ZBrushCentral

What is the performance impact of RAM Speed / Frequency?

ZBrush Usage Questions
#1

Been googling and searching but didn’t find anything substantially solid…

Just to be very clear… I AM ASKING SPECIFICALLY ABOUT RAM SPEED / FREQUENCY / TIMINGS and if any1 has experience testing this out

I UNDERSTAND the importance of RAM AMOUNT and CPU and STORAGE

I have a 2970wx, 64GB of ram (looking to upgrade to 128 anyways), and a NVME M.2 SSD

I understand amount of RAM is important… but does speed/frequency matter all that much? Especially with high-definition work involving like a single 50M+ polygon subtool with layers?

I have to replace my 8GB modules with 16GB modules… and so does RAM frequency make a big difference?

I was going to upgrade to 3200mhz… but if it makes a big diff, I’ll shell out for 3600mhz if not more

And again – I’m asking SPECIFICALLY ABOUT RAM frequency / timings impacts on performance

Like if any1 has experiences testing this out or has empirical data

no1 has said anything about RAM SPEED and RAM SPEED can improve videogame FPS by a noticeable amount… so I’m hoping MAYBE the same could be said here for ZBRUSH

Thank you for your time

1 Like
#2

Hi,

the amount of RAM has in general nothing to do with the speed, except you reach the limit of your RAM and the system begins to swap your data to the hard disk, then it will be much more slower.
Compare the RAM size with fuel for a car.
You can’t drive faster with an bigger gas tank but you can drive farther.
In other words you can handle more polygons with more RAM but not faster.
Most relevant for the speed of your car is the engine, isn’t it?
For the PC it’s the CPU.
Note: It’s also depending if the software can handle multicores or only singlecore.
E.g. There are some functions in my Cinema 4D which work only with one CPU core
Not sure if all functions of ZBrush works with multicore.
CU.

1 Like
#4

I have 64GB of ram upgrading to 128, so yes I get the size-limitation and how that works, but does ram speed along with the cpu speed as well as storage device speed, for example affect how quickly the fuel can be transferred from the gas-tank to the engine…

(I’m trying to continue the analogy, but it doesn’t really work with what I’m asking for)

That’s why I used the “50M+ polygon subtool with layers” example

If you’ve ever worked on a 50M+ polygon subtool, with layers then you know just the simple process of turning on and off the visibility and recordings of layers has a noticeable lag-time

I have a 24core cpu and a nvme m.2 ssd – I already know that those affect speed… but was wondering SPECIFICALLY how RAM Speed / frequency would affect ZBrush performance…

particularly with things like the “50M+ polygon subtool with layers” example

even SECONDS in difference would be AMAZING

and yes, just how some functions are multi-threaded and some functions aren’t… I was wondering if stuff like the “50M+ polygon subtool with layers” example WOULD benefit from increased RAM speed / frequency… SUCH AS being able to ACCESS the cached data more quickly for example I hope

just like how in some games, RAM frequency can boost FPS by 10-20+

The recommended specs I understand are guidelines, but I am interested in OPTIMIZATION

I was hoping someone has tested this out to confirm anything

#5

Hi @scoobybubu !

The official Pixologic Recommended System Specs have a section at the bottom about how to optimize a computer for Zbrush. If memory speed was a significant factor, they would have included it here. They are far more concerned with you having enough RAM, rather than faster RAM.

So while there might be some nominal benefit to having faster RAM, the answer to your question is that they are not especially important for the purpose of Zbrush.

#6

I appreciate the response, and am aware of that section as well as the bottom… but just as with the recommended spec sheets for videogames how they never mention the importance of RAM speed / frequency / timings… --------> youtube videos of people actually testing this out have gained 10, 20, 30, sometimes near/around/above 40 FPS improvement in modern games simply by improving their RAM speed / frequency / timings (and this would benefit greatly those who play competitively, especially in First-Person-Shooters, and who have invested in higher framerate monitors to take advantage of this)

so because it is not mentioned, how does any1 really know how nominal or greater the benefit is

especially with very high-definition workflows such as working with single subtools consisting of 50M+ polygon WITH layers

because even saving seconds switching between layers and recording I’d say would be worth it as this point

(when it’s only a few million poly’s… it’s almost instantaneous and not a big deal – and to continue with the videogame analogy… if you’re not playing competitively and not doing First-Person-Shooters WHERE THOSE SECONDS MATTER, and don’t have a 120 or 144hz+ monitor, then yeah it won’t really matter)

and so I was hoping if someone had tested this… but I guess I’ll just find out when I upgrade

#7

Crunching polygons in Zbrush is generally the function of the CPU, and will respond to differences there far more than any other factor. Even a small upgrade here would do more than the fastest RAM in the world.

Good luck with your testing.

#8

Thank you for your response

For you or anyone in the future:

I didn’t want to get into my specs or really explain the situation because I just thought maybe someone knows how much impact ram speed has on performance and could give their experiences but…

I understand CPU is important… which is why I got a 2970wx (And I went for more threads instead of less threads+morepower just in case I have to use the CPU for rendering like for Arnold or Vray CPU+GPU or maybe later try out Renderman’s upcoming CPU+GPU architecture too)

So I’m not THAT much of a noob. I’m trying to understand how much impact RAM SPEED has just to have a better picture

And I only have 64GB of ram… so I’d rather upgrade my RAM to 128 than shell out like $1500 for the 2990wx or wait for the 3rd gen threadripper or switch to the new Ryzen series and have to reinstall everything

I can’t seem to find any empirical data on RAM speed/ frequency / timing impacts on performance in ZBrush and since I have to upgrade my RAM anyways… I was wondering if any1 actually had tested this and just not spoken about it yet so I can be excited or not for a boost in ZBrush performance… and whether or not I should spend more for like a 3600 MHz than a 3200mhz or a 2600 MHz

I’m aware RAM Amount would improve how many polygons I can handle… and I’m aware CPU’s are great overall

I’m asking SPECIFICALLY about RAM speed frequency timings

And just like how I would have never guessed how much impact Ram Speed frequency timings could noticeably improve FPS in videogames… perhaps the same could be said here in ZBrush

Because no1 has said anything about it… and just like with videogames… just because the RAM SPEED is not specified in the spec sheets doesn’t mean it won’t improve performance by a noticeable amount

– just clarifying

And also… are you guys not curious as to how much performance impact it could make? Your answers sound more like theoretical guesses

If it made a 5% improvement?! WOAH… you wouldn’t want to know that? maybe it’s 10%, 15%…

So the next time you upgrade your parts or buy a new computer you can keep in mind:… “oh, well it’s a $50 difference between 2600mhz and 3000mhz… I’m down to spend $50 more for a 5% improvement” or “maybe i’ll get the new GDDR5 instead of getting the best of the last gen GDDR4”

I WANT TO KNOW STUFF LIKE THAT :]

I LOVE 3D VFX… I LOVE knowing those nuances

1 Like
#9

No one thinks you’re a noob, @scoobybubu!

However there are two users who have given you basically the same answer, backed up by official system recommendation that choose not to emphasize the importance of RAM timing, despite a financial incentive to do that if it were the case.

You appear to be disregarding these, which is your prerogative, but it’s also probably the end of what we can do for you here. Good luck with your own testing, and by all means report what you find.

#10

Well… just like how Maya is still buggy… or how you can’t even change the name of an XGEN description in the outliner without ruining things… or how the developers of MD removed the option to disable/enable auto-sync which is very workflow debilitating… or how the marketers of MD didn’t specify the limitations of not being able to draft layered clothing with the newly added GPU simulation feature… or just how Substance Painter still doesn’t have full UDIM support… or just like how Windows 10 Search for a long-time couldn’t provide accurate search results… or how even though there’s been GPU renderers for quite some time now – why it’s taken Arnold this long to start getting into it

There’s arguably financial incentives for these companies to [have] fix[ed] all these issues…

But I don’t expect companies, no matter how big, to be perfect :]

And I’d rather lean more on the side of being curious and inquisitive especially when there is no empirical data present (or even anecdotal), and as mentioned before with videogames, Official Game Spec Sheets From Developer Websites have zero mention of Ram Speed / Timing / Frequencies… and yet you can improve your FPS by 10, 20, 30 40+ BY improving your Ram Speed / Timing / Frequencies… which can give significant advantage to higher-end applications such as competitive e-sports

And yes it is also your prerogative to disregard the possibility of getting the most bang out of your ZBrush and you seem satisfied with that which is fine. We are coming at this from two different points of views and desires. But if you don’t know, you don’t have to say anything. This thread has been viewed multiple times - it’s hard to say by how many people exactly - but it’s likely they either don’t know too or don’t want to divulge anything, or maybe they clicked by accident or just don’t care.

But anyways I understand you are trying to help which is appreciated.

But for me, it is too inconclusive and there are too many reasons for me to stay curious :] also since I want to know how I can improve my workflow as much as I can and I have to upgrade for the long-term anyways

And just because you and Harrybee don’t know, doesn’t mean someone else does not. After all, 2 people is not exactly a large pool. Nor have you two mentioned anything like “yeah I overclocked my RAM and I didn’t notice a difference, even working with extremely high subdivisions” or anything of the sort

But yes, thank you I shall test away ^^

#11

ZBrush uses RAM to store the data that it expects to need handy. Everything else goes into virtual memory.

The amount of RAM will affect how much data ZBrush can hold without needing to write to virtual memory. This means that the more RAM you have, the less ZBrush will have to write to virtual memory.

The speed of your RAM minimally affects how quickly data can be written to or read from virtual memory. The reality, though, is that your disk’s data transfer speed is probably always going to be slower than your RAM speed, so the actual hangup will be with the disk rather than RAM that is too slow.

The other place that RAM speed can affect performance is of course when data has to be transferred between the RAM and the CPU. Faster RAM in theory allows this to happen more quickly. But then you have to take the bus into consideration. That is the pathway that the data takes when traveling between the CPU and the RAM, and is a function of the motherboard. So of course you need to ensure that the bus is capable of effectively using the data speed that the RAM is capable of.

AND your CPU itself has a data cache on it that basically acts as a small amount of super-fast RAM. The larger this cache is, the more effectively the CPU can crunch the numbers.

As you can see, it quickly starts to get murky.

You also have to consider that the maximum SubTool size in ZBrush is 100 million polygons, and even that is only accessible by temporarily boosting the MaxPolysPerMesh setting. In practice, your models are typically going to be topping out around 25-50 million polygons, with the occasional exception going higher. After all, a 4K map is only capable of holding a maximum of 16 million polygons’ worth of data. An 8K map (the absolute largest that ZBrush can create) is 64 million and a good 25% of that is probably wasted space in between the UV islands.

In my experience, RAM speed produces diminishing returns, with you getting significantly less useful bang for your buck the faster the RAM gets or the more you have. And generally it is more cost effective to add more RAM than to increase your RAM speed – the slower RAM is less expensive per GB, so doubling your existing RAM is going to cost less than increasing the speed of your RAM. To say nothing of both doubling it and increasing the speed! But while going from 4 GB to 8 GB shows a significant performance boost, going from 8 to 16 sees less and from 16 to 32 sees less still, etc. At 64 you’re already about as good as you’re going to get with ZBrush.

The 64 GB in your system is already plenty sufficient for working well with the number of polygons that you’re dealing with. Neither faster RAM nor a larger amount of RAM is likely to produce a discernible difference in performance. What would increase your performance would simply be to slightly change the way you work. If you start seeing performance slowing down, either break your model up into more SubTools or hide the parts of the current SubTool that you’re not currently actively sculpting. That data will then get offloaded to virtual memory and your performance on what remains visible will increase significantly.

4 Likes
#12

@aurick THANK YOU SO MUCH!!!

I wasn’t even thinking about the texture map res capacity… those numbers REALLY PUT THINGS INTO PERSPECTIVE. Those numbers, along with your recommendation to change my inefficient workflow, validate on a technical level my “intuitive guess” to revamp my basemeshes and how I plan out and handle my subdivisions and subtools

Like I said, I LOVE knowing about these nuances so I really do appreciate how you’ve described how these factors are connected

I’ve also decided to go with the more budget friendly and slower-speed upgrade to 128GB… and chose to continue with the 128GB storage upgrade due to often needing to have multiple programs up and running at the same time on different screens… and as you’ve described more precisely, which prompted me to think about it more – I think this has caused ZBrush to start reading/writing to/from the disk-part of virtual-memory, especially when working at the higher subdivisions / polycounts for detail-work, and thus not taking advantage of that more direct CPU-to-RAM data-transfer speed + MOBO bus capacity.

Again, I APPRECIATE you taking the time to explain this more in detail!

MUCHAS GRACIAS @aurick!