• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Radeon VII

Illuminist said:
Yeah I watched a video of a guy using the - 30 outside in America at the moment to cool his vega to 1 degree.
And then was very sad to see him completely fluff his way through overclocking it with no under volt at all. His YouTube has many thousands watching him and its just shocking. He got 1850mhz on the core and 1000 on the hbm and called it a day. .....
Click to expand...

I saw that and couldnt believe he didnt do an undervolt. His channel isnt too bad if you dont take it too seriously.
 
D.P. said:
No, Tensor cores have nothing to do with CUDA cores and they aren't similar in the slightest. CUDA cores are general purpose arithmetic units and fully programmable. Tensor cores are fixed function accelerators.
Click to expand...

Did you read what I wrote? Do you know down to architectural level what Tensor core is? I wrote in "layman's" terms what they are, yet you completely missed the point.
 
Panos said:
Did you read what I wrote? Do you know down to architectural level what Tensor core is? I wrote in "layman's" terms what they are, yet you completely missed the point.
Click to expand...


Yes, in laymans terms you are totally missing the point.
In laymans terms a GPU is the same as a CPU, so why bother having GPUs at all? If you are not going to distinguish between hardware based on their critical functionality then the entire point is moot.
 
D.P. said:
Tensor cores are very different to either CUDA cores or AMD's stream processors.

That is why they exist in the first place, dedicated hardware to accelerate specific functions.

You can compute ray-tracing in the CUDA cores/Stream processors, but doing so is on the order of 10-20x slower since it is run in software without specific acceleration of the intersection test, and moreover, all of those CUDA cores/Stream processors get taken up doing ray tracing instead of working on traditional razerization which is a requirement for the next 5-7 years until RTX hardware advances enough that the traditional rasterization can be entirely removed.
Click to expand...
If you notice on 3dmark on the port royale benchmark it says it will only work with dx raytracing enabled on the software drivers. This to me is very telling as it doesn't say rt cores or nvidia. It says in the software
 
D.P. said:
Yes, in laymans terms you are totally missing the point.
In laymans terms a GPU is the same as a CPU, so why bother having GPUs at all? If you are not going to distinguish between hardware based on their critical functionality then the entire point is moot.
Click to expand...

How you would understand what tensor core is other that Tensor cores are "cores" designed for 4x4 FP16 matrices?
 
Illuminist said:
If you notice on 3dmark on the port royale benchmark it says it will only work with dx raytracing enabled on the software drivers. This to me is very telling as it doesn't say rt cores or nvidia. It says in the software
Click to expand...

Of course, because DXR is not tied to a specific hardware implementation. Nvidia worked with MS to create an industry standard API.

This is obvious from the fact that Nvidia first demonstrated RTX and DXR using Volta which lacks Ray tracing cores. The whole point of Turing is it is aorund 20x faster than Volta despite having far fewer CUDA cores and those CUDA cores are then free to do traditional rasterization tasks.
 
D.P. said:
Of course, because DXR is not tied to a specific hardware implementation. Nvidia worked with MS to create an industry standard API.

This is obvious from the fact that Nvidia first demonstrated RTX and DXR using Volta which lacks Ray tracing cores. The whole point of Turing is it is aorund 20x faster than Volta despite having far fewer CUDA cores and those CUDA cores are then free to do traditional rasterization tasks.
Click to expand...

https://www.nvidia.com/en-gb/data-center/tensorcore/
 
Panos said:
She is gorgeous..... :D
But graphite thermal padding instead of TIM, that's new. Which means replacement time as I would more likely reduce 10C easily.
Click to expand...

Panos said:
The box was heavy due to the stand....

What worries me is the thermal capability of the graphite pad used. Might be of very good quality, and would work. But might be a cheap one, and need to replace it with Cryonaut.
Click to expand...

I read graphite has thermal conductivity of 1950 W/m.K. This is multiple times of any TIM or conventional solder.
 
You must log in or register to reply here.
Back
Top Bottom