Advice on new workstation build - Ryzen based

[springate]

I see now! Thank you! Guess that's it i'm gonna go for the x370 asus prime PRO and 32GB @ 2400 Kingston HyperX.

[Nejc Kilar]

@nkilar

That is interesting, I didn´t know that ECC RAM is supported. I thought this might be another benefit of the Thread Ripper / Naples chipset.
Which RAM / manufacturer did you buy?

Thanks

Actually I should have been more clearer. I bought the ECC ram for my Xeon workstation and similarly like with Ryzen its pretty nice to check if that exact RAM module is supported. I mean I presume just about any chip will work (probably even more true for the Ryzen build) but why go through the hassle if you can check what works and not.

Other than that Ryzen by itself does support ECC RAM but a) the motherboard needs to support it (not sure if there are any out there?) b) do you even need it? :)

[springate]

I have another question, folks 😀 is gpu needed for rt render?
I know its not about the main question of this topic but since we're sharing info on workstations i assume i'm not totally wrong asking here...

[maru]

I have another question, folks 😀 is gpu needed for rt render?

Nope, Corona is 100% CPU-based.

[Fritzlachatte]

@nkilar

thanks for your answer. To be honest, I don´t actually know, if those advantages of ECC-Ram are that meaningful for rendering purposes. Sometimes I have in mind, used ECC might sometimes be a good low priced alternative, but I am not so enlightened of the risks using used ECC-Ram.
I can imagine, AMDs Threadripper forces the user into using ECC-Ram but maybe they keep it as an optional choice. Tomorrow we will know more. 

[Nejc Kilar]

You are most welcome! Whats better than another opinion huh?! :)

Jokes aside, afaik there are pretty much zero improvements for rendering with an ECC RAM. They are generally a lot more stable but a little less speedy.

I do value stability and since with the CPU I have I had to go the ECC route I am happy to have it but I would be perfectly fine without. My previous workstation, that my significant other now has, didn't have ECC RAM and I don't ever remember that machine crashing.

Would be fun to hear what other people think, of course :)

[Jann]

Honestly just go with what's cheaper. Me and a friend have similar X99 based Xeon systems. I use ecc, he uses normal ram. Only reason for ecc was the better price at the time. I also used normal ddr3 on my previous dual xeon v1 box.

So just check motherboard compatibility lists, or if the ram you want isn't there, find if someone has it running with that setup. Amount > speed > ecc or not.
And for ddr4, get 16Gb on single stick, even if it means starting with just one.

[Ryuu]

Well the ECC RAM does not affect only the stability of the OS itself. The effect of a memory corruption could be just a single application crashing, data written on HDD/SSD being corrupted or (most probably) just randomly flipping a single bit in an image which is loaded in memory just for the purpose of being displayed (e.g. background image of a web page you are currently viewing).

[Fritzlachatte]

This exactly my problem these days, I would like to avoid with my next system. Yesterday Max started crashing, later on PS and now Firefox etc. bluescreens all over. A few weeks ago I changed some RAM because of these symptoms(but by far less dramatic) and now it is the same again. I don´t know if 4 years is an age for a PC (Intel 4770k with Corsair Vengeance) but I am a little bit disappointed.   
I expected AMD would tell us Threadripper comes in June but they just proclaimed "this sommer", which ends in September, by far too late. I feel uncertain buying a Ryzen now, I think 64GB is the same deadend like my 32GB was, when I bought the crashing PC 4 years ago. (btw. can "overfilling" with large Max-files damage the RAM, because it started after loading high detailed models and adding lot of polygons, then crashing and so the story goes on.)   

[Jann]

You should really stress test any new parts/system.
Same with your current crashing system. Stress test the ram. Every vendor has faulty sticks sometimes.

[Ryuu]

Unless you're in some kind of heavy radiation environment, I doubt that ECC RAM would help in your case. This seems more like a corrupted system or malfunctioning hardware.

[denisgo22]

This exactly my problem these days, I would like to avoid with my next system. Yesterday Max started crashing, later on PS and now Firefox etc. bluescreens all over. A few weeks ago I changed some RAM because of these symptoms(but by far less dramatic) and now it is the same again. I don´t know if 4 years is an age for a PC (Intel 4770k with Corsair Vengeance) but I am a little bit disappointed.   
I expected AMD would tell us Threadripper comes in June but they just proclaimed "this sommer", which ends in September, by far too late. I feel uncertain buying a Ryzen now, I think 64GB is the same deadend like my 32GB was, when I bought the crashing PC 4 years ago. (btw. can "overfilling" with large Max-files damage the RAM, because it started after loading high detailed models and adding lot of polygons, then crashing and so the story goes on.)

Why this is not a graphic card problem?:)

[Fritzlachatte]

I was just wondering why it came suddenly from one second to another while adding "details". Max doesn´t start anymore and firefox crashes the pc while openening. Chrome/Edge work well.And using it for simple things is stable. I would say it is somewhere in RAM + SSD or HDD. :-O
I use a Titan (1) but it never had hard work, because I am never gaming and Max/Corona shouldn´t torture the card that much but you might be right anyway and I have to check it and thanks for your hint.
But the fact that Ryuu mentioned, ECC could not prevent those sneaky errors is astonishing me (in a bad way) because there was a hope of limiting the risk of dataloss/damage in future systems.

 

[Ryuu]

I'm not saying that your problems couldn't be caused by data corruption in RAM, just that you would have to be extremely "lucky" for that having such a catastrophic effect :)

Most of RAM corruption is caused by high energy particles coming from space (unless you have a ball of plutonium sitting on your desk). These particles impact RAM cells and randomly switch value of a single bit. The resulting effects vary greatly depending on what is stored in that memory:

• it could be completely unused
• it could be an image displayed in web brower
• it could be frame buffer of a renderer - the effect can vary from hardly noticable like changing color from (255,255,255) to (255,255,254) to something more severe like changing float color from (1,1,1) to (1, 1, 8.5 * 10e37)
• changing a single bit in executable code will change a single instruction - if you are lucky, this will just cause immediate crash (either of the whole system or just a single application). If you are unlucky this will result in processed data being corrupted
• whenever a file is written to disk modern OS usually buffers the written data in memory first (so that the application can continue immediately assuming the data is written) and then actually writes the data to disk for some longer time period, possibly pausing when there are other disk requests with higher priority - so memory corruption can actually "change data on disk". Again if you are lucky, this just corrupts a single file, if you are unlucky this corrupts the filesystem itself. AFAIK most modern filesystems try to prevent this with various ECC schemes.

If you think that an event such as "cosmic ray particle hitting my RAM" is extremely rare then think again: http://www.cs.toronto.edu/~bianca/papers/sigmetrics09.pdf

I seriously hope that I'm interpreting the numbers from the study in a wrong way otherwise it would mean that in 32 GB RAM you would get 6-20 bit errors per hour.

tl;dr: Star that exploded thousand years ago can seriously screw up your render and/or even system. You can lower the chance of that happening by using ECC RAM.