> Can you use the IA tensor cores of nvidia to program ASIC of asteroids at home?

No, tensor cores do support low-precision data types like FP16, they're specifically built for AI workloads. The Asteroids app depends on FP64, and lower precisions aren't good enough. That’s also why none of the existing BOINC projects make use of tensor cores (or NPUs); they're not suitable for scientific computing. Like ASICs, tensor cores are highly specialized, great for AI, but pretty much useless for most scientific applications.

Technically the chip GB203 of rtx5070ti have tensor cores in FP64

When i received the chips AMD, i wil make the computers, about final of month may, in my country the transport is bad, and very slow.

He empezado las pruebas con el equipo AMD 7950x RX9070XT, esta 5 minutos procesando la tarea con la frecuencia de la memoria de la tarjeta entre 1200mhz u 1538mhz,, totalmente fria consumiendo 83wats, y despues , pasados los 5 minutos comienza a procesar, procesa 3 minutos y 18 segundos empieza a calientarse hasta los 2786mhz de memoria consumiendo 220-236wats, haciendo un total de 8minutos y 14 segundos, definitivamente hay algo mal en las tareas ps_250507_input_29478_69 (221603 bytes). porque solo trabaja a alta frecuencia la memoria los ultimos 3 minutos Las tareas tardan lo mismo que una rtx4060.

Los amd 7950x son especialmente calientes, sin overclok los 32 hilos al 75% de tiempo llegan 70ºC a 75ºC. suelen tardar 40 minutos cada tarea, una pequeña o ligera diferencia frente a los 45 minutos de loa amd ryzen 9 5900x.

Los amd7950x con ddr5 es la peor inversion que he realizado, rinden casi lo mismo que amd 5900x con ddr4, voy a dejarlo trabajando unos dias para ver los datos, creo que los tiempos se estan reduciendo.

Una cosa curiosa de rx9070xt , es que la temperatura de la memoria es 10 grados superior 71ºC, a hot spot 64ºC , cuando la gpu solo tiene 43ºC

Zen 4 offers some enhancements over Zen 3, higher clock speeds, a larger L2 cache, and support for AVX-512 (limited to 2×256-bit execution). The power efficiency is also better thanks to 7->5nm process node.

You can see up to 30% improvement in some apps, but in most cases, the gains are only a few percent, primarily due to the increased clock speeds. If you expected more, you'll be disappointed... The days when each new CPU architecture brought a 2× performance leap are long gone.

I disagree, I think it's software, given that the computing power of the RX9070XT is much higher than the RTX4060, and they get exactly the same score.


Estoy en desacuerdo, creo que es software, dado que la potencia de calculo de la RX9070XT es muy superior a RTX4060, y sacan exactamente la misma puntuacion

I disagree, I think it's software, given that the computing power of the RX9070XT is much higher than the RTX4060, and they get exactly the same score.


Estoy en desacuerdo, creo que es software, dado que la potencia de calculo de la RX9070XT es muy superior a RTX4060, y sacan exactamente la misma puntuacion

Yes, i using the rtx4060 on folding at home, and i obtain 4.000.000 ponts per day, but i now i using the rtx5070ti, and obtain 25.000.000 points per day.


In asteroids the rtx4060 work 1 unit 6-8 minute /wu = 50points
Buy the rx9070xt work 1unit 6-5 min /wu 50 points

I think is the software of application on asteroids

It's ultimately about the software. While you can compare the RTX 4060 and RX 9700 XT based on FLOPS in areas like FP32 for peak performance, the actual performance can vary depending on the specific workload.

you can check some compute benchmarks here
https://www.phoronix.com/review/amd-radeon-rx9070-linux-compute/3

Software and driver optimizations can certainly help, and the Asteroids GPU app currently doesn’t scale well on high-end GPUs due to memory bottlenecks rather then compute power, so there’s definitely room for improvement on the software side for both AMD/NVidia cards.

I disagree, I think it's software, given that the computing power of the RX9070XT is much higher than the RTX4060, and they get exactly the same score.


Estoy en desacuerdo, creo que es software, dado que la potencia de calculo de la RX9070XT es muy superior a RTX4060, y sacan exactamente la misma puntuacion

Yes, i using the rtx4060 on folding at home, and i obtain 4.000.000 ponts per day, but i now i using the rtx5070ti, and obtain 25.000.000 points per day.


In asteroids the rtx4060 work 1 unit 6-8 minute /wu = 50points
Buy the rx9070xt work 1unit 6-5 min /wu 50 points

I think is the software of application on asteroids

From what I read, I should be doing much faster on my Wu's with an rtx5070

yet it takes hours, they seem to take nearly no GPU load.

That’s because the system is using the AMD iGPU instead of the RTX 5070. Blackwell GPUs aren’t officially supported due to different CUDA requirements. You can make it work with a custom app, though.

I just noticed the Wu says ATI/AMD

How can I set that up to work with CUDA unit's

Probably is interesting, no buy the new rx9070xt, and buy rtx5000 series, by , in a other hand... without sli cross fire, ¿can run two graphic cards two wu isolate in the same motherboard?, because the motherboards have 1 pcie 5.0 and two 4.0, totally 3 pcie

Boinc does not support running the same task on 2 gpu's at the same time, only games do that.

Boinc STILL has one single cache that covers BOTH the cpu and gpu, running a single gpu task on 2 or more gpu's at the same time probably isn't even a thought in the Developers heads.

Yes lots of people run a task on every gpu they can fit into their pc, I've seen a guy running 7 gpu's in one box and it's not a mining setup. The one caveat to that is do not use the onboard gpu in the cpu chip for crunching, it can get way too hot and seriously shorten the life of your cpu. They make very nice stand alone boxes that laptops can use if they really want to run gpu tasks.

For gaming or rendering, the priority is having a single job finish as quickly as possible.

However, for BOINC, it’s better to run multiple independent jobs per GPU, which is fully supported, and PCIe speed doesn't really matter.
Using multiple GPUs to speed up a single job isn’t the goal. Even though BOINC allows assigning multiple GPUs to one WU, it would be more complex and less efficient.