Recommendations for Reliable/Powerful PC Components

After recent failures and decades of being Mac-based I've decided to go back to a Windows machine, custom-build of course. I haven't built my own rig since the mid 2000's and things have changed enough that I'm going to be doing a LOT of research on components.

One of the things I've noticed over the years is that the majority of trouble tickets I've seen here on the forum have been PC-based. I'm guessing it's most likely from tweakers who are pushing the limits of the board/CPU by hyper-clocking and other non-standard setups, but still it's a pothole I want to avoid to make sure Resolve gets fed what it wants to be efficient and not chew up and spit out the hardware.

I'm talking way beyond hardware recommendations, I'm well aware what resources Resolve needs to be healthy, but specifically if there any chip-sets/CPU's/GPU's that are problematic in nature and should be avoided. So I figure with the plethora of PC builders here I'd ask the questions.

Right now I'm looking at building a system around this core:

- ASUS Rog Strix B850-F board
- AMD Ryzen 9 7900X
- 256GB RAM (not sure what brand yet)
- 1500 watt power supply
- Internal M.2 SSD'd. 1 dedicated to the OS and programs, others will be setup as RAID-0 for ultra-fast throughput for all the working files.

I don't yet know which GPU I should be looking at, still researching.

I'm currently looking at AMD based boards because they seem to have the best bang-for-the-buck when it comes to number of cores and threads. Intel still has the typical server-type CPU's but they seem outdated compared to AMD - and a LOT pricier. Maybe I haven't seen the latest stuff from Intel yet.

The two main purposes for the machine is first and foremost to be the primary edit box for Davinci. I'm NOT going to be doing any work in Fusion nor anything beyond 4K. In fact I'll be shooting/editing in 2K and delivery will be high bitrate 1080p.

Secondary but still important is now that I'm going back to a Windows machine I'm going to enjoy MS Flight Simulator again, which also pushes hardware to the limits when sliders at maxed out.

Any recommendations/warnings would be greatly appreciated.

What are you editing? Codec and colour depth?

1500 watt are you in Europe? In North America I'd want to make sure your outlets can handle that.

IIRC in the US 18 amps is standard for residential 18x110 is about 2kw. 1500 watt PSU you better not have anything else on the circuit.

Canada IIRC is only 15 amps .

SeaSonic makes good quality PSUs. Don't get blinded by watts. Go for quality. The higher end SeaSonic are also more efficient. Not just at full load but also at partial.

Your GPU will matter more than the CPU for the most part.

Should have added think about things like ports. If you're bringing in lots of big files USB 4 or TB will help.

If you're needing large storage you might want an external NAS. Think about network

- Splurge at least for a 9900X (or any of the X3D versions if you like lower power consumption but same-ish performance and you play Flight Simulator <-- extra worth it for this). The difference, especially for that game, is worth it. Lookup some benchmarks.

- 192GB is max you can get. 48GB sticks are max.

- Skip RAID 0, not needed for nvmes and video editing. I'd rather have files separate on disks, less prone to failure, and if you need more space seen as 1 drive you can always do JBOD instead of RAID.

Regarding the GPU, look at benchmarks that cater to your specific workload and ignore people that say "you need this and you need that". You wouldn't believe the amount of people that go "CUDA, what if need to program in that language someday". But they barely know how to install Resolve...

For example my use case: I work with RAW and ProRes / DNxHR 96% of the time. 3% other codecs from Drones usually (which I transcode to Prores anyway) and 1% AI tools but its extremely rare.

So, for my use case, the 7900XTX was the way to go, price wise, VRAM wise etc but if you look at puget benchmarks overall scores, it puts it near a 2080Ti which is ridiculous. However if you narrow it down to RAW and ProRes that's specific to my case, its a no brainer. For example this chart:

https://www.pugetsystems.com/wp-content ... Raw_DR.png


If I was doing Magic Masks and SuperScale 4X and etc on all my clips then yeah, getting an AI card like a 5090 would be better. Otherwise the price difference is too ridiculous.

My two cents.

I don't care what GPU brand you buy, but if you follow the unscientific sampling of this forum over the years, you may discover that the volume of operational complaints falls more to one than the other. If I recall, BMD tech support confirmed the very same on this forum, though at the time Dwaine said he might have an explanation for why. What happened after that, no idea.

For the rest, very few people are willing to comment on this issue, given the ferocity and contempt of the response frequently encountered here. So maybe take the promotion of one or the other with a measure of skepticism.

Be careful OP on engaging with trolls around here. I was warned by the moderators to not engage with people that have absolutely no interest in what you ask and what your posts are about but are just looking for toxicity.

OP: the post above is exactly what I'm talking about. Do your research here, and you'll see. Decide for yourself who's responsible for "toxic" posts -- if the question has any interest for you.

OP, you ever notice how none of these trolls even tries to be on topic? I'm guessing its because you chose an AMD CPU as your first choice and they are absolutely seething and foaming at the mouth because of that?

And not a single word about what you asked in the original post? Makes you wonder.

When you post a question like this, you will get a crap loads of suggestions. I built my system base on my research, need and my workflow and what I can afford. So far this method works well for me. I currently have 3 desktop PC, all works really well in Davinci Resolve. 2 system are Intel CPU and 1 system is AMD CPU. All of my system are overclocked and 16GB GPU card. My latest system is i9 285K, I absolutely love this combo.

When you post a question like this, you will get a crap loads of suggestions. I built my system base on my research, need and my workflow and what I can afford. So far this method works well for me. I currently have 3 desktop PC, all works really well in Davinci Resolve. 2 system are Intel CPU and 1 system is AMD CPU. All of my system are overclocked and 16GB GPU card. My latest system is i9 285K, I absolutely love this combo.

As a counter to 285K, that CPU is nice for editing, not so nice for Flight Simulator (for OP since he asked for this).

(some tests with MSFS, 285K vs 9800X3D).

I'm talking way beyond hardware recommendations ... specifically if there any chip-sets/CPU's/GPU's that are problematic in nature and should be avoided.

Hope this helps:

1. Take a look at components Puget Systems is using and their articles on annual reliability statistics and learn from that.

2. Be aware that AMD CPUs and GPUS can't do hardware acceleration of 10-bit 4:2:2, nor can all Intel CPUs and Nvidia GPUs (some can). Again lookup Puget's articles on that topic.

3. You will likely end up considering Nvidia GPU but be aware it is not completely headache free. Lookup articles on melting issues with 4090 and 5090 plus recent stability problems with drivers that support 50 series.

If it were me I would consider getting Intel CPU based machine to get 10-bit 4:2:2 acceleration by CPU and Nvidia 40 series card that is less finicky than 4090 / less "living on the cutting edge" than 50 series. Or just getting system from Puget Systems guided by their advice which means you would be getting peace of mind too knowing you would be getting their warranty and tech support too.

2. Be aware that AMD CPUs and GPUS can't do hardware acceleration of 10-bit 4:2:2, nor can all Intel CPUs and Nvidia GPUs (some can). Again lookup Puget's articles on that topic.

Slight correction, H264 and H265 only.

10bit 4:2:2 like ProRes and DNxHR are different and work on basically anything.

If you value your work, apprecite highest stability and peace of mind, then don't build the system yourself. Instead, custome configure a workstation from reputable brands such as Boxx, Dell, HP, Puget Systems, Lenovo, ProArt and so on. Such companies have the needed resources to fully test their systems for optimal performance and stability, not to metion thier warranties.

Alternatively, you can build it yourself, and if you are lucky, everything should be fine

Like you, I used to build my own computers, but becouse of previous experinces, I only depends on HP and ProArt workstations for work.

But since I am planing to retire within the next few years, work is not priorty anymore, so I plan to bulid my dream machine this year. I have been doing my reaserch, watching what systerm integrators say and do, mainly Puget Systems, ProArt System Integrators and Maingear. So far I am leaning toward bundling a workstation based on ProArt case and components. I just fell in love with the ProArt PA602 Wood Edition Tempered Glass:
https://www.asus.com/eg-en/motherboards ... ass-panel/

Also check this video:

If you value your work, apprecite highest stability and peace of mind, then don't build the system yourself. Instead, go with with custome configration workstation from reputable brands such as Boxx, Dell, HP, Puget Systems, Lenovo, ProArt and so on. Such companies have the needed resources to fully test their systems for optimal performance and stability, not to metion thier warranties.

I disagree. Some of these companies use proprietary hardware and connectors and are vendor locking CPUs, GPUs and etc.

Here's a video about Lenovo :

@Yasser Saeed. You know that your bold statement here is none sense. I custom built all of my system and Overclocked all of my system. I never have had instability problems. Do you know why, because I properly run stability test to find the maximum clock can be achieved without crashing any programs. And do you know why you don't see I post any instability and problems with DR because I have none.

I disagree. Some of these companies use proprietary hardware and connectors and are vendor locking CPUs, GPUs and etc.

Here's a video about Lenovo :

Now days, it is rare for those companies to use proprietary hardware for workstations and use vendor locked componats and I think Lenovo stopped doing that, but for sure the other brands I mentioned do not do that so you can easily upgarde the most important components like CPU, RAM, GPU, storage .. etc.

Even my 10 years old HP AIO Z1 workstaion was upgradble to a degree, and I did upgrad its RAM, GPU and storage. I also upgraded the RAM, GPU and storage for my old HP Z2 Mini G5 workstation, and soon will also be upgrading the GPU for my current HP 2 Mini G9. Keep in mind that those are not even in a standard form factor, those are AIO and SFF and use proprietary motherboards, yet still upgradble by user. Even the CPU is upgradable on the SFF workstations.

However if you go with thier standard form factor desktops (full tower, mini tower and micro tower) then they are as upgradble as any system you build.

@Yasser Saeed. You know that your bold statement here is none sense. I custom built all of my system and Overclocked all of my system. I never have had instability problems. Do you know why, because I properly run stability test to find the maximum clock can be achieved without crashing any programs. And do you know why you don't see I post any instability and problems with DR because I have none.

Sorry, but your stability test is joke comparing to the real stability tests done by big companies. They have specialzed engineers that test every component for hundreds of hours using state of the art equipments which costs hundreds of thuosands, if not millions.

The stability test you do is a very common test that any end-user can do, however we cannot do what the big boys can in becouse of the huge resource they have and no end user can afford.

Now, it is not necessarily that you will face stability issues when you self-build your system, especially if you are extra careful in selecting the right compononts. However, the risk of components conflicts and stability issues in self-built computers are much higher than in those machines build by specialized companies. Why? Becouse of the extensive R&D and testings those company do with specialized equipment. Period.

I think I have repeated myself so many times in this forum trying to help others, but I am realy getting sick of it becouse of useless debates So take my advice or leave it. No body is forcing anybody.

I am done with such topic and will not share my experience anymore.

Oh, so my stability test is a joke ah? So, who the F are you to tell me it's a joke? My system performs and finished every projects without any issues. To my experience it doesn't look like a joke to me. In fact, I overclock every PC I built and it is none issue with stability. That because you had bad experience with your own stupidity or lack of the known how and now you telling everybody it is bad to OC? My AMD 1950X threadripper 16 cores Overclocked 4ghz since day one on a custom built watercooling and it is still running as good as day one. My i7 12700k Overclocked 5.1ghz on a custom built watercooling since day one. None of these PC have any instability issues. So, don't f'ing tell me its a joke. You are a joke.

What are you editing? Codec and colour depth?

1500 watt are you in Europe? In North America I'd want to make sure your outlets can handle that.

IIRC in the US 18 amps is standard for residential 18x110 is about 2kw. 1500 watt PSU you better not have anything else on the circuit.

Canada IIRC is only 15 amps .

SeaSonic makes good quality PSUs. Don't get blinded by watts. Go for quality. The higher end SeaSonic are also more efficient. Not just at full load but also at partial.

Your GPU will matter more than the CPU for the most part.

Codec and color depth are irrelevant and have no impact on performance. Just bitrate and frame size.

Should have added think about things like ports. If you're bringing in lots of big files USB 4 or TB will help.

If you're needing large storage you might want an external NAS. Think about network

Ports are irrelevant. All storage both for OS and working files are internal on M.2 SSD's.

- Splurge at least for a 9900X (or any of the X3D versions if you like lower power consumption but same-ish performance and you play Flight Simulator <-- extra worth it for this). The difference, especially for that game, is worth it. Lookup some benchmarks.

- 192GB is max you can get. 48GB sticks are max.

- Skip RAID 0, not needed for nvmes and video editing. I'd rather have files separate on disks, less prone to failure, and if you need more space seen as 1 drive you can always do JBOD instead of RAID.

Regarding the GPU, look at benchmarks that cater to your specific workload and ignore people that say "you need this and you need that". You wouldn't believe the amount of people that go "CUDA, what if need to program in that language someday". But they barely know how to install Resolve...

For example my use case: I work with RAW and ProRes / DNxHR 96% of the time. 3% other codecs from Drones usually (which I transcode to Prores anyway) and 1% AI tools but its extremely rare.

So, for my use case, the 7900XTX was the way to go, price wise, VRAM wise etc but if you look at puget benchmarks overall scores, it puts it near a 2080Ti which is ridiculous. However if you narrow it down to RAW and ProRes that's specific to my case, its a no brainer. For example this chart:

https://www.pugetsystems.com/wp-content ... Raw_DR.png


If I was doing Magic Masks and SuperScale 4X and etc on all my clips then yeah, getting an AI card like a 5090 would be better. Otherwise the price difference is too ridiculous.

My two cents.

That's the best 2 cents I've gotten on this thread. Thanks.

Oh, so my stability test is a joke ah? So, who the F are you to tell me it's a joke? My system performs and finished every projects without any issues. To my experience it doesn't look like a joke to me. In fact, I overclock every PC I built and it is none issue with stability. That because you had bad experience with your own stupidity or lack of the known how and now you telling everybody it is bad to OC? My AMD 1950X threadripper 16 cores Overclocked 4ghz since day one on a custom built watercooling and it is still running as good as day one. My i7 12700k Overclocked 5.1ghz on a custom built watercooling since day one. None of these PC have any instability issues. So, don't f'ing tell me its a joke. You are a joke.

Listen guys, if you're going to hijack this thread into another piss contest I'll just have a moderator pull the thread. With one exception none of you have anything worthy to input, so calm the hell down.

Capiche!!??

Hi,
I am skipping back to the start of the thread, because I understood it got in a "holy war" kind of situation (e.g. the faithful of AMD vs faithful of NVIDIA, .
Long time IT with some 3D graphics, video and Da Vinci (yes, I know the software uses different, but I write it the Leonardo way, and I don't mean the Ninja Turtle ) use here.

Never like the "holy wars", for IT at home I go with what I like, can get or can afford , for work, it is a matter of analysing what the usage scenarios, though the can get, can afford and other elements like support also matter (not that support does not matter in private life, but there, for computer stuff I often ended up getting also some second hand components, which may not have any support).

Ah, notice I am considering here Da Vinci Studio, because is the one I use (got cameras and a device from BMD).

---- AMD vs Intel.
In old times, AMD tended to run a bit too hot, but should not have that problem anymore, on the other hand, there may still be occasionally software that runs better on Intel than AMD, but that depends from the specific software, and for Da Vinci, as far as I know that is not a problem, in principle.
Some AMD CPUs can support more PCIe lanes that some Intel CPU of the same class, which can impact the quantity of PCIe slots and their performance, but that depends also on the motherboard, so, you should check that in combination, if you really want to fully optmise. Otherwise, you can just keep it to "CPU and motherboard compatible ? Good" approach .

---- RAM.
Put as much RAM as you can (afford) and your motherboard supports, faster is better, but in reality, it is better more RAM even if a bit slower, that less RAM faster. The nanoseconds for a few MHz don't count that much, the GB do.
Yes, I know there is the GPU VRAM, but without entering in details, either you always work with something that fits in the VRAM and do operations that run exclusively on the VRAM, or RAM is important for video and graphic editing (especially if you do not plan to run only one applicaiton at a time )

--- Disks.
Depending on the money and support from the motherboard (and don't forget power and space in the case) M.2 for speed, SSD for intermediate storage speed, SSHD for a bit slower, then maybe something slower bigger capacity for backup and the like (e.g. external, though with USB3, especially 3.2, "slower" can be relative).
Of course, you can also go for a 10Gb NAS and stuff like that, or even directly for Infiniband, but that would be expensively, complex, and honestly, I knew even professional that did not have that kind of setup (if you don't know what is Infiniband, special networking, older models, 40-56Gb, newere, 100Gb).
It depends again from what you can get, can afford, and like.
M.2 with a couple of TB of space are not hugely expensive nowadays, and if you have a bit more money, even 4 TB can be reachable - but I am referring to the disk to use for Da Vinci.
If you want to also boot from M.2, fine, but keep that separate from the Da Vinci disk.
It is an old trick of doing video editing on computers, even more useful nowadays that everything in theory is multithreaded and multicore (to be seen how good are the OS and the motherboard then in taking advantage of that, but that, is a different matter).

--- GPU.
Bigger contendents, AMD and NVIDIA.
AMD corrected a bit directions in latest models, but had some years ago bet big on the blockchain/graphic/gaming side, so in general if you go for that direction, it can be a good choice, being in theory a bit cheaper. NVDIA went already many years for another direction, computing/graphics/gaming, then adding up machine learning/"deep" learning (ML/DL, what is generally sold as "AI") as a kind of evolution/special focus of the "computing" element.
E.g. in blockchain, the energy cost per income of the AMDs some years ago was better than NVIDIA (never been in blockchain, looked for curiosity some year ago).
As you can see, there is an overlap, the "graphics/gaming" part, because they both tried to focus also on that (well, gaming is a big market), in the case of NVIDIA, even integrating/leveraging to some extent its ML/DL hardware (if the software uses it, but Da Vinci does).
So, which is better, depends on what you want to do.

If you plan to use the "AI" features in Da Vinci Studio, then NVIDIA may be the better bet, depending also on the model (see below) - NVIDIA focus on computation, then "converted" to hardware for ML/DL, goes all the way to the Tesla some 14 years ago, cards that did not even have a graphic output, because were only for calculation, and the "Maximus" configuration (one graphic card for calculation and output, and one headless card purely for computation).
If it detects a NVIDIA card with CUDA, Da Vinci Studio will optimise its ML/DL ("AI") functionalities for it.

About CUDA, true, most people that talk about CUDA never had programmed and may never program using it directly, but from point of view of programming, it is not necessary to program directly in C++ (CUDA is an architecture, but in practice the CUDA Libraries available for free are used with C++, though there is also a product with FORTRAN available) to use CUDA, e.g. things like tensorflow, and python can use CUDA acceleration in a transparent way thanks to libraries.
More importantly, a lot of graphic and video software, even free open source (e.g. ffmpeg, Blender, although Blender can also use OptiX on newer NVIDIA cards, and it also support specific AMD acceleration) can use CUDA acceleration without any need for programming, and using CUDA exploits to the maximum the hardware of the NVIDIA card.

The most diffused alternative to CUDA is OpenCL.
Advantage, it is an open standard, and works with CPU, all kind of GPUs (AMD, NVIDIA, Intel), and in theory it supports out of the box even multi-CPU, multi-GPU and even mixing (multi-)CPU and (multi-)GPU.
Disadvantages, aside from university experiments, multi-GPU support require the application to be changed and specific libraries to be used, so it is difficult to find a commercial application that uses it out of the box.
That is why people talk a lot about CUDA, squeezing the maximumo out of the NVIDIA GPU - though notice though that CUDA is proprietary to NVIDIA, it will be useless (should not even install) if you do not have NVIDIA cards.

I do not know if having two AMD GPUs, Da Vinci will use both of them when encoding/decoding, for NVIDIA, I know it will, though they should be same generation or at least either both post-RTX or both before-RTX, I saw it practically when I had two of the same in the computer, and I strongly advice you get anyway at least two GPUs.
Again, ideally same generation, but it is not mandatory. Most people will tell you to get the latest three slots card RTX4090, RTX5070 (well, forecast, not yet there), something like that. My advice, don't. It is better to have two cards, even if older (professional, the old NVIDIA Quadro line) than to have one.
Reason… If you have two cards of the same generation, both can contribute to encoding/decoding etc., it does not halves the time, but anyway greatly reduces it.
But even if you have one that is RTX and one older, you can use the older one for the video, and the newer one will be fully dedicated by Da Vinci for its operations (if it sees an RTX card, it will go for it and ignore the other, or at least, that is what I have seen), thus being anyway faster that if you had one card that has to do everything.
Can also try to have two RTX and one older one, I had that, but aside the potential cost (well, second hand is what I do... ), at lest the model of Quadro RTX I have seems to tend to go rapidly "hot".
It is in part by design (I checked the specs), but still, two of those crammed with a third card - though it may also be because the ones I realised after the ones I got had been used and abused a bit - and going for a long time, wasn't feeling too sure, so I modified the configuration.
Of course it is not 100% said you would have the problem, and if the case is big and you feel "adventurous", you could go for water colling and stuff like that - in my case I use a commercial workstation.

--- Note about power (not really for you, for other readers).
In Europe it is not unusual to have at home contracts with 3-3.5 KWh, and if you look at the specs of computers like e.g. HP workstation (so, American company), you will see that they power adapter can be rated (depending on the model) 1450W, but that in Europe in reality gives to 1700W.
Do not forget in Europe electricity is 220-240V, hence the increase, basic physics formula, W=V*A, if A stays the same, increased V means increased W, though it is limited by other factors, so it does not go twice as much, despite the jump from 110-120V to 220-240V.
Also, but that does not have the same interest for the wattage of the PSU, a frequency of 50HZ (that has an interest for you in video, it turns out it is the reason PAL is 25fps, not 30fps, and does not need to adapt to something like 24.97fps, while you need the 29.97fps in USA).

Hope this is useful