Sudden death of AMD Ryzen Zen 5 CPUs

SDK Chan said:
I hope I don't offend anyone, but the main advantage of the 3D-VCache is for me to compensate the poor performance of crappy "Japanese" made game engines.
Japanese games aren't really optimized for multi-core performance so, single-core performance is more important.
And if comparing the R9 9900X and R9 9900X3D in japanese games, then I can say that due to the 3D-VCache the R9 9900X3D delivers almost 20% more single-core performance than the R9 9900X in those games...
The more it is important while emulating games where you have every core for performing only one task.
Click to expand...
Its not only Japanese games ... its majority of games.
Only handful of games can utilize multicores. So 3D-VCache is acceptable for those hardcore gamers, and single core max speed still valid to this day not only in games but also with LLM`s - as with higher single core threaded cpu you can get higher token/second running interference.
Also, latest R9 9950X3D is a beast in productivity too ( as seen in benchmarks ) shame pricing is a bit Oooof.
 
You get the best from both worlds with X3D CPUs. :)
Yes, the majority of games use single-core, too, but japanese games (many of them) tend to be bad optimized.
Square Enix has a very good reputation of releasing very bad optimized games on PC.
Monster Hunter is bad optimized, but Octopath Traveler II is very good optimized.
But generally I don't notice so much of a performance gain in native PC games.

All X3D CPUs are not cheap, and the hype is partly the cause of it, I guess ?
The price is one thing, but if the use case of that CPU is justified, then the investment is also worth it, I think. ;)
 
SDK Chan said:
I don't think that the base clock is the max OCed clock of it, since it can clock up to 5.5GHz.
Click to expand...
For both Intel and AMD it still seems that they respect the traditional naming convention, where the "base clock" is the maximum clock speed without turbo, and with turbo they call it "turbo clock".
I rarely use turbo clocks because they are simply not sustainable, especially on my Intel dumpster fire. My laptop can sustain 70 degrees under full load at max base clock, but if I start slightly increasing the clock speed to above that it quickly reaches 80 and eventually even 90 degrees. Full turbo clocks under full load throttle the system because it surpasses 100 degrees, never again.
 
cracauer@ said:
I don't think photo editing is very memory bandwidth intensive. It loads the CPU pretty hard.
Click to expand...
It's pretty flexible
1920x1080 8bit/channel? Very low memory usage even for large and complex projects
4K 32bit/channel? Memory usage will be notably higher, but even then it's nothing that 16GB of system RAM can't do
 
Space Banana said:
For both Intel and AMD it still seems that they respect the traditional naming convention, where the "base clock" is the maximum clock speed without turbo, and with turbo they call it "turbo clock".
Click to expand...
I can't stop recalling "Turbo" as "normal speed" and with "Turbo" off as "mostly same speed as slooooow PC/AT".😅
 
All I can add is some anecdotal evidence. With an ASUS Rog Strix X670E-A board:
  1. AMD 7950X: fried after one year of use, including black burn marks on the mainboard. Mainboard exchanged under warranty by ASUS, CPU exchanged by AMD for a 7950X3D as they were out of stock on the non-X3D
  2. AMD 7950X3D (the one received as an exchange from AMD), dead after three months of use (CPU Status LED on mainboard remains on, no boot). Mainboard ok, CPU exchanged again by AMD.
Since these were different computer builds, all other components were different. Power supplies were ample (1000W) and brand name (EVGA, Seasonic) and so was the cooling. Also, these CPUs ran in AMD ECO mode (i.e. undervolted).
 
recluce said:
Since these were different computer builds, all other components were different. Power supplies were ample (1000W) and brand name (EVGA, Seasonic) and so was the cooling. Also, these CPUs ran in AMD ECO mode (i.e. undervolted).
Click to expand...
I never had burnt CPUs, but I read on the internet that bad PSUs are often the source of fault.
For almost 4 years now I have a EVGA 2000W gold labeled PSU running in ECO mode and everything is fine.

As for ASUS it seems that some users on reddit report problems with that board brand, too.
Generally MSI never let me down regarding GPUs, and motherboards.

Until now I only had one single problem with the new x870E MSI MPG board, and it only occurs when I try to reboot from Windows to FreeBSD.
The problem is a long boot which takes around 5 minutes, but if I only boot FreeBSD without changing to Windows the port responds very quickly from a coldboot/reboot.
 
An amount of electromigration of dopant atoms within the semiconductor crystal lattice sufficient to adversely affect chip performance seems unlikely after only a few months. It would seem reasonable to assume that the manufacturers (tsmc? global foundaries? ) have modelled chip aging processes of that type and can predict a reasonable service life, 4-8 months seems far too short to me. I suppose it could be the case, and that they know about it and are keeping quiet in the hope that most of them will keep working. A recall and swap of large numbers of chips might well be extremely expensive.

Other electrical/mechanical/thermal factors on the board might be relevent. Degradation of transmission line characteristics of pcb tracks caused by stray capacitance, dust buildup, ageing. Perhaps just plain poor pcb design. Capacitors aging, especially electrolytics, I can still remember years ago gigabyte had a big problem with low quality electrolytics used on their montherboards. The VRM module aging; if it is socketed, can you replace it?. Perhaps mechanical wear of the DIMM edge connectors, often the gold plating is a very thin 'flash plating' nowadays, much thinner than it used to be. Wiping with isopropyl alcohol might help. Heat sink compound drying out and becoming less thermally conductive. Or an incorrect application of heatsink compound in the factory. Loss of regulation in the system power supply due to psu components being faulty or aging. Fake components getting into the supply chain when the board was made (fake components are a major problem in production electronics). Again 4-8 months seems a rather short time for any of these kinds of problems to appear. Unless the board was marginal when they shipped it of course. At the end of the day its a bunch of physical electronics devices in a box, any number of things can go wrong.

Nonetheless all these types of faults usually have a rather low probability of occurrence with modern automated high-density SMD manufacturing techniques.

I guess you must have tried all the easy things - re-seat dimms and connectors, wipe edge connectors with alcohol, replace heatsink compound, vacuum the mobo clean of dust and dirt, try a different power supply, etc etc.
 
You must log in or register to reply here.
Back
Top