• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

So anyone got a Ryzen 5000 yet? DC benchmarks? ;)

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
StefanR5R said:
I took a possibilistic approach there, rather than a probabilistic one. ;-)
Click to expand...
The ECC memory wiki is an interesting read. My gigabyte aurus master x570 board does support unbuffered ECC memory although I've read you need to choose memory modules that are on the QVL list. Something to consider I think.
 
Assimilator1 said:
Just reading AnandTech's review of them now (so no idea when they're actually available atm). Performance looks good!

So I know some of you guys are on the cutting edge 🙂 , anyone ordered one? Or got one yet? lol
Click to expand...

I have started building a new rig last month in preparation of the Zen 3.
Only missing GPU(To be delivered soon) Asus 5700XT

Sadly in Canada they are all out of stock on Zen 3. I'm looking at the 5800X.
 
Related to the earlier peripheral discussion about ECC DRAM in this thread:

"The arguments against ECC were always complete and utter garbage. Now even the memory manufacturers are starting do do ECC internally because they finally owned up to the fact that they absolutely have to."​

says somebody who knows stuff. :-) Check the source for why this person advocates ECC for main memory. Hint: It's not just the often cited cosmic rays which flip your bits.
 
StefanR5R said:
Related to the earlier peripheral discussion about ECC DRAM in this thread:

"The arguments against ECC were always complete and utter garbage. Now even the memory manufacturers are starting do do ECC internally because they finally owned up to the fact that they absolutely have to."​

says somebody who knows stuff. :-) Check the source for why this person advocates ECC for main memory. Hint: It's not just the often cited cosmic rays which flip your bits.
Click to expand...

Linus Torvalds threadripper build. I can't believe he went with the cheap stuff!
"In an ideal world, Torvalds wanted 64GBs of error-correcting code (ECC) memory. He settled for a high-quality generic DDR RAM."

I ordered a couple of 32 GB sticks of unbuffered ECC RAM for my 3950X. I'll move the standard DDR4 from that build into other computers.
 
biodoc said:
Linus Torvalds threadripper build. I can't believe he went with the cheap stuff!
"In an ideal world, Torvalds wanted 64GBs of error-correcting code (ECC) memory. He settled for a high-quality generic DDR RAM."

I ordered a couple of 32 GB sticks of unbuffered ECC RAM for my 3950X. I'll move the standard DDR4 from that build into other computers.
Click to expand...
That case is $451 ???? And 2666 memory ? and not ECC ?? and what HSF ?? They show the fan, but not the cooler.
 
StefanR5R said:
Related to the earlier peripheral discussion about ECC DRAM in this thread:

"The arguments against ECC were always complete and utter garbage. Now even the memory manufacturers are starting do do ECC internally because they finally owned up to the fact that they absolutely have to."​

says somebody who knows stuff. :-) Check the source for why this person advocates ECC for main memory. Hint: It's not just the often cited cosmic rays which flip your bits.
Click to expand...
I read the link the other day, but didn't seeing anything solid about why a home PC would benefit from ECC, did I miss it??
 
Picked up a 5800X from Microcenter, but then got a 5900X off of eBay (For to much$) and returned the 5800X. Coming from a 9900K, the upgrade is pretty big.
 
Assimilator1 said:
I read the link the other day, but didn't seeing anything solid about why a home PC would benefit from ECC, did I miss it??
Click to expand...
With Ryzen, the builder has a choice whether to use ECC memory or not although some motherboards may not support it: ECC Memory & AMD's Ryzen - A Deep Dive
According to the above article, you can induce bit flips by overclocking the RAM or using timings that are too tight. My lesson from reading that article is to do extensive testing on new RAM modules and if you overclock the RAM or adjust the timings, more testing is required if you're using the computer as a computational device (DC work, etc.).
 
Assimilator1 said:
I read the link the other day, but didn't seeing anything solid about why a home PC would benefit from ECC, did I miss it??
Click to expand...
A potential need for ECC RAM in a home computer is comparable to a potential need for a data backup strategy at home. Some people care, others don't.

RAM bit flips happen. They can't be made to go away by not looking for them.

BTW, since we are in the Distributed Computing forum: Most Distributed Computing projects actively mitigate RAM errors (and other errors) on contributors' computers by means of result validation. Therefore, ECC RAM is perhaps nice to have in a computer which is dedicated to Distributed Computing, but not really required. These mitigations are very expensive (e.g., as you know, many projects implement it by requiring the same result being computed twice, on different computers). But the projects couldn't omit these mitigations even in a world in which all donated computers had ECC RAM, because other error sources still need to be addressed (e.g. the worst case of falsified results).
 
I still like Intel's response to the defective Pentium 90's...."the defect only affects users doing a lot of math..." You know, like all the engineers, accountants, and other businesses that upgraded to them. I suppose they expected people to be buying the latest and greatest for dialing in to AOL, writing letters to grandma, and making their own holiday/greeting cards. 🙂
 
StefanR5R said:
A potential need for ECC RAM in a home computer is comparable to a potential need for a data backup strategy at home. Some people care, others don't.

RAM bit flips happen. They can't be made to go away by not looking for them.

BTW, since we are in the Distributed Computing forum: Most Distributed Computing projects actively mitigate RAM errors (and other errors) on contributors' computers by means of result validation. Therefore, ECC RAM is perhaps nice to have in a computer which is dedicated to Distributed Computing, but not really required. These mitigations are very expensive (e.g., as you know, many projects implement it by requiring the same result being computed twice, on different computers). But the projects couldn't omit these mitigations even in a world in which all donated computers had ECC RAM, because other error sources still need to be addressed (e.g. the worst case of falsified results).
Click to expand...
Yea, so it would be of no significant use to me.
JFYI, I back up photos, some videos and some other select data, but I don't bother with the OS.
 
One of the computers which I bought several months ago is occasionally making a fuzz about an alleged fan failure on a header to which a fan is connected, but where I did not connect the RPM sensor pin (because Supermicro's BIOS has a rather high threshold to detect fan failure). This made me go look at the system health logs of those computers of mine which have a BMC ... and discovered one ECC error logged on January 6 on one of these computers (a correctable error).

That is: Anecdotal evidence suggests that bit errors in RAM happen. If your computer does not have the hardware to check for (and if possible, correct) these errors — or if your computer has it, but you never look at the log — this does not mean that you do not have these errors. It only means that you do not know about these errors.

PS, I'll stop now since this is mostly off-topic in this thread. Just one more remark: Today's computers have hardware check sums (or better) everywhere — internal to the CPU, on the PCIe buses, in storage controllers, on the cables to storage devices, in storage devices including their storage media, on networks. Prominently absent: Check sums in RAM at the "client" computing tier. But RAM is subject to the same trends as CPUs and mass storage and networking: Higher density, higher speeds, lower power.
 
Last edited:
StefanR5R said:
That is: Anecdotal evidence suggests that bit errors in RAM happen. If your computer does not have the hardware to check for (and if possible, correct) these errors — or if your computer has it, but you never look at the log — this does not mean that you do not have these errors. It only means that you do not know about these errors.
Click to expand...
My Ivy Bridge 2P has ECC RAM so I installed edac-utils and rasdaemon to monitor for errors. So far, none have been detected.

Code: 
mark@x40-linux:~$ ras-mc-ctl --error-count
Label                             CE    UE
CPU_SrcID#0_Ha#0_Chan#3_DIMM#1    0    0
CPU_SrcID#0_Ha#0_Chan#0_DIMM#0    0    0
CPU_SrcID#0_Ha#0_Chan#1_DIMM#1    0    0
CPU_SrcID#1_Ha#0_Chan#3_DIMM#0    0    0
CPU_SrcID#1_Ha#0_Chan#1_DIMM#0    0    0
CPU_SrcID#0_Ha#0_Chan#2_DIMM#0    0    0
CPU_SrcID#0_Ha#0_Chan#0_DIMM#1    0    0
CPU_SrcID#0_Ha#0_Chan#3_DIMM#0    0    0
CPU_SrcID#1_Ha#0_Chan#2_DIMM#0    0    0
CPU_SrcID#0_Ha#0_Chan#1_DIMM#0    0    0
CPU_SrcID#1_Ha#0_Chan#0_DIMM#0    0    0
CPU_SrcID#0_Ha#0_Chan#2_DIMM#1    0    0
 
StefanR5R said:
That is: Anecdotal evidence suggests that bit errors in RAM happen. If your computer does not have the hardware to check for (and if possible, correct) these errors — or if your computer has it, but you never look at the log — this does not mean that you do not have these errors. It only means that you do not know about these errors.
Click to expand...
What are the effect of these errors?
 
So, what benchmarks do people want on the 5950x? I’m currently Running WCG 100% on 32 threads with PBO enabled (it was on by default), I’m getting a solid 4.45ghz, 53C, and water temp of 25c. I only have a partial day of results returned with WCG, but it’s at 198 for about 14 hours of run time.
 
What memory timings do these Zen3 chips max out at in coupled mode? Zen 2 was between 3600mhz and 3800mhz depending on your motherboard. I heard 4000mhz for the memory but does Zen3 eliminate the memory latency of Zen2?
 
Assimilator1 said:
What are the effect of these errors?
Click to expand...
I promised to let it go ;-) but since you asked: There is a variety of possible effects.
  • If the fault is in a memory range which is not read anymore after the fault happened, then there is no effect at all.
    Otherwise,
  • a user program could crash.
  • The operating system kernel could crash.
  • In an uncompressed data file, a hard to notice small artifact could be introduced.
  • In a compressed file, that is, a file with low redundancy, a large artifact could be introduced.
  • A numeric simulation with generally good convergence could end in a result which is very slightly off.
  • A numeric simulation which does not converge easily could end in a result which is far off.
 
Hans Gruber said:
What memory timings do these Zen3 chips max out at in coupled mode? Zen 2 was between 3600mhz and 3800mhz depending on your motherboard. I heard 4000mhz for the memory but does Zen3 eliminate the memory latency of Zen2?
Click to expand...
I believe this will answer latency questions better than I can:

AnandTech Forums: Technology, Hardware, Software, and Deals

Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
www.anandtech.com
 
This is way off topic(sort of, it has to do with ECC), but RAM errors can be fairly catastrophic. A year or 2 ago I had setup a FreeNAS machine and copied quite a bit of data to it. I was not using ECC RAM,. but more importantly one of my memory sticks was failing and I didn't realize it. It caused my Zpool to fail in a way that it could not be recovered. I worked on it for several days(because it was such a large amount of data, and enough that I had removed most of it from my backups so that I could use the drives in the FreeNAS machine) including trying to mount the Zpool on other OS's, etc. and had posted in some various ZFS forums and stuff, and it was 100% unrecoverable. There's even some ZFS options that allow data loss but are supposed to mount even a bad Zpool and none of them could actually get it to initialize the zpool and mount and read data properly. To be clear, I knew there was a chance I would lose the data since I didn't also have backups, but it was stuff that was more annoying to lose than catastrophic, so it wasn't a huge deal. I made the conscious decision that the cost of drives outweighed the risk of losing the data, and then I lost out on that risk 😛 Anything important I had/have backed up several places.

That's probably a worst case scenario, and I don't think it would have happened had my RAM been non-ecc but not going bad, but it does show how bad it can be. I would imagine a bit flipped here or there is a much smaller issue, but I still understand the need for ECC.(but I don't bother with it on most machines). ZFS is much more likely to have issues like this than most filesystems, though, since it does checksumming and other stuff behind the scenes, so don't be worried you're going to lose your normal file system because you're not using ECC. It's commonly recommended to use ECC with ZFS.

As far as benchmarks @Endgame124 I'd be interested in any projects that you want to run and post benchmarks on.
 
@Icecold

current plan is to get 3 days of results for WCG with PBO on, then 3 days with it off and a lower power target. I still have to figure out how to do all the eco mode stuff though.

Then I’ll switch to F@H for a few WUs just because it will be easy. Then I’m not sure, maybe rosetta?
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top