|
Daeus posted:Are many folks having this experience? I am going to be in this boat soon and wasn't sure how common this kind of error when switching over was I moved from a Intel Xeon 2430v2 (socket LGA1356) to an Athlon 200GE without needing to reinstall Windows I moved from that same Xeon to a Ryzen 2400G without needing to reinstall Windows I moved from a Ryzen 2400G to a Ryzen 3100 without needing to reinstall Windows My brother moved from an Athlon X4 860K (socket FM2+) to an Athlon 200GE without needing to reinstall Windows And later he moved from the Athlon 200GE to a Ryzen 3600 without needing to reinstall Windows so it's possible that the specific example I wrote about, which was the latest move I did, was just the specific one that was problematic, or that I just wasn't being patient enough with getting Windows to sort itself out
|
#
?
May 19, 2021 18:01
|
|
|
|
| # ? May 15, 2024 20:10 |
|
|
I threw together my build last night. Easiest one ever. 3600x into a MSI B550 board, booted it, flashed the bios, installed windows and it was a breeze withought a single problem. I'm amazed by this, but then again, I'm an old and still have ptsd from windows 98 installs. I also forgot that the 3600x comes with a fan/cooler and I lazily just slapped it on with the existing thermal pad and everything is fine. Not to worry though, I'll soon be poking around the mobo with sharp metal things, applying thermal toothpaste (pepsodent is the best) to the I/O ports, etc. and will manage to gently caress it all up in an insane quest to get it to "run more better."
|
|
#
?
May 19, 2021 18:11
|
|
|
Went from Coffee Lake to Zen3 with no probs. It booted straight in during the test boot cause the keyboard didn’t recognize in time, lol. Guess that’s a good sign!
|
|
#
?
May 19, 2021 18:17
|
|
|
Daeus posted:Are many folks having this experience? I am going to be in this boat soon and wasn't sure how common this kind of error when switching over was The trick is change the hard drive controller in device manager to the ACHI (generic driver) then swap the drive. It should boot up.
|
|
#
?
May 19, 2021 19:37
|
|
|
AutismVaccine posted:Do you have multiple programs running which try to read the sensor data? Maybe close everything in the background before starting HWMonitor. No - nothing else. I found this on reddit, I wonder if this is true: "some reddit guy" posted:I believe this is a CPU specific problem rather than BIOS version or motherboard. Apparently some of these 5600X/5800X have two CCD's instead of one, and that other CCD is permanently in deep sleep mode. So basically your 5600X/5800X is a failed 5950X, and ryzen master is very likely looking at the wrong CCD since that CCD is still sucking in some power.
|
|
#
?
May 20, 2021 09:48
|
|
|
Shemp the Stooge posted:No - nothing else. Hmm. I suppose that makes sense, but if the core complex is bad, wouldn’t they have binned it out before mounting it to the package? Or maybe it binned fine and then failed QC after and that’s why it is turned off and the chip downgraded from a 59xx? You’d have to delid the CPU to find out for sure
|
|
#
?
May 20, 2021 20:06
|
|
|
Shemp the Stooge posted:I updated the bios on my 5800x and msi X570 based board to try and fix the usb issues. The update went fine but now Ryzen master shows my temp as 0C and PPT as 0W.
|
|
#
?
May 21, 2021 00:15
|
|
|
Malloc Voidstar posted:Are you using the latest Ryzen Master? I had to update mine after a BIOS update. yep, 2.6.2.1818
|
|
#
?
May 21, 2021 00:36
|
|
|
I've got two 16gb sticks of 3600 RAM, for a total of 36gb in my dual channel system (3600x+B550), running at 3600 via the XMP settings. I happen to have a third 16gb stick of the same manufacturer/brand/model. Question: If I put this third stick in, will performance suffer in terms of timings? I don't need the extra memory, but it would just be sitting around otherwise, so why not. I guess I could find out, but I'd like to know if 2 sticks in one channel and only 1 in the other is really, really dumb right off the bat.
|
|
#
?
May 21, 2021 01:00
|
|
|
If it's the same spec stick, then I think the only drawback is that the 3rd stick will run in single channel mode, so the first 32 GB will be dual channel, then any data stored in the remaining 16 GB is single channel performance. Depending on your use case, it may or may not matter. FWIW, on Intel systems, I believe I read it's typically a 3-5% performance hit on average, but obviously results will vary.
|
|
#
?
May 21, 2021 01:28
|
|
|
Ok, cool, thanks. As I understand you, I'll suffer a comparative performance hit of 5% on the data that makes it to that one stick. Sounds good to me -- I'll give a try.
|
|
#
?
May 21, 2021 02:01
|
|
|
It'll probably be worse than that, because AMD's memory controller is not nearly as good as Intel's. It's likely that your frequency and timings will have to regress. Two sticks is always the ideal if possible, especially in AMD systems where it's much harder to get optimal memory performance.
|
|
#
?
May 21, 2021 02:14
|
|
|
K8.0 posted:It'll probably be worse than that, because AMD's memory controller is not nearly as good as Intel's. It's likely that your frequency and timings will have to regress. Two sticks is always the ideal if possible, especially in AMD systems where it's much harder to get optimal memory performance. Again, it's all situational dependent: https://www.reddit.com/r/XMG_gg/comments/iv8b7q/psa_performance_impact_from_asymmetric_dual/ Gaming can take a noticeable hit, but otherwise "it depends".
|
|
#
?
May 21, 2021 02:53
|
|
|
Outside of some weird outlier circumstance I would just run the 2x16GB. RAM should always be matched and paired, and then the arguments about 2 vs 4 sticks become a different animal. I went with 4x8GB personally, but others would absolutely go for 2x16GB. But 3x16GB is weird and if you're already at 32GB total seems like it would almost certainly hurt performance while providing essentially no benefit. Unless you're doing something where 32GB is somehow not enough memory, which seems like an edge case.
|
|
#
?
May 21, 2021 03:07
|
|
|
SourKraut posted:Again, it's all situational dependent: A laptop with a Zen2 APU and memory at extremely slow clocks and timings is not a good analog for a Zen3 CPU successfully running 3600 XMP. At 2666 CL18/19 the memory controller is chilling and performance is in the garbage to begin with. At 3600 and hopefully decent timings you are often pushing the limits of what a given AMD IMC can manage and the signal degradation that comes from more than one stick per channel frequently will cause failures, resulting in either instability or probably more likely lower clocks/higher latency on the memory. If he was doing work with a 40gb data set and had to deal with it, sure, do what you have to do, but it's not ideal and in a "why not" scenario, this is the why not. No harm in playing around with it, but don't leave it like that and make sure your memory is running properly when you're done.
|
|
#
?
May 21, 2021 03:30
|
|
|
SourKraut posted:Again, it's all situational dependent: Ah, thank you for this and for the others chiming in. I'll just stay with two sticks. It's one thing to take a performance hit on that individual stick....it's quite another if it drags down performance in general. I realize I will never notice any of this, of course. Why is my system 21 nanoseconds slower today? edit: and I'm not trying to simulate a black hole or anything, just thought I'd rather use the extra stick as long as it didn't hurt anything. it's no biggie. thanks again. AARP LARPer fucked around with this message at 03:38 on May 21, 2021 |
|
#
?
May 21, 2021 03:31
|
|
|
Speaking of ram, I have 4000mhz ram on Zen3 (it was a stock thing, I really needed new ram and that was all they had in the amount I needed.) Should I be underclocking it to 3800mhz or 3600mhz on a 5900x or something? Is it even gonna matter that much?
|
|
#
?
May 21, 2021 05:06
|
|
|
I would try the RAM at 3800MHz with the FCLK at 1900MHz, if it can't do that and you have to go to 3600MHz and 1800MHz FCLK you're not losing much.
|
|
#
?
May 21, 2021 05:11
|
|
|
BrainDance posted:Speaking of ram, I have 4000mhz ram on Zen3 (it was a stock thing, I really needed new ram and that was all they had in the amount I needed.) I personally hit a hard-stop wall on Zen 3 running >=3800:1900IF. Best I can do is 3733/1867. I would not count on it running at 4000 mhz as it says it will. I mean I guess try it, but odds are you probably can't run it that high without either extremely good motherboard + cpu + ram sillicon lottery and a lot of tweaking of voltages you'll probably have to bring it down to 3733 but maybe you'll have some luck at 3800
|
|
#
?
May 21, 2021 05:19
|
|
|
With my 5600x and Asrock B550M Pro4, I got 3200 CL14 b-die stable at 3800 and 1900 FCLK at CL16. It would pass 6 rounds of Testmem5 with the anta777 extreme test, but generate thousands of WHEA errors when doing anything processor intensive or running the OCCT large AVX2 test. I am convinced AGESA 1.2.0.2 is flawed, but I didn’t want errors. I dropped it down to 3733 / 1866 and tightened to CL15 and get basically the same AIDA scores but without the WHEA errors. Edit: https://github.com/integralfx/MemTestHelper/blob/master/DDR4%20OC%20Guide.md is a great guide if you can be home to mess with BIOS settings while testing timings. bobfather fucked around with this message at 12:23 on May 21, 2021 |
|
#
?
May 21, 2021 11:54
|
|
|
I thought you were supposed to match the IF and memory speed, ya'll are indicating that faster IF is better, period. Little help?
|
|
#
?
May 21, 2021 12:57
|
|
|
LRADIKAL posted:I thought you were supposed to match the IF and memory speed, ya'll are indicating that faster IF is better, period. Little help? you're supposed to match HALF of the memory speed with the IF [3800 / 2] = 1900 [3733 / 2] = 1866 and so on
|
|
#
?
May 21, 2021 13:06
|
|
|
It's DDR, I stated the question that way intentionally.
|
|
#
?
May 21, 2021 15:21
|
|
|
LRADIKAL posted:It's DDR, I stated the question that way intentionally. Memory speed is usually reported in transfers/s, FCLK is always going to be cycles/s. Same for MEMCLK and UCLK as they have CLK in the same. 1:1:1 is the best for those clocks. The only reason to try a disconnected faster FCLK is to see what the absolute maximum FCLK your CPU can handle independent of the memory. That way you can tell if the cpu is holding you back or if it’s the memory.
|
|
#
?
May 21, 2021 15:43
|
|
|
I missed the announcement earlier this year, but I read today where Jim Keller ended up: https://betakit.com/hardware-startup-tenstorrent-raises-over-200-million-in-funding-claims-1-billion-valuation/
|
|
#
?
May 21, 2021 16:14
|
|
|
Subjunctive posted:I missed the announcement earlier this year, but I read today where Jim Keller ended up: Amazing that this saga started with a Canadian professor taking a trip to Russia and ends with Keller spearheading a new startup.
|
|
#
?
May 21, 2021 18:20
|
|
|
Penisaurus Sex posted:Amazing that this saga started with a Canadian professor taking a trip to Russia and ends with Keller spearheading a new startup. A Canadian professor and alt-right cult figurehead taking a trip to Russia to coma-detox from his crippling benzo addiction and red meat only diet. There's a LOT to unpack.
|
|
#
?
May 21, 2021 19:15
|
|
|
Jim Keller is such an interesting figure to me. A relatively low profile engineer that just bounces around the best tech companies in the world on a whim, distributing groundbreaking architectures whever he goes. He should be more famous than his grifter hack of a brother in law. Is there a good write-up on his career anywhere? I'd love to see how influencial he really was in each position.
|
|
#
?
May 22, 2021 06:48
|
|
|
https://videocardz.com/newz/amd-next-gen-am5-platform-to-feature-lga1718-socket Apparently AMD is moving to an LGA (pins on the motherboard) socket for AM5
|
|
#
?
May 23, 2021 05:04
|
|
|
BurritoJustice posted:Jim Keller is such an interesting figure to me. A relatively low profile engineer that just bounces around the best tech companies in the world on a whim, distributing groundbreaking architectures whever he goes. He should be more famous than his grifter hack of a brother in law. 8 out of the 10 great CPU architectures all had jim touching it, he is the greatest CPU designer on the planet.
|
|
#
?
May 23, 2021 06:58
|
|
|
What's the list?
|
|
#
?
May 23, 2021 07:01
|
|
|
gradenko_2000 posted:https://videocardz.com/newz/amd-next-gen-am5-platform-to-feature-lga1718-socket No PCI-e 5.0 for desktops - not surprising really, but so much for the dreams of running 3.5GB/s SSDs off single PCIe lanes.
|
|
#
?
May 23, 2021 18:53
|
|
|
ConanTheLibrarian posted:No PCI-e 5.0 for desktops - not surprising really, but so much for the dreams of running 3.5GB/s SSDs off single PCIe lanes. I wonder if there will be a refresh Zen4+ that will add it, because Alder Lake on Intel supposedly will have it. Also I want to see motherboard manufacturers + case vendors join forces for EDSFF E1.S backplanes + drive bays and do the single lane per drive thing too!
|
|
#
?
May 23, 2021 19:03
|
|
|
Genoa (Zen4 servers) will have a PCIe 5 I/O die, so it’s def physically possible for Zen4 to support it. There isn’t much of a use case for PCIe 5 on the desktop quite yet, and supporting it will add cost to the boards I imagine, so it’s probably just a timing thing.
|
|
#
?
May 23, 2021 19:09
|
|
|
Daeus posted:Are many folks having this experience? I am going to be in this boat soon and wasn't sure how common this kind of error when switching over was i refuse to ever reinstall so I always take the time to fix it. It's usually as "trivial" as going into recovery, hitting command prompt and using the bootrec/rebuildbcd tools. The one that took me the longest was when the vendor included their own special snowflake nonstandard boot location in UEFI that was getting found before the standard microsoft one. One repaving of the UEFI partition later and wouldn't you know it, that nonstandard vendor BCD wasn't getting picked up anymore.
|
|
#
?
May 23, 2021 20:26
|
|
|
https://twitter.com/ExecuFix/status/1397173117487828992 I've never heard of this particular person before in the circle of known "leakers" so maybe they're blowing smoke up our rear end but they keep dropping info about AM5
|
|
#
?
May 25, 2021 14:27
|
|
|
FuturePastNow posted:AMD will always have lovely chipsets and building them into the processor won't change that. They are pretty ok now! Im picky af and been happy with the X570.
|
|
#
?
May 25, 2021 15:10
|
|
|
gradenko_2000 posted:https://twitter.com/ExecuFix/status/1397173117487828992 Up to 170W TDP? Okay, I guess we don't care about TDP anymore now. I'm building a 5000-series machine with the best parts that I can get and sitting on that for a while. 170W TDP, jesus christ.
|
|
#
?
May 25, 2021 15:52
|
|
|
Maybe if they manage to squeeze in a 32 core CPU on AM5, otherwise 170W TPD seems kind of out there.
|
|
#
?
May 25, 2021 16:50
|
|
|
|
| # ? May 15, 2024 20:10 |
|
|
redeyes posted:They are pretty ok now! Im picky af and been happy with the X570. They are good now! Me too!
|
|
#
?
May 25, 2021 17:08
|
|
