- Joined
- Dec 18, 2000
I'd love to be the attorney litigating a case based on that. They provide a tool for AMD customers to use, but if you use it, your warranty is null and void. Really AMD? Seriously? Give me a fricken' break!
-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
FEATURED Official AMD Ryzen 3 Vermeer (4***/5***) Rumors and Discussion Thread
- Thread starter Janus67
- Start date
That's how it's always been, yes (Read: overclocking from the BIOS)? The only difference is going through the BIOS there's the honor system (you O/C it, you break it while OC... it broke the terms of the warranty, you eat it). Whereas with a Windows software component, they can get you because you accepted the terms and may have a record of it. If not, it's still the same honor system as neither company will warranty overclocking (Intel will with an additional fee/plan... I don't think AMD offers this?).I'd love to be the attorney litigating a case based on that. They provide a tool for AMD customers to use, but if you use it, your warranty is null and void. Really AMD? Seriously? Give me a fricken' break!
Stability is not a scalable measurement. It's either stable, or it is not.
- Joined
- Mar 7, 2008
I've previously taken the position you can't prove stability. You're only stable until you're not.
In the context of overclocking, there are different levels of stability:
Can't boot
Boots at all, but can't do anything useful with it before crashing
Stable enough to load OS
Stable enough to run benchmarks once, a.k.a. bench stable
Stable enough to run chosen software, such as games (typical casual OC)
Stable enough to run Prime95 (I thought Zen CPUs with power limit would negate this, but I'm fighting a stock Zen 2 system over this right now. Prime95-like workload is unstable whereas everything else appears stable. Suspect mobo power delivery scaling at fault. Reducing power limit makes it worse)
Stable enough to run anything/everything
- Joined
- May 30, 2004
That usually means you at the least, need an oh-shucks-ton of Vcore! (At least in the past!)
OTOH, if Prime95 fails within 2 hours, then you're almost there! Just need a minor Vcore bump, in my experience.
I agree there are levels of instability as you have described, and the closer to the top of that list indicates how far you are from being stable. It is also true that even stable systems will experience a random "glitch" that can lock or freeze an otherwise stable system. Setting aside the glitch outlier, a stable system will be able to run any task indefinitely so long as temperatures remain in check. Datacenters and miners do this constantly.
There is no 100% stable PC and that's why there are various technologies which are helping it to remain stable as close to the perfect condition as it's possible.
There used to be something like a number of potential calculation errors but no one is talking about it anymore. In Intel P4 era it was something between 8 and 30 possible errors per CPU and Intel was actually reporting that (some may remember when was a lot of noise after Prescott release). It doesn't mean that anyone will see it. It's more like a potential case when the hardware may crash, or not. There can be still that 0.00001% cases that will cause it to crash but I feel it's a waste of time to talk about it as it's going nowhere. For end-users stability means it will work without problems for a long time and as long as the user feels it's fine then it's stable enough.
Going back to the Ryzen 5000, I've already noticed while testing new coolers last weekend that 5900X was going down to 1.0xxV during work when I don't remember Ryzen 3000 series to drop below 1.2xxV in the same conditions. Maybe I missed something but I feel it already has a wider range of voltages during a typical work.
I'm not sure what is bringing us AGESA Patch A, B, C as it's never described. Some BIOS versions have it marked as Patch A-C, some not. Some go straight to the next number without Patch info. I wonder if motherboard manufacturers modify something already or not.
There used to be something like a number of potential calculation errors but no one is talking about it anymore. In Intel P4 era it was something between 8 and 30 possible errors per CPU and Intel was actually reporting that (some may remember when was a lot of noise after Prescott release). It doesn't mean that anyone will see it. It's more like a potential case when the hardware may crash, or not. There can be still that 0.00001% cases that will cause it to crash but I feel it's a waste of time to talk about it as it's going nowhere. For end-users stability means it will work without problems for a long time and as long as the user feels it's fine then it's stable enough.
Going back to the Ryzen 5000, I've already noticed while testing new coolers last weekend that 5900X was going down to 1.0xxV during work when I don't remember Ryzen 3000 series to drop below 1.2xxV in the same conditions. Maybe I missed something but I feel it already has a wider range of voltages during a typical work.
I'm not sure what is bringing us AGESA Patch A, B, C as it's never described. Some BIOS versions have it marked as Patch A-C, some not. Some go straight to the next number without Patch info. I wonder if motherboard manufacturers modify something already or not.
"Stable for your uses" is a line I like to drop... Because one man's stable, may not be another.
There is another thread here where the guy input some random FFT length in P95 trying to get stable... If that isn't applicable to work, why bother. There is something said for using your PC to test stability, lol.
There is another thread here where the guy input some random FFT length in P95 trying to get stable... If that isn't applicable to work, why bother. There is something said for using your PC to test stability, lol.
- Joined
- Feb 1, 2011
- Location
- Republic of Texas
Isn't the idea of stress testing for stability a pass or fail test? It seems that the various tests available have been designed to stress either various or all components to determine what will give and when. Certainly you can argue that running a PC at stock settings is differently stressing the system than OC'ing one or more components, so stress is inherent, just sayin'.
Pretty sure that walking to the mailbox would have any attached monitors showing me as stable; trying to run the O'course or triathlon would make me bluescreen. Pretty much anything in between would again show me stable as well.
Pretty sure that walking to the mailbox would have any attached monitors showing me as stable; trying to run the O'course or triathlon would make me bluescreen. Pretty much anything in between would again show me stable as well.
- Joined
- Mar 7, 2008
I'd presume that to activate the undervolting mode, you have to go into effectively the same operating mode as overclocking. There is no difference between them in that sense. You're going outside original specification. Also I'd argue that under most workloads, undervolting will NOT result in any power reduction, since the CPUs typically work at power limit in multi-core workloads unless you have already overclocked by removing that power limit. A lower voltage at same power limit implies higher clock speeds.
Leaving it here just because I can
32GB @DDR4-5300 CL18 1.55V stability test
If this thread was only about Ryzen 5000 then I wouldn't have anything interesting to post. All Ryzen 5000 that I tested couldn't boot above DDR4-3800 1:1 or above DDR4-5000 1:2 (actually only one could make more than 4800). It's like having Ryzen 3000 but with lower latency.
32GB @DDR4-5300 CL18 1.55V stability test
If this thread was only about Ryzen 5000 then I wouldn't have anything interesting to post. All Ryzen 5000 that I tested couldn't boot above DDR4-3800 1:1 or above DDR4-5000 1:2 (actually only one could make more than 4800). It's like having Ryzen 3000 but with lower latency.
This makes me laugh - https://www.techpowerup.com/275565/...mitation-intel-chips-since-haswell-support-it
There was news with SAM tested on Intel and it works pretty well. Somehow AMD technology won't work on their own chips which are not so old.
I wouldn't be surprised if Nvidia added SAM in their cards through firmware+driver update. There were rumors about it recently.
There was news with SAM tested on Intel and it works pretty well. Somehow AMD technology won't work on their own chips which are not so old.
I wouldn't be surprised if Nvidia added SAM in their cards through firmware+driver update. There were rumors about it recently.
- Joined
- Mar 7, 2008
I'm debating if I should post that on another forum and see how the hardcore AMD fanboys react to that!
So based on the above, on CPU side pretty much any still relevant Intel system could support it. AMD only support with Zen 3. GPU wise, we have support on AMD 6000 series, I'm not aware of any talk for earlier. nvidia are reported to be working on support for 30 series, but no news for older. I hope they go back to Pascal if not impossible due to hardware, but I'm not holding my breath for it.
I'm not sure if anyone has an experience like me with Ryzen 5000 but I feel like my 5900X is already degrading. When I got it, I was able to run stability tests with memory at DDR4-5000. Now I have problems to boot at this clock. There are some other situations when I had no problems and now it sometimes can't boot or pass tests. One more example, I was able to run 16/32GB modules at 1:1 up to DDR4-3733/3800. Now it has problems to run them above DDR4-3600 when IF is 1:1. It simply freezes during boot.
- Joined
- Nov 7, 2016
There were stories of people degrading the memory controller on the 3000 series wasn’t there? Push the memory above 3800 and some of the voltages could go above spec and cause damage from what I heard of some people.
It's more like something with IF/memory controller. DDR4-3600 1:1 and 1:2 up to DDR4-4800 are fine, but it was working 1-2 ratios higher with all my RAM when I got it.
It was mostly with cores that were working at manual 1.4V+ for longer.
I had 5600X and 5800X before that 5900X. 5600X had faulty IMC or something, and couldn't boot with memory at more than DDR4-3200. 5800X was a bit disappointing but it was working about the same as 5900X right now.
It was mostly with cores that were working at manual 1.4V+ for longer.
I had 5600X and 5800X before that 5900X. 5600X had faulty IMC or something, and couldn't boot with memory at more than DDR4-3200. 5800X was a bit disappointing but it was working about the same as 5900X right now.
Last edited:
OP
- Joined
- May 29, 2005
- Thread Starter
- #160
Similar threads
