4152 hours of Ryzen Long Term Testing: We have Good and Bad News
Vložit
- čas přidán 10. 06. 2024
- Support me on Patreon:
/ der8auer
---------------------------------------------------------
Save 10% on your iFixit purchase: DER8AUER10
eustore.ifixit.com/der8auer
Pro Tech Toolkit: bit.ly/2JOFD8f
---------------------------------------------------------
Find my products at Caseking:
Delid Die Mate 2: bit.ly/2Rhv4y7
Delid Die Mate X: bit.ly/2EYLwwG
Skylake-X Direct Die Frame: bit.ly/2GW6yyC
9th Gen OC Frame: bit.ly/2UVSubi
Debug-LED: bit.ly/2QVUEt0
Pretested CPUs: bit.ly/2Aqhf6y
---------------------------------------------------------
My Equipment:
USB- Microscope*: amzn.to/2Vi4dky
My Camera*: amzn.to/2BN4h2O
(*Affiliate Links: If you buy something, I will get a part of the profit)
---------------------------------------------------------
Music / Credits:
-
Outro:
Dylan Sitts feat. HDBeenDope - For The Record (Dylan Sitts Remix)
---------------------------------------------------------
Paid content in this video:
- Seasonic Spot
Samples:
- ASUS B550-E Gaming Mainboard, Corsair A500, Corsair Vengeance LPX 3200C16, Seasonic Focus PX-650
---------------------------------------------------------
Timestamps:
0:00 Intro
2:50 Ads
2:56 The Setups
3:59 First Impression
4:43 OC-Settings for the test
7:12 Benching after the test
9:39 ok, this happend
12:15 The results
16:15 Outro - Věda a technologie
This is the type of content that separates you from a lot of the other reviewers on youtube. I remember when you first did that long term test, but I assumed I would never see the results. Its just so easy for something to come up in real life that invalidates the test, or you just forget about it. Really appreciate this content!
i fully agree. this is a lot of work done for 16 mins of (amazing) content
The cat makes this content roughly 56% better.
There's no other way to get this info, you have to put in the work and spend the time. Fantastic!
This is the type of content we need. It puts some science around the typical misinformation that gets spread around the internet. Here's how it works: one person in reddit or forum gets a bad CPU that should've been RMA'd in the first place, then everyone else starts freaking out, as apparently, this one bad sample becomes the general truth. Zero methodology, one sample out of a million. No one cares about everyone else with successful experiences, the bad CPU is the one to look at-- as if RMA didn't exist. I commend you on this research, calling it just CZcams content, with the amount of low effort videos we have nowadays, should almost be offensive.
salve concordo com tu
I think you're correct. I had a launch day 3600X that was a horrible piece of silicon, it wasn't worth an all-core OC, and PBO made it worse than stock. I ran that CPU on stock settings for about 9 months, then it started BSOD'ing randomly once every few days. That eventually became once every few hours, before I RMA'd it. The 3600X I got back from RMA was a completely different chip, boosted over 150MHz higher and ran cooler to boot. It also didn't BSOD which was nice. I suspect if I had been running an all-core OC, I likely would've suspected it was degradation, when it was just a faulty chip.
It's not just bad CPUs - sometimes people flash newer version of BIOS which has tweaked LLC settings / disabled stretching or some other changes (RAM related for instance) that make the system unstable (or it wasn't stable in the first place, it just worked because clocks were stretched, for example) and they go to Reddit and start to tell everyone how their CPU degraded
@sbcontt YT I don't know dude but I live in Indonesia, and I have yet to see such problem with cpu RMA's in one of our largest PC enthusiast facebook group, granted there might be one or two but most of the times it was just supply issue, took a bit longer for them to get the replacement
@sbcontt YT it just doesn't make sense for me honestly for them to do such thing, people would eventually realize something was wrong
1.45vcore is nutty when pushing these chips at 100% heavy load. This video is fantastic thanks for doing this experiment!
Thats the Vcore I used to set in my old Phenom II x6
@@ikarus1111a a few years back it was normal to have a higher voltage, it depended on the architecture the given cpu was built upon. Bigger nodes could handle higher voltages more consistently, smaller ones require less and less to slow down degradation.
My 5600X is reporting burst voltages up to 1.49GHz.... never sustained for more than a second or so before dropping to like 1.36v but still, should I turn down my scalar?
@@nothing.mp3 1x scalar max, with just pbo on auto frequency and voltage then whatever pbo / the smu decides is 100% safe. whatever this video shows is 1000% damaging
Still pushing 1.48 volts (vdroop to 1,42v on loads) on my Fx8350@4.75Ghz (Aircooled, fsb Overclock 237x20, without LLC). I had it for 9 years, but i began OC it 5 years ago.
Keep in mind that voltages really depends on the CPU you're running and most importantly, the temperatures.
I'm glad someone did this. Thanks for the work you put in, you'll put a lot of minds at ease.
If it's viable for you, I hope you continue doing these tests in the future.
I still use my AMD FX 8320 with a slight O.C. It crashes and freezes from time to time but it is 8-9 year old CPU. I don't think the OCing is the killer of CPUs, it's crappy PSUs. Never cheap out when buying a PSU. I had extremely bad experiences with crappy PSUs.
yes! power delivery quality. my homie has been having a problem with 2 different computers i built for him, i ended up using a power filter power strip from the wall, and he has had no issues since.
Same here. I have an FX 8350 that I just replaced with a 5950X system. Everything was working great except the 13 year old PSU when I decommissioned it.
@@glsracer thats an upgrade and a half
@@ThunderingRoar yeah, it's pretty nice. I use it as a VM server, though the RX590 could do 1080P gaming.
I had 2 Hipper PSU -both failed and almost killed the motherboard and at the end they did kill the cou or motherboard since I downclocked it to half to keep it running.
People don't realise that transistors haven't really shrunk by much over the past few generations of lithography. The increases in transistor density have come mostly from better resolution from the masking processes allowing for the transistors to be packed closer together without interfering with each other.
It's much much more complicated than this. Certain high performance areas or "hot spots" require more spacing, some areas require more room for interconnects etc. You could also say that even Intel's 14nm process had so much optimizations thanks to continuous improvement of IC design (e.g. better software) that it could compete with nodes with lower nm number in their marketing name...
AMD is mostly about jumping nodes and less on refinement... Because refinement doesn't resolve silicon lottery from happening anyways.
Yup, Physics is a bitch. I don't know when exactly but at some point transistors will just reach a point where they never get smaller again. These masking processes will also hit their limit at some point, so we're going to be stuck in terms of raw transistor density at some point.
@@Broniath we still can come up with better designs and things better than transistors
@@khhnator At some point there will be a an impassable hard limit though. Would actually be interesting to calculate what the theoretical limit of computation is, probably something related to the electrons you're moving getting close to the speed of light, although I suspect the practical limit is much lower.
I would love to see how a 12900k fairs running at high voltage and 95c for a year. It runs very hot.
Probably not well. It is a 240 watt CPU after all if you leave the UEFI settings the way the board manufacturers have them. Gotta have those auto-overclock features enabled to make your motherboard seem like the fastest one. -_-
Think of the power bills xD
Intel have been in the game a long time and they have the best chip voltage and thermal regulation in the game, I think Alderlake will be just as reliable as all the chips before them
@@TrebleSketch lol the power bills…0.78$/24hrs. 136$ for 4200hrs, which is almost 2 years at 8hrs/day full blast…Power is really irrelevant bro…If you spend 750$ on a top of the line CPU and then complain about 68$/year in hydro you need to get your priorities straight.
@@radugrigoras I guess I didn't realise power was that cheap where they are.
It's like ~0.40AUD/kWh here (plus ~1$/day of supply charge), if it's just CPU heavy benchmarks for ~300W of total system power that'll be ~0.13AUD/h and ~3.2AUD/day +1$ supply charge. For 4.2AUD/day and 4200 hours is around 175 days.
So for ~175 days (~2 quarters) it would be, 560AUD W/O supply charge (735AUD with SC). Guess the cost just sounded crazy on my end 😅
outside of extremely heavy use with over clocking, these chips will be good for then next 20 years, if the past is any indication. I consider a working 486 to be usable in certain circumstances.
I'll bet in normal desktop use they'll do 20 years with absolutely no issues. Doubt someone will be using them for 20 years though. Intel CPUs from 10 years ago are still very much usable for a lot of tasks but you can upgrade to so much faster with not much more money.
Outside of extremely heavy use - which I do not do at all, I cannot find a reason to change my i7 2600K at 4.2 GHz 1.35 V. Even gaming 1080p runs everything fine with an RX 580, because who can afford a GPU now...
@@rkan2 I agree. The only tasks that are really unfun with older multi-core chips is >1080p/30 video playback and any gaming if using the igp. Normal net browsing, email, and word processing is totally fine.+
I partly disagree. Even with typical browsing/email/etc. OS nowadays are running more in the background and assuming you at least have a quad core.
Now if you run a lightweight linux setup, yeah good to go...
@@rkan2 yes my i5 3740 can testify to that, almost 10 years and still going, not that strong but i can wait another year.
"After 5 years upgrading would be totally fine..." Watching on my 11 year old dual Xeon PC... 🙃
There is something that i think this test is revealing.
The 5800X has perfect CCX's as it requires all cores on each ccx to work.
The 5600x can have a CCX that failed QA for the 5800X as it doesnt require all the cores to work as 2 can be disabled, so being lower quality silicon.
most 5600x use perfectly fine chiplets, zen 2 has yieldrate of over 90%, amd doesnt have enough " bad " chiplets to use for a 5600x and 5900x
@Georg Warhead On zen 2 in time newer chiplets were better in OC and lower voltage so it can deppend on revision (New revision = better).
statistics sample is too small to assume that
and second 5600x have had lower voltage then 5800x by default for the same frequency
Kind of thinking the same thing, stock clock for 5800x is usually the highest clock, but that more cores are usually lower clocks keep in power threshold.
It doesn't mean that though? In fact I could come up with a simple model that suggests the 5600x should have better CCX. On average the number of CCX chips that are going to have all four be working, *and* all four be higher quality, would be lower than the chance of having a chip of 75% of the area (basically a 6 core CCX) having better quality silicon. If you look at it statistically assuming that each area of the wafer has an equal chance and that one area being biased doesn't make the area next to it biased, then statistically you would be more likely to get more 5600X CCX that are better than the same at 5800X.
In reality my model isn't accurate, because my assumptions don't fully hold. But your assumptions don't fully hold either. It's too hard to say which one of our assumptions is correct. In reality we're both going to be correct to a degree. Both of them combine + tons of other factors to give very complex yields. And when you're talking about this sort of far out degradation it becomes even more complicated, and there's a good chance that no one (not even TSMC) has good data on this.
This is a great study! Wish for more like this.
Awesome test! Thank you!
To me this means: Just re-OC your processor after some years. Or maybe when you are dealing with some sudden instability after some years - just recalibrate the OC.
Besides -- Ryzen 5000 runs faster on cooler temps anyway. If you use less voltage, temps are lower, which increases stability, and gives you a faster clock speed. My 5900X only needs 1.375V to run 4.875GHz
You wouldnt need to run 1.45V. So if you approach the CPU that way, it probably can have a much longer life too.
It should mean “I don’t need to worry”
He is running 1.45v, the degradation increases exponentially with voltage. Even 1.4v is barely comparable.
Agree. Even running a 5600x at 4.750 mhz all core, can be done easily at way less then 1.31v and for someone that use, let's say, a pc for 8 hours every day, 4 of them gaming, it should have no major issues for years. Just need to have care of the cpu.
@@andreabriganti8621 And the sad shit is most people care so much about their CPU only to sell it off after 2-3 years and upgrade to something.
I'd love to know how that compares to the 12900k Vs an overclocked 5900X.
@@col.hanslanda2013 I can't judge that. It's theyr money and we don't know theyr reasons and " needs " but I got what you mean. I'm used to buy hardware and use it for 5/6 years at least. I'm without a pc from almost 2 years now. Still waiting for better prices and avaible gpu, while trying to save money. I don't feel safe to go into used market unless I don't have any other options but, if there are people that sell theyr cpu/gpu in good conditions, there will be people, in a worse situation than mine, able to get it and have a good use of it. I will never understand the urgency to have " the best of the best " always but, again, I really don't feel to judge that. Sorry for the essay btw.
Thanks so much for doing this - both for yourself, so ppl will finally shut up about OC comp voltages - and for the general users out there. Super useful :)
If you can do it for AL that would be awesome, plus it’s getting into winter so maybe you can actually use the heating now? :)) 👍🏼
This is fantastic information and a massive service to the community. Thank you so much for doing this.
Been waiting for this, tyvm for doing this test.
Great video - would have liked to know the voltages these CPUs were at under load (during long term testing). With vcore set to 1.45 and LLC=4, assuming something like 1.33v? Not familiar with ASUS LLC settings/behavior.
Considering In the video while running r23 it hovered right below the input voltage id say the llc is pretty linear
LLC4 is a slight slight vdroop
01:55 You can disable updates to prevent that. Or you can simply disable automatic reboot. Not via the UI, of course (because Microsoft), but via settings in group policy.
Just unplug the network.
@@xeridea That's nonsensical. Just disable your network adapter.
disable network while chia farming .... ding ding ding we have winners.
Awesome testing. Nice work man!
Good job sir!! I have learned much, thank you.
I remember people spreading this in the Zen+ days, saying some low setting for Vcore was *way* too high and would murder your CPU. Half the time I swear it was trolling. But either way, after Zen2 I pretty much stopped feeling the need to do manual overclocks on these CPUs. AMD does a great job squeezing performance out of them.
Well one of my 3900X at stock (bios "enhancements dissabled") it peaked at 1.45V alot and drew around 155 watts fully stressed and didnt even reach the base clock. I do love AMD for what they have done to the market but boi do they have hellish variance
With pbo the 5600x is a beast, no need to play with voltages and frequency
That is some awesome testing
Thank you, this is very reassuring! I was even worried about going above 1.3V. Kudos!
I cannot wait to shove this in people's faces.
Great video as usual .. I really waited for this video , i want to see more similar videos like this and do this test on ADL
I have an AMD FX 8320E 3.2ghz OC@ 5.1Ghz/ 1.433v core since 2013 when it came out.
runs same clock and voltage with same MOBO, PSU, and RAM since 2013. and still runs same temps 50-65c @load. no crashing no degradation that i can see. my son has been using it since 2017 when i upgraded to my Threadripper. Make sure your PC is maintained and repasted when needed and clean from dust and dirt, it really helps alot.
Im scared too pull out my AIO from the CPU. The hose are stiff from 2013.😄
Kudos for running a long term test like this! I would agree with you that if you are already overclocking and overvolting, then even a 40mv tweak to keep it running is no big deal. Perhaps you could continue that same torture test for another six months?
And yes, absolutely, I would love to see you do a similar torture test on Intel Alder Lake CPUs. :-)
But I do not envy your power bill! LOL
Great video. Thanks for this well done testing and reporting
Great video, I would like to see more testing. Thanks!
Would be interesting to see more thorough stability testing. I mean, passing three runs of Cinebench is a very low bar.
But I guess the most important thing is that, well... the CPU might degrade after a few years of heavy OC. As long as it still works after a small increase in voltages and/or reduction in frequency, it's not a huge deal.
Whole "burning" heavy load test was stability testing. If there were any issues he'd notice them in benchmark error, event viewer messages, etc.
@@sliwa95 Which test are you referring to?
@@sliwa95 What? you need to test a CPU in various applications to draw conclusions, I had a dying 3700x and it crashed more in light games where in Cinebench it ran fine.
My girlfriend's 6600K recently got stuttering in some games after being overclocked to 4.5ghz at 1.25v for 3-4 years. I increased the voltage to 1.3 and all is good. After the next 4 years I'd increase it to 1.35, but I'm sure we'll upgrade her system before that.
May be even the PSU, not even the CPU.
Great long term test Roman, really insightful, much appreciated. You mention not being sure of full actual up time because of Windows updates, curious how come you didn't just set the update service to "pause" for however long and then when you're ready, do a manual check and update on all machines one time?
Thank you for doing this test! Awesome!
Very interesting. I undervolted my previous intel i7 4790 cpu not much after purchasing it, and it was stable for some months then it started crashing when I was rendering. I increased the voltage (still undervolted) then after some days it crashed again. I finally had to increase it again but it was still undervolted. Since then there have been no problems, it is still slightly undervolted. I was using it heavily for rendering, then Blender and no further degradation happened ever. It has been working for 7 years and its still working perfectly, I just upgraded to a ryzen because I needed more performance. And my 11 year old laptop i7 (despite overheating issues) stillmworking without crashing. Im pretty sure without overclocking most cpus can work for decades, and just as this videomshowed too there is no reason to worry even when overclocking.
Very impressive. 1.45v is much more than I used when max out the oc. 1.35v at most
1.35 is very manageble imo
my r7 1700 got 1.3975v for 5 years no degradation
If you see that even 1.45v with thousands of hours of prime 95 is no problem you should really not be concerned about normal overclocking even if the cpu is used like 10 years.
@@ThaexakaMavro
My r5 1600 is undervolted 1.289v for 5 years, but only 3.8GHz. Temps are very cool though, 27C idle on single tower air and only draws 22 watts package lol
Amd in fact suggested this cpu should be at 1.35v max to be " safe " but even until 1.39v there's no major issues. Expecially the 5600x. Can be o.c. to 4.750 mhz all core at Lower then 1.31v if undervolted with pbo and then manually o.c.
fantastic video, this is very informative, wish a lot other tech sites would do this kind of tech journalism.
Thanks a Der8auer for this experiment. It clears few important doubts for me and many others. Thanks again for putting time and effort into this.
2 decades ago, I ran an AMD "Athlon" Pluto 700 MHz @ 1 GHz (1.85V with Golden Finger device - yes, I still remember, I'm a nerd) for 3 years, with very heavy usage (and 3 years back in those days would've been a lifetime for a PC, because a new bleeding edge computer was already obsolete after 1, maybe 2 years). Anyway, being a retro enthusiast, I continued to use this CPU for all these years (I even used it yesterday for 4 hours), and it just works without a hitch. So... yeah, I would say that 21 - 22 years is pretty good for a very heavily overclocked CPU. :-)
I have a very old computer in my house with a pentium 75mhz lol, still works fine
I remember having an athlon xp 3200 that I left powered on 24/7 for months and the one time I decided to shut it down I was presented with an athlon 2500 being detected at the next boot which back then was the difference between game runs or game don't run. Ended up selling the potato with a lower spec gpu and sold the Geforce 5950u I had for a good whack (today's equiv would be a 3090 ti) and bought a fkin bike instead. I suspect the CPU was a graphite trace scam chip from dodgy stock on Ebuyer. Being around 17 at the time the bike was a better choice anyway :P
Have you overclocked your kitty? Tail looks a bit unstable ;) Great content.
Thank you for doing the research! Would love to see the same for Alder Lake chips 👍
Thanks for the interesting talk in the voltages
I wouldn't call inability to hold specific overclock at some voltage as "failure", I bet it is perfectly stable as stock setting, which means you still definitely get what is advertized on the box.
I remember being told the 22nm process is too fragile and unstable, so Ivy Bridge can't take more than a mild overclock when I built my system. I'm still running that overclocked 3770k today. It's fine. It's still good.
Same here with a 3770k
Things been churning along for years just fine, already eaten one PSU.
As Roman said: people say this for every node shrink. But it's always the young whippersnappers who say it because this is their first rodeo. I got into overclocking on my AMD K6 400MHz, and I've been overclocking the snot out of everything ever since, simply because buying a beefier cooler is cheaper than buying the next tier CPU :P
@@andersjjensen Same, started with K6 and have overclocked every processor I ever owned since then to their limits. And I still have all of those CPU in working condition. Including my very first K6
Awesome testing you've done, and you're right- it'd be better if there was a bigger sample pool, but I still don't think CPU's degrade as fast as people think, unless voltage is really jacked up. Been rocking my 6700K full time at 4.7G, and 5 years later, it hasn't hiccupped or given any issue. Side note, really loving the new ink!
Great video! Thanks for making it!
Yes please do it with the alder lake if possible. Thanks !
It's been like this since the 90s. There's no reason to waste time and effort to come to the same conclusion. Unless you're overclocking competitively, or is a complete tool who just hammers the voltage on a crappy board with a crappy PSU you're not going to get degradation worth a damn...
I have a Ryzen 5 3600, and using Ryzen Master, auto overclock, it boosts at 4.3ghz but my voltages shoot up to 1.4-1.45v regularly. I started manually overclocking, and I can run stable at 4.4ghz with only 1.2v. I did this because I was scared of the degradation. I have actually successfully pushed 4.5ghz at 1.275v but I haven’t done any heavy testing yet at those settings. Seems like maybe I won’t have to worry as much as I previously thought
4.4ghz at only 1.2 volts? youre lucky. my 3700X needs 1.25Volts at 4.2ghz
Damn, my 3600x did 4.4@1.294v. If I was able to get it at 1.2, I definitely would've pushed for the highest frequency I can till I hit 1.3v.
Great video, great tests. Thanks a lot
Enjoyed this plenty, super cool test!
I didn't see any long term temperature data. Was that monitored even periodically? Reason I ask is I think those Corsair A500's the air coolers that had cold plate quality issues? Heat affects silicon life as well, it might be worth taking a closer look at the A500 that was on the problematic 5600X.
Those initial reviews were all terrible on that A500 due mostly to the issue you describe. Because of all that bad press, they were selling these at like 30-50 dollars. When I bought my system this year, because I'm a huge fan of huge air cooling towers, I just said 40 dollars, why not. . . And I got it for my own 5800x system.
I can tell you that my A500 is completely flat on the bottom. I'm sure like any plate to plate its not perfect, but it certainly isn't noticeable to the human eye or noticeable in its performance. I am of the belief that initially these A500s had a huge machining problem, and due to all the bad reviews, the product was essentially doa with making a splash.
I'm happy to report that I've tested my 5800x using both the A500 and a noctua d15, I run a custom overclock, and the temperature I get with it 24/7 is essentially the same as the d15. I max out at 82c running torture tests long duration with the A500. With the D15 I was maxing out at 81c. It is essentially a wash. A 40 dollar air tower versus a 100+ dollars air tower, getting basically the same temps. Idle, load, torture. Aside from more noise, I wouldn't be able to tell the difference.
I think this was a cooler that has been thrown into a heap of negativity after the terrible release, forcing corsair to try to just sell them off even AFTER they seem to have fixed the machining issue. For 40 dollars what I constantly see it advertised at, it's an unbelievable bargain due to the bad press. I suppose maybe I could have just gotten lucky, but I assume they probably did fix their process after launch. This obviously is my personal experience with it/opinion of what happened. For me, It's been a wonderful cooler for the hot 5800x. At a steal of a price. Nothing looks better in my 15 year old modded HAF 932 then a gigantic air cooler!
@@mayaddy6228 Nothing I said suggests they couldn't be good. The issue was quality control and consistency. One cooler working out doesn't offer any insights into his 3-4 all being equally reliable.
I'd love more data. I'm talking multiple CPU's running at multiple different voltages. To see at what point degradation becomes intolerable. Of course gathering such data would be a long project, as this was.
As a tangent, my r5 1600 was capable of 3.9Ghz at 1.375V new, i ran like that for 4.5 years, occasionally dropping the clocks as the system became unstable (High res gaming coupled with occasional CAD work), in the end i was only able to get 3.75Ghz at the same voltage.
Lol, I had same CPU and used my motherboard's default oc settings which would push 1.4v at 3.9Ghz, I used it for 3.5 years and it never skipped a beat. But I did have a decent cooler so the CPU never went over 60c during gaming.
Mine did 3.85GHz at 1.3625V. Dropped to 3.825GHz after 4 years.
Niiiiceee!...I was sure they were going to fail at some point but is very nice to see they did not really degrade nearly at all.
Ty again very much!
I intend to OC my 5600x and now I know that will be fine for the next 5 years
I'm thinking how will this effect the new CPU mining meta,i hope it does we don't need new GPU situation happening with CPU and scalpers,
CPU miners will undervolt for max effect per watt
Unless they are stupid
Stock CPU can last much longer than ten years. More like twenty or thirty years under NORMAL use... possibly even more. I own many CPUs that are ten or more years old. In all my years I have never had a CPU die no matter how hard I pushed it (Intel Coffee Lake, Skylake, Ivy Bridge and Sandy Bridge at or above 1.4V and AMD Zen 2 up to 1.35V).
yea that's why oc degrades em faster... but its also irrelevant, instead of 20 years.. 10? ok its trash by then
That is definitely the case with older cpus. But honestly with these we have no idea how they will last even at stock clocks and voltages.
great stuff. I just might overclock now (X470/5600x/5700XT). As with others, thanks for doing this and spending the time and money to gather this data.
I'm getting the same CB results on a 5600X with stock settings @ max 75W package power. I did heavily modify the cooling system, so it can keep boosting @ 4.641Ghz (modified Apple G5 air cooler max 60°C). So I don't really see a need right now to overclock this CPU. Cool video, great to see this chip will last !
I am a bit weird, but I have never pushed my CPUs hard on the voltage.
But in my world, my CPUs aren't done with their life after 5 years, once it is done in the primary rig, it goes over to the home servers and do a workload there.
So that is why I still have Core 2 CPUs in daily operation today.
I have no use for cpus slower than a smart phone
i JUST retired my 3570k this year. And by retired i mean gave to my buddy who i using it to game with me on the near daily >:)
Best cpu ever made
@@Steve-ph7qn than you aren't very imaginative or you don't respect the tech. There are lots of cool projects you can do with an old PC. Pretty much anything you see someone do with a raspberry pi can be done with a PC with a core 2 duo. Pi-hole, smart walls, servers, security systems, etc, etc. My main computer might be up to date but I still have a laptop with a old pentium and a aio with an i5 2400 doing work. Plus my server which is an atholon 200e (or something 2gen am4 atholon the exact name escapes me).
@@TechGuyBeau I don't know if it was silicon lottery or what, but my 3570k was able to run stable at 4.7 with only a smallish voltage bump. I ran that sucker for years before I sold it on eBay and I have little doubt it's still out there crunching numbers somewhere.
@@GummyBearRacing Yeah, I don't respect 20 year old tech, you don't respect power efficiency and like to horde. I agree, we aren't the same.
I had my 2500K@5.1GHz/1.5v running 24/7 for many, many years. I ended up with 4.9GHz/1.52v before upgrading this year to 5600x and I keep this one per core at -30 -30 -10 -17 -30 -30 @4.65GHz because I didn't see much of a benefit keeping it OC at 4.8GHz.
Sandy bridge, the little CPU that could. I had a 2600k for 8 years before i upgraded. Excellent OC'er
Thank you so much for this video
I was heavily concerned about my 5800x degrading faster because of high temps but i guess it was overthinking of me
Awesome experiment.
One thing to mention is that most aging effects (e.g. Electromigration and Bias Temperature Instability) have a relaxation effect. So when you stop providing stress, the "aging" effect is actually reduced. So the idle time can negate some of the stress and the long periods of constant stress should have a much more dramatic aging effect than the comparative cumulative aging of a system that is used 3-4 hours a day. Electromigration for example, immediately begins to recover in the absence of current in the respective interconnect. In other words, your testing scenario is likely more conservative than real life, even more than what you conclude.
When designing chips these days, companies do a similar experiment as this, they use extreme conditions under unrealistic circumstances. They also design to a specific lifetime, e.g. 10 years. But when they design to 10 years, what they are really designing for is the first 5 percentile or so of chip failures. So the vast majority of chips will actually last much longer than this reliability design point. This of course assumes the chips are without defect, but many companies do a burn-in style test that tries to induce failure in these defective chips.
Wow had no clue you would even try to attempt this test crazy 🤣 love this so much no one is making interesting content like this nice one derbauer I have a 5800x so good to know my 1.3v 4800mhz is gonna last a long time but I'm probably gonna go to am5 anyway but I'll sell this at a fair price to a kid so he can have a really good rig.
based on observation early Ryzen 3000 was not voltage tollerant and there was a lot of variance chip to chip on max OC and fit voltages, this became less an issue over time, my early R5 3600 quickly stopped being able to do 4.4Ghz but some of my use cases were pretty close to prime95, currently I find the max stable overclock of my R5 3600 is 4.0ghz at 1.25V.
basically Im saying later Ryzen 3000 i.e mid 2020 or newer and ryzen 5000 is probably more voltage tollerant, if you still are paranoid just test FIT voltage under prime95 and assuming your work load is gaming you can probably go 50-150mv higher or if its rendering then maybe 25mv higher at the most
it's very hit or miss and that's just the risk of overclocking but the 3600's were much more prone to the variance in cpu's since they were just failed 3600x's.. i have one of the early chips which does 4.2Ghz all core boost with i think 1.35v or what ever the stock voltage is at 66C full load. but honestly there's no reason to overclock the 3600, better off leaving it using the stock boost with better cooling.66-68C is the threshold for when it starts dropping the boost below 4.2Ghz but even then it'll do 4.15Ghz up to 72C.
@@sirmonkey1985 yep, for most use cases stock boost is more than enough but blender rendering or similar loads is the only time i recomend a manual OC because you can squeeze a bit more performance out of your CPU at low voltages, seems to be typically 200-300Mhz which might add upto quite a few hours off of a render.
i ordered a 3700X on day 1 release and it only does 4.2ghz at 1.25 Volts
Can you do any testing on the main boards that you used to see if they degraded at all?
Also, what temperatures were the CPU's running at roughly over this time period when under load?
Amazing experiment! Would be nice if you could also measure the opposite scenario: a underclocked/undevolt cpu and used a stock one, as control, and estimate how much life the underclocked cpu gained in comparison. I realize my 5950x dont need to run faster than 3200mhz for any of the tasks I put it through, and I also lower the v.core to 0.7v all cores! These chips are simply wonderfull! Cheers for the amazing work!
Overall this is good a test and good to know. Some thoughts I have from it.
When the cpu is running 1.45v you are using llc4. Thats very droopy voltage. I wonder if the cpu is seeing a much lower voltage than this. Low llc will help with transient spikes to help with current peaks and lows. The reason people would run higher vcore is to run higher cpu ratio. Here you use 4.7ghz which is stock. A cpu running lower frequency will degrade less because the transistors switching are doing less work.
The ram used would be very easy on the memory controller part of the cpu. Again less degradation.
The heat cylces are unfortunatly extremely unrealistic compared to real world useage. There was basically one or two because of the windows update. Think of a lightbulb. If you left it on it would last a lot longer than turning it on and off constantly. It would be interesting if you did the same test but with a stop start appoach.
Also temperature average would be useful. Because if the cpu was thermal throttling that could have reduce voltage for some time.
Again this testing is very interesting and good to know. I just hope people dont have a false sense of security and experience bad degradation and consequently need another cpu.
Idk when this fear mongering of "degradation" started. In the past people used to overclock the piss out of sandy bridge CPU's to 5GHz and nobody ever talked about the lifespan of the CPU going down. Nowadays people say running your GPU or CPU overclocked or at 80 degrees is going to kill it, wtf 🤣
Yeah imagine telling them 90-95c on a Pentium 4 was just normal.
Very nice video, in my experience hard but stable OC is good to run for a upgrade cycle 4-6 years. I only ran into deg when I did some rather high voltage and clock speeds at close to zero degrees C.
and even then it was not that the CPU could not run a high OC under normal conditions but it was not able to repeat the absolute maximum.
Could board variance between the motherboard the CPUs ran on vs the one for testing account for some of the differences in vCore setting? Even if they use the same components or BIOS. It seems like not a big deal to me. Especially since you used a higher voltage than what is needed for most air or water overclocking on Zen 3.
All those long term tests are invalid unless kitty was vigilantly watching them at all times. And I suspect kitty was found sleeping on the job a lot! 😂
it's like CPUs have had the 1.4 daily limit for quite some time now
Thanks. ~4.3@1.25 all core Tdie@80C w/ actual power draw 170W 24x7 feels more confident now.
Good video. Got me thinking if you could get CPU's, ram, gpus etc etc thats been heavily used either from private persons, or maybe from a server farm and test those too.
My 3770k is amused
My FX-6300 is amused too :)
LOL I still have 3770K running 4500Mhz for 10 years now.
don't blame the cat most of that is your dead skin cells
I really enjoy this content! I will say I've seen a lot of odd crashes among Ryzen 5000 CPUs, but it's something around 5% or less of what the company I work for sells. It probably looks higher to me since I work RMA as well
What about running samsung b-die on 1.5V, will that shorten life span on cpu enough to matter within normal upgrade cycles?
Last week I bought my first Seasonic PSU, I had only used Corsair before. The power of your ads!!!
Thanks for this vídeo!
Great video, thank you
Excellent presentation. 10/10
too often we hear how fast and a barrage of tests with software that an average user would never run. As a person that builds their own PC's for home use and gaming etc I want the best value for my money and often spend a couple of months trying to juggle the pro's and con's between red blue etc. I recently built my 1st AMD system as the final condition was the power consumption. The AMD was simply more efficient but always in my mind was the question of reliability. I spent a lot of my savings on a x570 master plus a 5900x but do my upgrade every 5th - 6th year. Your presentation has just eliminated any minor worries I had that longevity might be an issue. Thank you.
This is really good content. I would like to see would there be change if you ran another 4k hours. Also yeah intel test to compare their quality too.
I was concerned about long life on my 5800x, also im running 4 dimms with 2 different ram kits at a slow 3000mhz CL18. I set a 1.28 max voltage with -22 all core curve. Doing 5900+ on C20 without OC seemed a nice result performance wise, is it? slight above 80ºC worst temps.
Thank you for your efforts! I'm studying CPU architecture and may I put this videos link for reference?
I run my R9 5900X with a set overclock of 4.5 GHz set by the CPU Core Ratio in the BIOS, with an Offset Negative Automatic Voltage. During normal operations (anything other than gaming), the CPU runs at 1.1 volts, and I leave it on from 11 PM to 6:30 AM running a video on my monitor with tropical underwater scenes, and soft sleeping music. The computer comes on when I get home at 5 PM to game, and do my email, visit forums, and watch CZcams. Been doing that every night for the past year-and-a-half, and it's still going strong. Calculated time being on is about 6300 hrs on this CPU.
omg we have the same cat! They freaking look alike 100% including the body size!
Fantastic video, I would love to see more with Intel. Ty very much for the work done. 🤘😎🤘
Great work! I would like to see you continue this test for another six months on those CPUs, and add an Alder Lake (preferably i7 or i9) at 1.55 or even 1.60v to this test. I would suggest doing CPU mining as your test load, to offset your power consumption.
Note that you can set your connection as "metered" to reduce Windows Update reboots. Also note that you can look in your System Event Log after a reboot to determine exactly when the system rebooted. Chances are, this history goes back the whole six months!
Why intel at even higher voltages? Would it not make sense to use the same 1.45v to compare how monolithic gets affected vs chiplet? I mean you could not run 1.6v unless you cooled intel down with liquid nitrogen lmao.
i have been running 9900kf at 5.0 all core 1.36v 24/7 for 2+ years, good to know that im probably in the clear until DDR5 is a bit more fleshed out and the next series comes out, but i would still be interested in this on 10-12th intel as well and maybe even Threadripper
All ur knowledge and experience and you cant block the update server in ur DNS or Firewall?
The hero we don't deserve but need. Please do this with the Alder Lake aswell! Sick series
What was the temperatures during those prime95 and 3d marks tests? I believe temps also matter for degradation
So happy I invested in the 5800X this spring. Finally payed enough to secure a 3070 and am a happy camper
Nice test, i do have a suggestion.Certainly do it with alderlake. Heck. Do it with every generation of cpus. But find a way to add tests. For instance: it would have probably been great to also have tested to what degree these cpus were able to undervolt with the curve optimizer. It seems this would have given a more sensitive measure of degradation.
Thank you for the tests, do some for 3rd gen Ryzen since i feel like those are very wide spread now & all
Hey Derbauer do you still got your old 3800XT? And do you plan on doing anything with it for a retrospective review too? 🤔
Here’s a subscriber from NYC, thanks for the great review, can you please test the i5 12400?
My first 2 5800X's I could never get stable stock, many people thought there was problems with the CPUs after 9 months I got my 3rd CPU which was perfectly fine. Similar Experience for others too. Maybe it was an issue with early silicon. It was a big headache, fun at first to troubleshoot but in the end annoying, however now that's in the past and I'm enjoying my 5800X quite a lot.
Thanks for the vid, good to know I don't have to worry about going through the above again.
What are the GPU's on your shelf on the back? I don't recall any particular ASUS GPU with fan&shroud design like that.
14:07 you shook me there for a sec i thought it was to the viewers
Hi. I'm working on my thesis doing docking with flexible receptors and I've noticed an incredible jump in performance going from an I5 7600k to a Ryzen 5 4600h, using 50% less ram. I'd love to see scientific performance on new cpu architechtures but I can't seem to find stuff like that on yt