Welcome Guest ( Log In | Register )

22 Pages « < 4 5 6 7 8 > » Bottom

Outline · [ Standard ] · Linear+

 AMD Ryzen, AM4 / AM5 Platform

views
     
Bonchi
post Dec 27 2020, 12:23 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(diehard1979 @ Dec 27 2020, 12:05 PM)
So this happen to both intel and amd cpu?
why they say not all? is it related to some model only?
*
issue is only on ryzen 5000 series documented by a few testers. while I’m a 3000 series here, no issues on my end. Intel is problem free as well, it’s just the article quoting without source.

perhaps it’s the same like zen2 microcode issues that is causing infinite loops last time.

This post has been edited by Bonchi: Dec 27 2020, 12:27 PM
Bonchi
post Dec 27 2020, 04:04 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(SSJBen @ Dec 27 2020, 03:38 PM)
I agree. Understandable they wanted to chase numbers because it's the only way to bridge the gap to Intel, as far as majority of consumers are concerned that is.

But they've chased the numbers game for far too long now and it's mind boggling that Zen 3 being a refined Zen 2 has more problems than its predecessor which already had over a year of after launch "fixes".
*
not really, zen3 overhauled the CCX to improve memory latency. the only thing still the same with zen2 is the chiplet approach.
Bonchi
post Dec 27 2020, 04:33 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(SSJBen @ Dec 27 2020, 04:14 PM)
Yes but that's almost like saying I re-painted my room and added a few new shelves and it's now a different room.

They're really still the same arch, re-arranged and refined at the end of the day. The overhauling they did is not as ground breaking as they're advertising.
*
The rearranging ccx layout is quite a big feat especially when it comes to the I/O portion. As you know the CCX dont have to move back and forth as often with the IOD so there’s alot of things has to be re-written in the microcodes.

It’s like engine swap a wira with an evo engine... it wont be plug and play, the suspension, brakes, gears, chasis stiffening etc all need to be done before the car is drivable.

That was why initially they dont intend the zen3 to have backwards compatibility on chipsets. So now Im even more curious to see how it’ll work on a B450/x470. I was originally planning to test that out, but i lost patience with all the problems b450 gives with my zen2 so ive switched to b550 before zen3 lol.
Bonchi
post Dec 31 2020, 12:07 AM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(Andrewtst @ Dec 30 2020, 05:29 PM)
Completely removed the MPV Player included de-associate the file extension.

This MPV crash actually make my PC Dllhost.exe application error even using others application to playback video later.

I just did sfc /scannow and also DISM /Online /Cleanup-Image /RestoreHealth to solved this error issue.

Beside this, I believe F31q BIOS also solved the crash issue.
Yup, and I think this F31q BIOS do solved all my crash issue, now not even an issue for letting my RAM in XMP Profile 1 auto.
*
which means it's facing an infinite loop. A very typical problem with ryzens especially when doing development for API as their microcode is full of bugs. Running on VM solves most of the issue.
Bonchi
post Jan 1 2021, 02:38 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(SSJBen @ Jan 1 2021, 01:56 PM)
Actually if you guys read deeper into the BSOD and WHEA reports, most of them stemmed from early production silicons of Zen 2 and of course, Zen 3 now. Users who has issues are those with manufacturing dates early on in the product's production cycle, where as those who bought much later (like 8 months after) does seem to have no issues, this was with Zen 2 at least. Seems like Zen 3 is following the same trend.

Yes, sly AMD for keeping hush about it. But similarly, Intel was hush hush about their security flaws too. Both companies hide shit under the carpet.
*
it’s just more of design error in early production and also a microcode that is full of bugs.. and not to mention agesa that is not transparent.

my amd cant do alot of things currently unless i run the stuff inside a VM lol... maybe cuz the stuff i do rapes the ram and cache more than average gamers /video renders / benchmarks and required high accuracy. So it’s crashing or errored out very frequent.. Especially during the times when I use my own rig while working from home. Remote into my company’s redundant xeon server and it works flawlessly.

Till today, amd is still full of bugs and they dont seem to be fixing it... they just ditch and move on to the next gen.. like the agesa support for b450 like completely stopped. I changed to B550 out of frustration and the stability problems gone away due to the newer agesa. But many old bugs is still not fixed.

In the end we gotta see which flaw weighs more than the other. And regarding “security flaw”... as a guy who studied cyber security and ethical hacking i wouldnt call it a purposed design error or a cover up... it’s a security breach that was found by third parties. No system is perfectly secure... it’s more on how much effort an attacker wants only. If a similar effort is being done on amd, pretty sure alot of things will pop up as well.

And lo and behold... a bunch of breaches was found on amd recently laugh.gif where one breach is vulnerable since bulldozer.

This post has been edited by Bonchi: Jan 1 2021, 02:48 PM
Bonchi
post Jan 1 2021, 05:41 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(SSJBen @ Jan 1 2021, 03:17 PM)
Some while back I believe AMD said they'll never make AGESA transparent because they don't want their competition to know what their so called "secret sauce" is. Yes, bullshit 99. This was somewhat verified by the stilt in the Zen+ days too, I don't remember where he posted that on OCN.

Ah yes when the whole security thing blew up, AMD damn lansi saying they not affected. And of course recent security flaws they sweep under the rug.
*
vulnerabilities always discovered or studied on mainstream or enterprise stuff... back then intel was kinda the monopoly so of course there are millions of cyber security students and professionals used intel as their test bed for their thesis and research. Now when amd is growing big, these people will start scrutinizing on amd.. and in just a short window, a bunch of vulnerabilities are already found... and i believe there will be more and more.

Altho as home users, just like the vulnerabilities intel poses.. we dont have to worry la.. and we’re not really affected unless got hidden sex tape or some super profitable blueprint or something lol.
Bonchi
post Jan 1 2021, 10:18 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(SSJBen @ Jan 1 2021, 10:05 PM)
Whether it affects home users or not is besides the question, that's another topic. But doesn't change the fact that there are security flaws on both AMD and Intel, yet both of them tries to hide it under the rug as best as possible. Doesn't happen doesn't mean it's not there, ya know? If we want to talk about transparency, both Intel and AMD were never transparent.
*
from last I remembered, intel didnt hide it but rather released a patch rather quickly but it has a significant performance penalty where back in the days amd fanboys laughed at intel remember? They didnt hide on the security flaw but they hide the fact that the patch causes performance loss.

you should know la.. intel cannot really hide lol.. plus they dont have hardcore fanboys to defend them apart from corporate/enterprise rats lmao.

Amd meanwhile.. hehehe they just remained silent thus far laugh.gif (well to be fair their software engineers still have so much more major operation flaws to fix in the microcode before they got even time to look at security)
Bonchi
post Jan 4 2021, 09:34 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(sai86 @ Jan 4 2021, 09:23 PM)
https://docs.google.com/spreadsheets/d/1dsu...t#gid=527992713

Collection of Ram used, tested by user. stumble this from dunno which Youtube reviewer ady  hmm.gif
*
pretty pointless actually as OC is very subjective to lottery... unless if it's regarding XMP + mobo compatibility then yeah, that will be super useful.
Bonchi
post Jan 6 2021, 04:17 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(xxboxx @ Jan 6 2021, 12:06 PM)
Maybe Linus Torvalds don't know things.
*
he’s probably crashing like no tomorrow and end up running on VM right now lol.

Then again he only mention good about the price per core and ECC support on threadrippers/Epyc jer. Which is the basis of his switch laugh.gif. Nothing to do with ryzens.
Bonchi
post Jan 7 2021, 04:18 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(nrw @ Jan 7 2021, 03:07 PM)
follow what Tristan said.

Technically the situation for DIMM's with 16Gbit IC's (which wiill be on 32GB DIMM's) is pretty easy.

There is
1) Hynix CMR
2) Micron Rev.B
3) Samsung A-Die
4) (SpecTek which is relabeled Micron Rev. B)

All of them are having little issues reaching speeds until 3800 on Ryzen. After that Hynix and Micron ones separate themselves from the Samsung ones.

If you wanna run XMP. Proceed like TristanX proposed.

If you wanna OC you can save a lot money by buying 2400/2666MT/s rated sticks and overclocking them. In that case however you want to make sure to have the same IC's on both sticks you buy.
*
personally using micron rev B sticks. it doesnt like to go on low latencies but speed wise it can go quite high.. like 4000Mhz high.
Bonchi
post Jan 7 2021, 04:44 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(nrw @ Jan 7 2021, 04:24 PM)
Yup, cmr will behave similarly. Thanks for confirming.

DR 16Gbit (in 2x32GB or especially 4x32GB config) with CMR and Rev. B will hit the limit of the FCLK and even IMC first afaik.
*
Another interesting thing to point out is the micronB also tend to prefer lower voltage which is why could not really crank the latencies. Im running my sticks at 3600 CL16 on 1.35v. Temps also low because of that

Cant find much user feedbacks on these IC. Got them by chance when buying a lowcost 2x16GB dimm expecting dual rank but it turns out to be single rank with the micronB. got it at RM300 for 32GB so cant complain.
Bonchi
post Jan 7 2021, 07:49 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(Lagoo00 @ Jan 7 2021, 06:55 PM)
Hi guys , i recently upgrade to 5600x from 2600 , just wonder if i ever need to upgrade my ram too, currently runnin g.skill trident z 3200mhz cl16 at xmp
mainly focus gaming on 1440p with 3080 , so i guess the margin of fps increase is less than 10%?(10-15fps) if upgrade to 3600 cl14  :confused:
*
just try to overclock and see can reach that speed or not. Dont need to upgrade.
Bonchi
post Jan 7 2021, 08:32 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(Lagoo00 @ Jan 7 2021, 08:09 PM)
try once , and gotta reset cmos to startup again then gave up lol , sound too complicated to me 
also dunno y , thaipoon burner cant show what die my ram is
*
they should be B-dies. reset cmos when OC settings gone bad is normal. Depending on your mobo, difficulty of setting the overclocking values varies as well. You can start by following dram calculator, when you get the hang of it then you can calculate your own. Maybe more experienced sifu here can help you.

I only made my super cheap 3200mhz cl16 to run at 3600cl16 and didnt really tried to overclock the max out of it.
Bonchi
post Jan 8 2021, 05:19 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(pikacu @ Jan 8 2021, 02:16 PM)
i read here that first gen Zen 2 buyer have bsod issue, silicon issue?

im first gen Zen 2 user..bought on launch. how to check for.bsod?
*
silicon and agesa issue. Agesa can say like 90% solved ... only if you happen to use or do something in that 10% of instruction then it will crash, mostly seen in new launch games/softwares or windows updates.
Bonchi
post Jan 8 2021, 11:32 PM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(pikacu @ Jan 8 2021, 05:24 PM)
can give example to test bsod?
*
my examples are not gaming related..

It’s mostly when running some python or nodejs instances for analytical data processing which are very cache and memory intensive. Especially on forking child process which allows nodeJS to use multiple threads.

Another that tends to crash alot is when training models on the cpu.

Arguably, some might say it’s a coding problem. However these problems never occur on intel consumer cpu and xeons and there’s no error catch found. So I’ll just put the blame on amd.

Im not a gamer so cant say much on that.

This post has been edited by Bonchi: Jan 8 2021, 11:34 PM
Bonchi
post Jan 9 2021, 12:10 AM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(pikacu @ Jan 9 2021, 12:01 AM)
can provide the code.and compiler used? guide to compile would be nice..so i can do 1:1 simulation. wanna try it on my machine
*
it’s my former company’s property so i cant share. evenso, I dont have a big enough dataset to give you to reproduce those errors.

python and node dont need compilers lol. It’s an error that arises during execution. almost always certainly memory related.

from what my former team suspect, it’s somewhat .. in easier explanation, the execution caused the OS to run out of memory and crash out. Intel maybe have a better IMC so dont suffer these issues.

This post has been edited by Bonchi: Jan 9 2021, 12:26 AM
Bonchi
post Jan 9 2021, 12:39 AM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(pikacu @ Jan 9 2021, 12:24 AM)
alright noted.

if anybody got method to bsod zen 2..please share. wanna test it on my machine.
*
you can try google notmyfault or write a code that will rape your rams using an infinite loop. But kinda pointless because this is 100% will cause a crash.
Bonchi
post Jan 9 2021, 10:15 AM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(Fhaarkas @ Jan 9 2021, 05:57 AM)
Where the heck do you guys get Ryzen 5000 from? Paying cukur price?
*
alot of shops in lowyat offer full pc bundle preorder. pay full price up front and wait like a week to a month and you'll get it.
Bonchi
post Jan 9 2021, 11:27 AM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
QUOTE(pikacu @ Jan 9 2021, 11:08 AM)
1. last i check common users/gamers dont run prime95 casually nor for fun.

will arrange footage for 2, 3 and 4.
*
that's where it will simulate and force the 10% of instances that will crash a ryzen PC. By right it shouldn't crash at all on stock.

we should be expecting error free for 99%. It seems intel usually has some sort of better error catch controls so at least on my former company servers, when a task fails, the task just fails and the other services are still running. Meanwhile on an opteron/epyc, there's a chance it will just fail, bsod and reboot and you walk in the next day to find all the services has been reset. mega_shok.gif then it will be hours of data integrity check by the team.

My former company did gave an intern a ryzen7 thinkpad and she's kept getting errors we dont get on our i7... like android/ios virtualization, deployment etc so perhaps it's not reliable for dev use yet.

altho one thing I see beneficial on epyc is the price per core so you can run alot of VM which you can protect it with failover clusters.

However if it's for a browsing, gaming, rendering cpu where the 10% chance of crashes is not an issue, then the ryzen is not a bad choice. Cuz personally im also using one.. which is actually still fine if doing light work like testing simple scripts or replying emails when working from home.. However due to reliability, I still teamviewer into my laptop which i left in office to do most of the work.

This post has been edited by Bonchi: Jan 9 2021, 11:46 AM
Bonchi
post Jan 10 2021, 03:20 AM

KittehPowah
******
Senior Member
1,649 posts

Joined: Sep 2008
end of the day, from my more nuetral perspective based on my personal usage since i used both intel and amd for pretty intensive work..... cuz im in data anlytics line.
» Click to show Spoiler - click again to hide... «


TLDR:
AMD marketing boasting on consumer benchmarks because that’s where they can show their edge over their competition. Eventhough real world use is inconsistent. But no doubt the ryzen5 is a damn good buy. just the 8 and 9 kinda “depends”.

Intel marketing keep showing real world use cases because that’s where it still excels and they have better supports on certain features that will give them an edge (non gaming) however lose kawkaw in consumer benchmarks to the competition.

This post has been edited by Bonchi: Jan 10 2021, 03:44 AM

22 Pages « < 4 5 6 7 8 > » Top
 

Change to:
| Lo-Fi Version
0.4013sec    0.38    7 queries    GZIP Disabled
Time is now: 3rd December 2025 - 09:20 PM