Intel i9 or Ryzen?

I’m in the middle of building a new desktop and am torn between either a Intel i9 series or Ryzen. In the past Ive had Intels after getting burned with dual (whoa) AMD Opterons in 2000’s. I started my build with the i-12900 (non-k) and I discovered these things run hot when specifically rendering. I had to redo my whole build as initially it was ITX format and am now trying to move to ATX with better cooling. But its got me torn and considering Ryzens since they run a lower watt for now and it seems its slightly faster in Vray. I do plan to upgrade to which ever chip in the fall and it seems Ryzens wont have the low wattage like they do now. They are bumping them up a lot in the 7000 series it looks like. Also it seems it will max out at 16 core where as i think Intels Raptor will have 20-24.

Thoughts anyone?

Hi @sean_brown !

I have a similar personal experience with 12900k, I’m not impressed by this CPU/platform overall. It is fairly new and there are many issues.
I own a 5950X as well, it is faster in rendering, runs cooler and a very stable platform with hardly any issues in production.

I advice people tovoid Intel 12th generation, here are the issues I faced
>> Runs very hot, depending on the motherboard. For the advertised performance, you will be at 100 degrees with a Noctua NH D15 which is not normal
>> You cannot run 128 GB of RAM reliably with it, I tried myself and I pulled my hairs out. I used very expensive memory and an expensive ROG board
>> You will need to use windows 11 for Intel 12th generation, there is no way around this
>> Many of the new Z690 boards don’t use SLI, so no NVLink if you add a second GPU later on
>> You will need an expensive board for this CPU, 300 to 400 Euros or more.. And for the 5950X it is like 150 Euros for a good board.. In a nutshell, the 12900k uses nearly double the power draw at stock of that 5950X.. so good motherboards will be a lot more expensive for sure.
>> Don’t wait, prices are very good now for GPUs, CPUs and SSDs… Ampere GPUs are even cheaper than the first release 2 years ago, you never know what will happen when the new CPUs or GPUs get released
>> 5950X is a great value CPU at the moment, and will still hold this value for some time .. yes it is the end of AM4, yet the 5950X at 700 Euros is a killer value and a very stable platform for production

Best,
Muhammed

I had this conundrum. After reading everything I could, they both are pretty equal in most things. Ryzen does get kudos in renderings, but even those numbers are marginal. I went with Intel and I have had no issues. The main reason I went with Intel over Ryzen was pretty much availability. Had the Ryzen need available I might have chosen that.

Thank you Mohammed for your very detailed response. Well after trying to do two intel builds, one in a itx case with a Noctua d12 hitting 100c and another in a atx case with a Noctua d15 hitting 92-98c I’ve decided to try AMD again because the next intel chip will have 8 additional e-cores and I can see those temps rising further. I can appreciate AMDs efficiency. I don’t like these high intel temps and don’t wish to do AIOs or custom loops.

I had a question about dual gpu. I purchased the ROG Crosshair VIII Hero mobo which has dual PCIe 4.0 slots that will run at 8x. My primary gpu is a RTX 3090. I have a spare 1080ti. If I insert the 1080ti into the second slot without SLI or NVlink, will vray be able to make use of the two cards? Or do I need to buy a second 3090 and use NVlink for any kind of multi gpu rendering?

You can use any combination of GPUs for rendering with Vray GPU as long as they are all Nvidia. The series and model number doesn’t really matter as long as they all support CUDA. You don’t need Nvlink to render on 2 or more GPUs simultaneously. The available GPU memory in this case will be the amount of memory of the GPU with the smallest capacity, if that makes sense. So if you install a 3090 and a 1080 Ti and you use both for rendering, the available video memory will be 11GB. Nvlink is used when you want to combine the memory of several GPUs into one large unified memory pool. So if you get two 3090s and Nvlink them together, you will have 48GB of memory available to fit your scene in when rendering instead of 24GB. I’m not sure if you can link different generation cards with Nvlink though.

Hi @Muhammed_Hamed
So I completed the ryzen build and on Vray benchmark the CPU is high higher, which is great, but for some reason on both RTX and Cuda tests, my scores are half of what they used to be. Would you know why this is?

Update: So i see the numbers have gone back to normal for RTX benchmarks but only if i wait 1 minute after boot up. If i try to do it as soon as i hit the desktop then the score is reduced to half. Odd. Never saw this on the Intel Build. I wonder if its the mobo that has a sub system still starting up. Does anyone with an AMD cpu notice a similar phenomenon? try doing a RTX bench the moment you hit the desktop and then another after 1 min. they should be different by up to 50%.

Update 2: Ive noticed Phoenix simulation dont utilize the cpu 100%. Is that normal? Like i did the waterfall sim and it never exceeds 75%. My 8th gen intel is pegged at 100%. What im wondering is are there any bios settings i should be aware of. I dont exactly want to overclock but I dont want my cpu held back. I hardly ever see it hit boost clock speeds in HW Monitor.

I’ve just tried running the RTX bench immediately after rebooting and I don’t have this issue.

What motherboard do you have?

Asus ROG Strix TRX40-E Gaming

ok so i dont know what to make of this. Using GPU-Z, i can see that when the RTX benchmark score is cut in half at first boot up, that the TDP is 35% or about 150 watt power draw. When the score is the correct result, which i just run the test again or wait a minute, the TDP is 85% or about 350w power draw. Now here is the really weird thing. when its on the low performing benchmark, If i open the windows Snip tool, the TDP goes back up the %85. the moment i close it , it drops back down to 35% . I was trying to take a screenshot of the low TDP and everything I would open it, the TDP jumped back up. Does this make sense to anyone? im loosing my mind here. Is it the GPU, the CPU, the Mobo or Vray Benchmark. I still dont think its the CPU or GPU because when i run other tests at first boot, like the regular Vray benchmark or Heaven, its fine.

That is indeed weird. I’m not sure how much of a “problem” this is though if it goes back to normal a minute after Windows boots. You can’t even launch Max/Maya etc. and load a scene in less than a minute so how does that negatively impact you really?

I agree and I was all set to accept it but I benchmarked again while the computer was on for an hour and the scores dropped again to half. So its possible it can happen anytime and spending this much on a computer where it can lose half its gpu rending performance at any times is not ok.

Hi @Muhammed_Hamed , could you take a moment to reply to some of my questions and help me? Ive been trying very hard to solve this on my own and could really use your input after going on your suggesting with AMD. As you can see my benchmarks are very inconsistent after switching to AMD and I would like your help in solving this. I know its not the video card because the same thing happens to both my spare 1080ti and 3090 on their own. Ive also changed the motherboard to the Asus Crosshair Dark Hero and it is still happening. And the cpu benchmark never had issues. And why does activating the windows Snip app cure the issue. Is this a Vray Benchmark bug or do I need to try different hardware?. Im exhausted rebuilding my computer over and over so i really need your assistance.

Thank you.

Did you format your system disk and install a completely new fresh copy of Windows when you changed from Intel to AMD? You should definitely do this to avoid any leftovers from the previous installation that may interfere with the new hardware. Did you install the latest BIOS for your motherboard? Also, it’s very important to install the AMD drivers for you CPU and chipset. You can find them here.

If these still don’t solve the issue, I would try running the V-Ray benchmark from a live Linux USB to rule out problems with the Windows operating system. Here’s a quick guide on how to do this. After you boot into Linux from the USB, make sure you install the Linux version of the benchmark and the latest Nvidia drivers for Linux.

Yes , I do a fresh windows 11 install in each build. I even tried a fresh windows 10 install just to rule out issues with 11 since its so new. i updated my bios. i downgraded my bios, I tried older version of Vray Benchmark. I just ordered a second CPU to try. I got 2 more motherboards on the way.

This is taken over my life and considering going back to Intel and deal with the high temps.

Hey Sean,

I will reply to your questions shortly

I have the same ROG Dark Hero, same CPU and same GPU
Could you please list all the system specs, power supply, drive, memory configuration. It will be helpful

I haven’t seen this behavior on my system, should be solvable I think(as we both have the same hardware)
What Nvidia driver are you using?
Do you have PowerToys installed? we have reported similar behavior to Nvidia recently on using this tool

Use the latest bios, default settings, install the recommended driver 512.59 (make sure to check clean install in the install option)

When you Benchmark or Render, never use the 1080ti with the 3090.. mixing both specially in RTX mode is not gonna work well
The 1080ti uses software raytracing, the 3090 uses hardware raytracing(tensor cores).. mixing both will slow down everything

What you should do is have the 3090 in the first PCIe slot, then install the 1080ti in the 3rd PCIe slot(I know it is x4 bandwidth shared, but you need to give breathing room to your 3090, this is priority)
Then connect your monitors to your 1080ti so it gets used for viewport performance, or GPU denoising in V6 and you keep the 3090 to your GPU rendering/Vantage

In Nvidia control panel, change Power Management mode to Prefer Maximum Performance
Swap OpnenGL rendering GPU to your 1080ti (this device is what Max/Maya use for viewport)

If you are on Windows 11, HW accelerated GPU Scheduling should be enabled by default. If you are on Windows 10(Recent builds) you need to enable this manually

https://cdn.discordapp.com/attachmen…mage_66836.jpg
https://cdn.discordapp.com/attachmen…mage_66837.jpg

Sorry about not replying sooner, I have been busy

Best,
Muhammed

when i run other tests at first boot, like the regular Vray benchmark or Heaven, its fine.

So this happens with specific scenes?

Sometimes the simulation goes way too fast, before all the threads can synchronize and the next sim frame comes up. If you make a heavier sim that needs a bit more time the CPU should stay at 100%.

My dear dear Muhammed, you did it! You solved the issue. It was the Nvidia studio driver 512.96. After you suggested using 512.59, I used DDU to remove 512.96 and clean installed 512.59 and the problem was solved. So make note of this bug with that version of driver. It will cause the card to idle (35% TDP) during RTX renders. Especially at first boot. But will happen randomly later as well.

You have saved me days of computer building. I had 2 more builds planned. One with a MSI board, another with the Asus ROG x570-E board. And then I had a second CPU on order. I am very relieved. I was getting tired of smelling CPU thermal paste cleaner. I liked everything else about the AMD cpu and was not looking forward to going back to intel. I was also considering going with a Mac. Im primarily a Mac user as I do most of my work thats not related to 3D on my Mac. I had a Mac Studio 20-core on order with a 2 month wait but upgrading my gaming PC seemed the smarter move. But to see the performance cut in half made me consider just getting the Studio even though its much slower (10x?) without GPU supported rendering. Im very happy to now officially be part of the AMD crowed. The temps are fantastic. I was hitting 98C on the intel and now they are 65C and faster too. My intel cpu score was 15000 and my AMD cpu score is 19000. No overclocking on either.

Thank you again. Maya always warned me I was not using 512.59 drivers, I will never ignore that warning again.

Btw this is my PCpartspicker list. I wondered what are the differences in our builds Your Account - PCPartPicker . If you have a list I’d love to see it. Or if you have anything you suggest I change let me know, Im still within my return window on everything.

Yes only with CUDA and RTX benchmark scenes . Never with the V-Ray benchmark scene.

Sean