|
|