My 13900k has been fine, but it’s also been on the Bios fix since it (they) came out, and I had a voltage limit of 1.35v before that.
AMD has the better CPU’s overall, (I’m running a 9800X3D in my sim rig), but Intel has an advantage with this particular software. I suspect it has more to do with Intel’s ability to run faster RAM and TVAI loves fast RAM. AMD can technically run RAM at fast speeds, but it has to “decouple” the RAM speed from the IMC speed and takes a big penalty when you run over 6000-6400MT.
Still, peace of mind has its own value and AMD hasn’t had the reliability issues Intel has.
Bios “Fix”, Voltage reduce in case of prevent from CPU-Degradation. In the company I have several HP Z 13900k /14900k Workstations which have a CPU utilization of 100% in idle mode, because the CPU is damaged
As Mellow said. RAM is likely to be the bottleneck for any high-end machine.
Note, the SPEED of the RAM is what matters, not the amount.
I got a linear speedup when I upgraded to faster RAM. From 4200 to 6000 it got me a ~40% speed improvement on some models. The app is still bottle-necked on RAM throughput though, so I’m eagerly following all memory hardware advancements
Note that TVAI uses very little amount of RAM. Most models use no more than ~2GB of VRAM, some less than that, and amount of system ram for a 1080p → 8k upscale requires about 20GB system RAM. So 32GB system DRAM should suffice if you’re only doing single-clip encodes. If you have a fast graphics card such as the 4090, then a 64GB DRAM machine would be enough to eke out every ounce of aggregate FPS throughput for a two-clip parallel encode.
As has been covered here on the forum previously, the BSOD isn’t Topaz’s fault. It’s that the systems people use to run TVAI are unstable. TVAI is a unique stress test that stresses the memory controllers along with CPU and GPU like no other benchmark. You may run prime95 along with Futuremark’s furmark for instance and not have a crash. But that’s just because neither of those is stressing the memory controller at the same time.
Everyone who had a crash and when through a stability tuning exercise seem to have had those issues resolved according to past posts in the forum.
Nope. Only in gaming (and that’s questionable per title) and on paper and from shills. AMD has huge issues as well. I mean… just go to any AMD’s social handlers or AMD’s forums.
I overlooked the post, ok that’s new for me, the question is what this function does exactly, I don’t trust TVAI with such things^^
I downloaded Avidemux V2.8.1 for Windows from here I don’t know why the program is in German, because my Windows language is englisch and I can’t change the program language, maybe it takes region code when installing.
import a video first then you must choose a codec (ffv1 for lossless) then register “video” and “filter” the Resample function is under “Transformation” this function must be there if not its strange.
ExtremeTech has an article with the owner of the card who admitted he screwed up by not using the cable that came with his 5090 FE. He used a cheap aftermarket cable that looked cool, but didn’t meet the electrical specs of the factory cable. Der8auer has sensationalized the story, just to generate views on his channel. It’s clickbait.
The only thing i can see that would bottleneck the GPU in TVAi is RAM read/write.
I have a 7800X3D which is far from 100% load, barely even 50%. However, I’m running overclocked ram at 6200mhz at a latency of 63ns, and I’m seeing read/write peaks close to highest read/write my ram can achieve. Average read/write is about 50% of max.
Can definitely see someone running eg. 4 sticks and being unaware of ram tuning would end up with 2x the latency and half the read/write. Resulting in a bottleneck for their GPU.
Did you watch the video? His own card was pulling over 20 amps on a single wire. Two wires carrying the bulk of the load while the others were effectively along for the ride. That’s NOT okay.
I just stated that 7800X3D and decent ram doesn’t bottleneck TVAi. As it’s heavily gpu-bound due to the models mainly use tensor cores. Only using a CPU to render with TVAi, I wouldn’t even want to know how long that would take. Even for a threadripper.
who to trust these days?
