Would love to hear people’s take on this.
Faster single-thread operation and QuickSync would be my two big reasons to switch. I’m worried about PCI-e lanes and bifurcation though.
Currently running a 5965WX with 256GB DDR4-3600, RTX 4090, six gen 4 NVME’s, four of which are in RAID-0, three SATA SSDs, two SATA HDD’s, 400GB Intel Optane P5800X, 10gbe.
Six gen 4 NVME’s is 24 lanes which already blows past Z790’s twenty. And that’s without any of the other storage or gpu.
On the other hand, QuickSync would allow me to edit my video footage without transcoding as it has 10-bit 4:2:2 hevc support which is lacking in both Threadripper and Nvidia. The multithreaded performance is about on-par with my Threadripper in Adobe Premiere and in single threaded tasks Intel blows it out of the water for things like Photoshop and After Effects.
I considered throwing an Intel Arc GPU in my Threadripper system to get the codec support but after the HUMONGOUS 4090 and my nvme-pcie cards there just isn’t physical room to fit it without some exotic riser configuration and I’d still have to invent a way to mount it in the O11DXL so it gets airflow.
What would you do? Is it worth the pcie bottlenecking to get quicksync and faster single-threaded performance?
Do you use more than 128GB of ram?
Not yet. I currently have 256 but my usage hasn’t gone above 100 or so.
You could possibly move to the 13900k and end up with money back in your pocket at better than your current performance depending on how you sold your current parts
But maybe you have some part of your workflow not mentioned here that would seriously be impacted from going from 8 channels of memory to 2 and from your equally impressive and overkill sounding storage down to something more average.
Move the 4090 to the bottom slot and get an Arc for encoding/decoding video.
That already tells the story. Unless you’re fine with downgrading storage and forego any future upgrade the Threadripper offers superior CPU and IO throughput.
Does the threadripper really not have the grunt to handle your 10 bit HEVC needs in software then?
I mean, only you know how much difference that will make to your workflow.
Similarly, only you know how much faster single thread performance will improve things for you.
Ultimately it is a compromise either way, as inevitably you’re losing PCIE if you switch, but I’d be inclined to somehow try the Arc gpu to get an idea of how that will affect things, as you seem to feel it’d be quite a big deal - could you use a lesser Nvidia GPU in the short term? Or temporarily bodge a ribbon cable setup with an open bench?