|
https://www.microcenter.com/product/510035/TUF_B450M-PLUS_Gaming_AM4_mATX_AMD_Motherboard How is this board? It's $50 after buying a CPU with it and I think I read that it will support PCIe 4 from CPU? They also have an open-box 3700X for $280.. $330 then 16GB DDR4 3200 would be about $65 so It would be around $400 to upgrade.
|
#
?
Jul 17, 2019 14:48
|
|
|
|
| # ? Jun 5, 2024 15:46 |
|
|
AMD put the kibosh on PCIe 4 on non-X570 board last i heard, not that it's very relevant. Can't speak to that board specifically, but if you're going to microcenter make sure it has a 'Zen3000 ready' sticker on it, that's how you know it had an updated bios.
|
|
#
?
Jul 17, 2019 15:27
|
|
|
My microcenter (tustin CA) is now out of every 3000 series ryzen except the $400 3800X. I have waited too long and now I am sad. 
|
|
#
?
Jul 17, 2019 17:09
|
|
|
fargom posted:My microcenter (tustin CA) is now out of every 3000 series ryzen except the $400 3800X. I have waited too long and now I am sad. Order from Best Buy. Their site still says that you'll be able to pick it up in store on the 22nd, at least for me here in California. Of course its possible you might be able to snipe it off Amazon if you're quick, or if some other site gets them in stock also. Best Buy is the only place where I see an actual date given for stock though. I managed to order a 3700x off Amazon but it went out of stock right after I finished my order.
|
|
#
?
Jul 17, 2019 17:12
|
|
|
Alpha Mayo posted:https://www.microcenter.com/product/510035/TUF_B450M-PLUS_Gaming_AM4_mATX_AMD_Motherboard It's not great - it'll work for a 3700X but it'd be pretty marginal for a 3900X and the feature set is lacking compared to mid-range ATX B450s. I'd consider getting the ASRock B450 Pro4 instead, which should also save you another $30. It has the same tier VRM, but a slightly better audio chip, 8 rear USB ports instead of 6 (including 2 USB 3.1 gen 2 where the TUF just has one), and a second SATA-only M.2 slot. Unfortunately the only B450 mATX that has a better feature set and VRM is the Mortar, which is discontinued in the US. E: Like ItBreathes said, don't buy it for the PCIe 4.0 - news keeps going back and forth on whether it'll be supported, but it seems unlikely at this point. If you really want PCIe 4.0, get an ASRock X570m Pro4, which will also get you two full PCIe 4.0 x4 M.2 slots (but costs $175). Stickman fucked around with this message at 19:11 on Jul 17, 2019 |
|
#
?
Jul 17, 2019 19:09
|
|
|
The Gate posted:Also true, and I'll be honest I haven't built a PC in 7 years so like, I wanted this time to go as smoothly as possible. I'm kind of feeling this but part of me is really excited to set up nothing but a PSU and a motherboard outside of the case and flash the BIOS with only a single led to tell me what's going on.
|
|
#
?
Jul 17, 2019 19:57
|
|
|
After reading multiple posts, here and on reddit, about undervolting and clock speeds on 3X00 CPUs, I took another look at my own 3900. What I found was that it, indeed, was running at lower clocks than the rate I had set in BIOS -- with a desired clock of 3.55GHz, I had one core running at that speed, most running at various speeds between 3.3GHz and 3.1GHz, and one core running at 2.9GHz. This was confusing, because I did all of my initial testing with a constantly running load of scientific workunits. I knew that it took 10 to 20 seconds for the processes to get everything they wanted loaded in memory and start crunching in earnest, which I had allowed for. What I didn't notice until I took a second look, was that clock speeds would hold stable for a few seconds after that before starting to diverge from my desired speed. I just didn't watch long enough to see the behavior people have been talking about for the past ~day. I watched the system for hours to assure myself of stability, but I trusted that clocks were not deviating. I'm now re-evaluating clocks and voltages, and going much more slowly than before. So far I've established a baseline of 3.45GHz, with all cores stable at that speed (within 10MHz of it, which is perfectly normal in my experience). Vcore is set to 1.0875V and Vcore offset at "auto"*, which is pulling 96W under load. That's a lot more than the 78W I was seeing with Vcore offset at -100mV, so I assume that the system is using that offset to step voltages up to where it needs them to stabilize clocks at 3.45GHz. Experimentation continues... * On my mobo, setting Vcore offset to 0 actually toggles it back to auto, so I suppose I need to set it to some marginal amount like -1mV to get actual control over voltages and find a sweet spot. mdxi fucked around with this message at 20:38 on Jul 17, 2019 |
|
#
?
Jul 17, 2019 20:36
|
|
|
I'm seeing posts on reddit that indicate that undervolting is undesirable and you should try limiting TTP limit as that hard limits your power draw and the CPU actually tries to optimize itself inside that power limit.
|
|
#
?
Jul 17, 2019 21:02
|
|
|
TheCoach posted:I'm seeing posts on reddit that indicate that undervolting is undesirable and you should try limiting TTP limit as that hard limits your power draw and the CPU actually tries to optimize itself inside that power limit. I saw that too, but I can't find anything in my BIOS settings that looks like that setting (either the name, or the description). I'd love a setting like that, because I know that the CPU's own voltage controls are way better at this than I am. Edit: Found it, 3 levels deep. Playing with that now. Edit 2: On an ASRock BIOS, it's not under AMD CBS or AMD PBS. It's at Advanced --> AMD Overclocking --> Accept --> PBO --> PBO Limits [Manual] -> PPT Limit [W] I'm done with this until the linux kernel knows how to accurately report the temperature of Zen 2 CPUs. My primary concern here is actually thermals; power limits are the knob that I can turn to affect that. mdxi fucked around with this message at 21:34 on Jul 17, 2019 |
|
#
?
Jul 17, 2019 21:04
|
|
|
ItBreathes posted:AMD put the kibosh on PCIe 4 on non-X570 board last i heard, not that it's very relevant. I've seen some tested and while they are able to do numbers higher than PCIe 3, none of them quite meet the PCIe 4 standard, and as a result AMD can't have anyone advertising things falsely.
|
|
#
?
Jul 17, 2019 21:28
|
|
|
Per-CCX OC sounds like it could be a fun thing in a Zen 3 arch. The CPU market is so awesome right now.
|
|
#
?
Jul 17, 2019 23:02
|
|
|
ufarn posted:Per-CCX OC sounds like it could be a fun thing in a Zen 3 arch. It feels like the P4/AthlonXP days all over again. 
|
|
#
?
Jul 17, 2019 23:39
|
|
|
Athlon XP all the way.
|
|
#
?
Jul 17, 2019 23:48
|
|
|
Is there a way to turn off the RGB LEDs on the wraith prism cooler? I've got it on a CentOS machine so I can't use the windows software to do so.
|
|
#
?
Jul 18, 2019 00:44
|
|
|
Cojawfee posted:Athlon XP all the way. My dude.
|
|
#
?
Jul 18, 2019 00:46
|
|
|
Actuarial Fables posted:Is there a way to turn off the RGB LEDs on the wraith prism cooler? I've got it on a CentOS machine so I can't use the windows software to do so. I was actually wondering about this too. Can you not just unplug the connector from your mobo?
|
|
#
?
Jul 18, 2019 00:52
|
|
|
DeadFatDuckFat posted:I was actually wondering about this too. Can you not just unplug the connector from your mobo? I only have a single cable plugged in to the CPU fan header on my board. Unplugging it would also remove power to the fan.
|
|
#
?
Jul 18, 2019 00:56
|
|
|
Actuarial Fables posted:I only have a single cable plugged in right now. Unplugging it would also remove power to the fan. Oh shoot, I thought there was a separate cable for the lights. NM then.
|
|
#
?
Jul 18, 2019 00:57
|
|
|
DeadFatDuckFat posted:Oh shoot, I thought there was a separate cable for the lights. NM then. There are two optional cables meant for connecting to either a USB or an RGB header. Maybe if I connect the USB cable it'll stop, as it'll wait for a controller. e. Nope Actuarial Fables fucked around with this message at 01:07 on Jul 18, 2019 |
|
#
?
Jul 18, 2019 01:01
|
|
|
Depending on your motherboard, you might be able to plug it into an RGB header and disable it in the bios.
|
|
#
?
Jul 18, 2019 01:16
|
|
|
Zen 3 is still going to hit AM4 and SP3r2 sockets, right? Still debating whether I should sit it out for a year.
|
|
#
?
Jul 18, 2019 02:07
|
|
|
Actuarial Fables posted:Is there a way to turn off the RGB LEDs on the wraith prism cooler? I've got it on a CentOS machine so I can't use the windows software to do so. Does the hardware store the settings? You could hook it up to a windows machine and turn it off then hook it back to your other machine.
|
|
#
?
Jul 18, 2019 02:24
|
|
|
skylined! posted:Does the hardware store the settings? You could hook it up to a windows machine and turn it off then hook it back to your other machine. Worth a shot. I'll give it a go tomorrow and report back.
|
|
#
?
Jul 18, 2019 02:33
|
|
|
Combat Pretzel posted:Zen 3 is still going to hit AM4 and SP3r2 sockets, right? Still debating whether I should sit it out for a year. To the best of our current knowledge.
|
|
#
?
Jul 18, 2019 02:45
|
|
|
I hope there's going to be some more elaborations as to quad-patterning DUV vs. EUV and what that'll do for power efficiency and max. clocks, as well as new rumors for Zen 3 and what's so special to it (there's been allusions to that in the past). Don't really care for the rumored 3-way SMT in Zen 3, tho.
|
|
#
?
Jul 18, 2019 03:17
|
|
|
I went to swap HSFs today and found that the thermal paste was so stuck I ended up tearing off the lid by accident. gently caress me, right? I went and grabbed a 3600 and stuck it in my Asus B350M-A, everything seems to run fine and it benches like it should... Except Windows only sees 16GB of 32GB of memory, while CPU-Z/BIOS see the full 32GB. I have two sets of CMK16GX4M2B3000C15 that were bought at different times, one reports as Micron DRAM and the other Hynix. E: updated the bios and it no longer posts lol e2: even after using the cmos reset jumper and pulling the cmos battery  nice.
BlackMK4 fucked around with this message at 04:35 on Jul 18, 2019 |
|
#
?
Jul 18, 2019 03:49
|
|
|
Combat Pretzel posted:I hope there's going to be some more elaborations as to quad-patterning DUV vs. EUV and what that'll do for power efficiency and max. clocks, as well as new rumors for Zen 3 and what's so special to it (there's been allusions to that in the past). https://imgur.com/Xu3PSy6 7FF+ is the EUV process on that chart. 7FF is the current DUV process. So I'd expect some decent power efficiency improvements and/or a very modest clock increase from 7nm EUV. The clockspeed limitation with Zen2 probably has more to do with its design than its process. Either that or AMD is binning the bejeezus out of the good dies for server parts and desktop is being stuck with the left overs. Combat Pretzel posted:Don't really care for the rumored 3-way SMT in Zen 3, tho.
|
|
#
?
Jul 18, 2019 05:21
|
|
|
This CCX OC stuff is really interesting, from the people I have seen so far it looks like a typical 3900X consists of one good bin CCD and one more average bin. The good bin CCD can have the two CCXs run at 4.45-4.5+ while the other CCD tops out around 4.35GHz, which is why all-core OCs of the 3900X have tended to top out around there. Having 6 of the cores running at 4.5+ GHz should give some actual gains over stock Precision Boost where manual all-core OC and PBO + Auto OC at this point really aren't, cant wait to see someone do a full test of this.
|
|
#
?
Jul 18, 2019 06:20
|
|
|
BlackMK4 posted:I went to swap HSFs today and found that the thermal paste was so stuck I ended up tearing off the lid by accident. gently caress me, right? I went and grabbed a 3600 and stuck it in my Asus B350M-A, everything seems to run fine and it benches like it should... You mean it ripped the heat spreader off of the CPU? A few months ago I had a similar thing happen with my 2700x, I took the fan off but the whole CPU came out with it. The socket lock on AM4 seems less than robust. When this happened I put the CPU back in, used less paste, and tried again only to find half of my RAM wasn’t being recognized. I took the CPU out one more time, reseated it, and everything was fine.
|
|
#
?
Jul 18, 2019 13:33
|
|
|
PC LOAD LETTER posted:I think this mostly says it all here as far as possible process improvements go for EUV vs DUV:
|
|
#
?
Jul 18, 2019 13:36
|
|
|
GutBomb posted:You mean it ripped the heat spreader off of the CPU? I did this with an AM2 socket. CPU came right out and bent every single pin. I can't even remember why I was taking the heatsink off, probably something dumb like replacing the stock paste with arctic silver for no reason other than someone said it's better. I was a broke highschool kid at the time and had a WoW raid in an hour. I spent the next hour with a tiny screw driver from a glasses repair kit and a bright light and bent them back in. That CPU continued to run perfectly fine for years afterwards without any signs of being worse for wear. I'm sure it still works I bet it's still in my parent's basement where it was stored after I built the replacement computer.
|
|
#
?
Jul 18, 2019 13:50
|
|
|
|
|
#
?
Jul 18, 2019 13:55
|
|
|
GutBomb posted:You mean it ripped the heat spreader off of the CPU? Yeah it can happen with any ZIF socket, not just AM4 but older Intel and AMD ZIF's you could do the same thing. To avoid it happening you're supposed to twist the HSF back and forth a bit while sliding it off to the side to pop it free and break the suction effect from the thermal paste so you don't pull the CPU out of its socket. Combat Pretzel posted:Hmmm, the 10% power reduction does nothing for me. I hope eventually there's going to be elaborations as to what improvements are coming with Zen 3.
|
|
#
?
Jul 18, 2019 15:51
|
|
|
PC LOAD LETTER posted:I think this mostly says it all here as far as possible process improvements go for EUV vs DUV: 7FF+ will probably include improvements to the standard cell libs to allow for higher clocking transistors where needed, and the area/power improvements will come for free basically. It's probably best to think of it as 7nm++ in intel parlance (well, 10++ since intel 10 ~ foundry 7). Intel 10 is rumored to have really awful clocks though so the era of 5ghz stock processors might be temporarily over.
|
|
#
?
Jul 18, 2019 16:10
|
|
|
https://www.reddit.com/r/Amd/comments/ceakbs/if_you_want_to_save_powerreduce_thermals_reduce I think this is the thread where y'all have been undervolting. Tried this?
|
|
#
?
Jul 18, 2019 18:21
|
|
|
Holy poo poo the wraith prism is a piece of poo poo with its locking mechanism. It unseated my 3900x and broke a pin off.
|
|
#
?
Jul 18, 2019 20:48
|
|
|
e: oh its the socket A clip back from the dead, yeah good ol clips
|
|
#
?
Jul 18, 2019 20:51
|
|
|
Funny that the "lesser" cooler that came with my 3600 required me to remove that dumb bracket.
|
|
#
?
Jul 18, 2019 20:56
|
|
|
Yeah, that's one thing I've always hated on AMD sockets was those god drat clips. That was the first thing to come off after I built my first AMD build in years back in 2017 and got a third party cooler (Noctua NH-D15).
|
|
#
?
Jul 18, 2019 21:06
|
|
|
|
| # ? Jun 5, 2024 15:46 |
|
|
Whenever I built an AMD PC, I was always worried I was going to break the drat thing when putting those clips on.
|
|
#
?
Jul 18, 2019 21:08
|
|