|
Fix it yourself, how hard could it be?
|
# ? Dec 20, 2021 22:01 |
|
|
# ? Jun 6, 2024 21:33 |
|
Dr. Video Games 0031 posted:Or not, because he's working through a massive backlog and has stopped accepting GPUs. Oh. Sorry, Had no idea.
|
# ? Dec 21, 2021 00:19 |
|
Klyith posted:Articles about it: Well, I did apply the latest bios and AGESA crap. Didn't work. Just added in a pcie USB card and life is good. No cut outs, no issues. Weird!
|
# ? Dec 21, 2021 00:22 |
|
My Asus 570 Mobo with 3700x has USB make disconnect reconnect sounds occasionally. Nothing ever stops working though, and after several Mobo updates and windows 11 it seems isolated to my new mechanical keyboard I spilled wine on and fixed by putting it in the dishwasher and then letting dry for 2 days.
|
# ? Dec 21, 2021 00:46 |
|
I had issues with my ASrock gaming phantom board with USB issues, keyboard and mouse would randomly disconnect for 2-3 seconds at a time. Installing the latest BIOS did nothing but what ended up fixing the issue for me was going into the BIOS and manually changing the PCIe speed from PCIe4 to PCIe3. Haven’t had a single problem since.
|
# ? Dec 22, 2021 23:16 |
|
https://twitter.com/anandtech/status/1474156169094701065?t=ZQa-Y6O-28hd7xIBqjs_5Q&s=19
|
# ? Dec 24, 2021 04:23 |
|
gradenko_2000 posted:https://twitter.com/anandtech/status/1474156169094701065?t=ZQa-Y6O-28hd7xIBqjs_5Q&s=19 this isn't a bad thing while i want the I/O die to move to a better node this is fine for now, problem is Apple won't let anyone buy 5nm or 3nm tsmc.
|
# ? Dec 24, 2021 04:27 |
|
AMD spent so many years trying to get out from under the wafer agreement, and now they need to expand/extend it. Oh how times have changed. Also this is probably my fav part: quote:First Amendment to the Amended and Restated Seventh Amendment to the Wafer Supply Agreement
|
# ? Dec 24, 2021 04:40 |
|
Cygni posted:AMD spent so many years trying to get out from under the wafer agreement, and now they need to expand/extend it. Oh how times have changed. The issue with the WSA before is it was exclusive, they could only ever buy Gloflo, now they can buy gloflo and others.
|
# ? Dec 24, 2021 04:51 |
|
I asked in the Tech Support thread, but this is likely a better place at this point. I have a 3600X running at stock that I've been stress testing. To make a very long story short -- it fails exactly one stress test. Prime95 Small FFT while the CPU is thermally throttled. Always the same core, only when thermally throttled. Other AVX intensive tests seem rock solid even if the CPU is throttling. I've tried many other fpu-intensive stress tests like ibt and x264 stability test and other Prime95 tests. Most can't get it to thermally throttle. Even those that do are rock solid stable with no failures reported. Is this somewhat to be expected, as in, running at stock (with "Automatic Overclocking" in the ASRock bios which I assume just use AMD's built in precision boost overclock) thermal throttling will introduce failures, or even if thermally throttled should it be able to pass a power vampire FMA/AVX test at stock speed? I'm seriously considering RMA'ing it just because it is going to be an absolute pain to resell someday with this defect. Also tempted to upgrade from a $25 BeCool Pure Rock Slim to a $60 Scythe Fuma2 which I strongly suspect will stop the thermal throttling (5/6 cores 10/12 threads runs around 85-90 degrees in the SmallFFT test, with 95c being the point where it starts to throttle), just to satisfy my OCD even though it will never hit these conditions in real-world use. But I don't know if this is considered normal for modern processors with built-in "turbo boosting" and is actually expected to fail under these conditions. Any opinions? Chuu fucked around with this message at 07:09 on Dec 24, 2021 |
# ? Dec 24, 2021 07:00 |
|
Either return it, or quit intentionally crashing it.
|
# ? Dec 24, 2021 07:35 |
|
If it's perfectly stable while not at extremely high temperatures, then I don't see an issue there. Don't let it get that hot. It's really quite bad if you're letting it happen on a regular basis. I also feel like your cooler should be able to handle a 3600X at fully stock settings, even during AVX loads. Check to make sure that it's mounted properly with good thermal paste spread and adequate mounting pressure.
|
# ? Dec 24, 2021 07:52 |
|
Dr. Video Games 0031 posted:If it's perfectly stable while not at extremely high temperatures, then I don't see an issue there. Don't let it get that hot. It's really quite bad if you're letting it happen on a regular basis. That's a good point - the Pure Rock Slim is rated to 130W, so should have some headroom. I have a Dark Rock Slim which is rated to 180W because I wanted the extra headroom for upgrades (and I wanted a black cooler).
|
# ? Dec 24, 2021 10:11 |
|
Chuu posted:Is this somewhat to be expected, as in, running at stock (with "Automatic Overclocking" in the ASRock bios which I assume just use AMD's built in precision boost overclock) thermal throttling will introduce failures, or even if thermally throttled should it be able to pass a power vampire FMA/AVX test at stock speed? I presume that with automatic overclocking you mean "Overclocking Mode" set to "Auto"? That shouldn't do any overclocking so unless you have PBO enabled, it's just the stock boost algorithm afaik.
|
# ? Dec 24, 2021 11:27 |
|
Klyith posted:Articles about it: Combination of factors. The chipset updates did resolve the worst stability issues I had, and make my g2 more usable. Then HP admitted that in the specific case of the g2, the psi was also underpowered, compounding this problem. They just sent me a new cable on a fatter psu, and part of me thinks I could probably go back to the earliest zen9 driver for my board and stuff might still work, but not spending the time to find out. some index owners reported the same stuff, though.
|
# ? Dec 24, 2021 13:56 |
|
The 5800X continues to creep down in the price but the 5900X seems to be stubbornly refusing to move much. Even with the 12700K undercutting the latter by 20%. I guess people just keep on buying the 5900X.
|
# ? Jan 3, 2022 14:10 |
|
In the UK the 5600x even went up in price, which is the exact opposite of what I was expecting I got it off ocuk for £240 (as part of a bundle) and now it's £320
|
# ? Jan 3, 2022 14:18 |
|
AMD CES keynote stream for tomorrow: https://www.youtube.com/watch?v=_jX-hKvUQDU
|
# ? Jan 3, 2022 18:02 |
|
Be careful when buying AMD based OEM boxes second hand. Lenovo/AMD are rolling out their "Vendor Locking" trash to all AMD Pro based systems, including Threadripper and Ryzen Pro. https://www.servethehome.com/lenovo-vendor-locking-ryzen-based-systems-with-amd-psb/
|
# ? Jan 3, 2022 20:24 |
|
Cygni posted:Be careful when buying AMD based OEM boxes second hand. Lenovo/AMD are rolling out their "Vendor Locking" trash to all AMD Pro based systems, including Threadripper and Ryzen Pro. Holy poo poo this is terrible behavior. I hope the vendor keys that are used to lock the processors are leaked by a spiteful employee.
|
# ? Jan 3, 2022 23:56 |
|
Yeah, the part where it will then ask to perma-lock any future Ryzen CPUs you stick in there is just nasty. Feel bad for anyone who doesn't know what they are doing, and its going to make buying 2nd hand Ryzen CPUs a chore. Good thing i stick to buying outdated Bulldozer dogshit!!
|
# ? Jan 4, 2022 00:46 |
|
Truly the tables have turned -- some Intel like behavior here.
|
# ? Jan 4, 2022 01:02 |
|
movax posted:Truly the tables have turned -- some Intel like behavior here. Eh, it's a legitimate design on AMD's part. The point of having this is that once it is turned on, it is very very hard for an adversary to do anything malicious at the firmware level. The firmware itself need a cryptographic signature, with the key that hopefully Lenovo / Dell / whoever is keeping safe. That means it kinda has to be irrevocable in the hardware, versus something like the TPM module that the bios can reset. The alternative would be for AMD to have a preset key in non-writable hardware, and for AMD to control the key. You don't have ewaste CPUs that are forever locked to one platform, but it's a single point of failure and slows updates for everyone. When MSI wants to send out a BIOS update, they'd have to first send it to AMD to get signed. MSI's BIOS updates are slow enough already thank you. In many ways this dynamic blown-fuses thing is a much nicer & more open method. It doesn't lock out new entrants to OEM markets. And some organization that's sufficiently large & paranoid, say the DoD, could even work with their vendors to use a different key that they kept themselves. So instead of an attacker wanting to compromise Dell's systems to nab the key, they'd have to go Mission Impossible the NSA or whatever. tl;dr AMD make a gun intended for self-defense, Lenovo and Dell picked up the gun and thought "I can mug my customers!" I'm not sure what AMD could do, other than demand that their partners never use this feature like that and always leave it something that the end-user has to turn on themselves? Klyith fucked around with this message at 02:48 on Jan 4, 2022 |
# ? Jan 4, 2022 02:46 |
|
I don't think there's a good argument that the CPU has to be locked to the machine. Sign the CPU to the machine so that motherboard and that CPU will only boot together in super trusted mode, sure, but what's the benefit of stopping that CPU from working in another motherboard?
|
# ? Jan 4, 2022 02:50 |
|
Klyith posted:Eh, it's a legitimate design on AMD's part. The point of having this is that once it is turned on, it is very very hard for an adversary to do anything malicious at the firmware level. The firmware itself need a cryptographic signature, with the key that hopefully Lenovo / Dell / whoever is keeping safe. That means it kinda has to be irrevocable in the hardware, versus something like the TPM module that the bios can reset. It's true, the below was what I was really thinking as you said: quote:tl;dr AMD make a gun intended for self-defense, Lenovo and Dell picked up the gun and thought "I can mug my customers!" I am chuckling a bit now at the idea of something like this on GPUs shipped in prebuilts, something like old DVD player region codes. "OK, you get 3 machines you get to move this GPU too, and then no more!"
|
# ? Jan 4, 2022 02:56 |
|
K8.0 posted:I don't think there's a good argument that the CPU has to be locked to the machine. If I've understood things correctly, an affected CPU will work on other motherboards with the same key signature. Presumably that means "any compatible Lenovo system", though I guess there's nothing stopping Lenovo from using multiple keys to make things even worse! It's effectively a vendor-lock, not a machine-lock. K8.0 posted:Sign the CPU to the machine so that motherboard and that CPU will only boot together in super trusted mode, sure, but what's the benefit of stopping that CPU from working in another motherboard? Because if you allow a CPU to boot in non-super-trust mode, what stops the firmware attacker from silently using that pathway when they compromise the system? Enforcement would have to be done by the OS, with some method to directly ask the CPU if it really was in super-secure mode during boot. But then your super-sophisticated firmware attack might have a way to gently caress with the OS before it even loads. I can see the line of thought where the only safety from a sophisticated firmware attack is for the entire machine to turn into a pumpkin rather than boot.
|
# ? Jan 4, 2022 03:50 |
|
Klyith posted:I can see the line of thought where the only safety from a sophisticated firmware attack is for the entire machine to turn into a pumpkin rather than boot. This usually what secure boot means in embedded land.
|
# ? Jan 4, 2022 03:54 |
|
Klyith posted:Because if you allow a CPU to boot in non-super-trust mode, what stops the firmware attacker from silently using that pathway when they compromise the system? Enforcement would have to be done by the OS, with some method to directly ask the CPU if it really was in super-secure mode during boot. You don't allow a signed CPU to ever boot in non-super-trust mode, if it's in a super-trust motherboard. You make the CPU identification of standard vs secure motherboard a hardware feature, not in anyway subject to firmware. How does that have any relevance to motherboards that don't have the super trust feature? K8.0 fucked around with this message at 07:44 on Jan 4, 2022 |
# ? Jan 4, 2022 07:41 |
|
Remember to order a CPU ASAP if you want it. Someone developed a new "raptoreum" coin which mines with CPU and benefits from large L3 cache. You can imagine what will happen to the 3D cache ryzen availability. I'd imagine the 5900x equivalent 3d cache version being $1000-$1500 if the raptoreum coin takes off. You can earn $4 already excluding electricity costs today with current L3 cache models.
|
# ? Jan 4, 2022 09:29 |
|
Ihmemies posted:Remember to order a CPU ASAP if you want it. Someone developed a new "raptoreum" coin which mines with CPU and benefits from large L3 cache. You can imagine what will happen to the 3D cache ryzen availability. I'd imagine the 5900x equivalent 3d cache version being $1000-$1500 if the raptoreum coin takes off. You can earn $4 already excluding electricity costs today with current L3 cache models. $4 per what? The existence of a coin isn't a death knell for a hardware product. There are a million different crypto coins that are mined using all kinds of different poo poo. Chia came and went pretty quickly. Is there any sign that raptoreum will actually drive up demand and create a scarcity situation?
|
# ? Jan 4, 2022 09:32 |
|
Spring? Boo.
|
# ? Jan 4, 2022 11:38 |
|
Just the 8-core one with augmented cache? Bleh. Also, H2 2022 is probably minute supply around December. --edit: Zen 4 is probably gonna be expensive as gently caress, if they're launching both of those that close together. Because within that timeframe, former has barely reason to exist. Combat Pretzel fucked around with this message at 12:39 on Jan 4, 2022 |
# ? Jan 4, 2022 12:35 |
|
Either they only have capacity to do one chip and they've decided to focus on the 5800X as most beneficial to profit (it's the unloved one that has seen the biggest price drop, and if the cache is most useful to gamers, they think the 5800X is the best price point) OR: the focus on gamers on the slide seems contradict the 'leadership in every segment' message and there's other slides for other segments which will have other 3D chips.
|
# ? Jan 4, 2022 12:49 |
|
Combat Pretzel posted:Just the 8-core one with augmented cache? Bleh. 3D versions of the 5000 series should be popular enough for people on AM4 looking to tide over the DDR5 shortage and the GPU market being hosed. I'll probably replace my 2600 unless the pricing is bad or they cut off B450 boards.
|
# ? Jan 4, 2022 14:14 |
|
I wonder if DDR5 is going to still be expensive enough in 2022 that AM5 is going to support DDR4
|
# ? Jan 4, 2022 14:22 |
|
Ihmemies posted:Remember to order a CPU ASAP if you want it. Someone developed a new "raptoreum" coin which mines with CPU and benefits from large L3 cache. You can imagine what will happen to the 3D cache ryzen availability. I'd imagine the 5900x equivalent 3d cache version being $1000-$1500 if the raptoreum coin takes off. You can earn $4 already excluding electricity costs today with current L3 cache models. do the people who developed raptoreum have the money or influence to pump their shitcoin into FOMO "takes off" territory? also a minor google search suggests raptoreum isn't new and has been around for a year or two. how many people are mining it now? how many new CPUs added to the mining pool would it take to drive that $4/day to $0? I'm asking because the counter-example is that chia was very prominent at launch, and only caused a month or so of disruption to SSD availability and prices. So anyone who ran out and bought a SSD ahead of chia because they thought "ugh SSDs will be as hard to get as GPUs" ended up mildly screwing themselves. Chia got hyped in like May, SSD prices barely budged, and then late summer through now have been steadily declining. Zero long-term impact. tl;dr take all crypto "big news" with a grain of salt, particularly anything dealing with alt-coin crap
|
# ? Jan 4, 2022 14:59 |
|
Keynote starts in five.
|
# ? Jan 4, 2022 15:54 |
|
Link here: https://www.youtube.com/watch?v=_jX-hKvUQDU
|
# ? Jan 4, 2022 16:01 |
|
Klyith posted:do the people who developed raptoreum have the money or influence to pump their shitcoin into FOMO "takes off" territory? also a minor google search suggests raptoreum isn't new and has been around for a year or two. how many people are mining it now? how many new CPUs added to the mining pool would it take to drive that $4/day to $0?
|
# ? Jan 4, 2022 16:04 |
|
|
# ? Jun 6, 2024 21:33 |
|
Shipon posted:i don't know if it's because of chia crashing but picking up 2 TB samsung 970 evos for $200 each during black friday was pretty great I think that's due to NAND flash oversupply (caused by early COVID shortages).
|
# ? Jan 4, 2022 16:08 |