On the GPU front we can see that since cache pre-fetch is the issue that all classes of GPU/CPU & other processor class with cache may well face issues.

https://insidehpc.com/2018/01/nvidia-races-patch-gpu-drivers-spectre-meltdown/

https://nvidia.custhelp.com/app/answers/detail/a_id/4611

***
Intel information with sub-tabs (of interest)

https://www.intel.com/content/www/us/en/archi...s-and-intel-products.html

Running all AMD, 2 win 10 FX-8320 and 2 win 7 Phenom II x6 1035T...fully updated...no issues so far.

OK, now I can update my Linuxes.

& nVidia drivers. Without any fear. ;)

Just for the record:
On 2018-01-13, I updated two Athlon II x4 based machines running LinuxMint 17.3.
The kernel update (3.6.x > 4.4.0-109) ran well and the machines work fine.
Cheers,
Yves

"The micro code be updated to flush & or contain the the speculative data in a data-cycle secure storage,
Within the framework of cache and ram/virtual-ram?
Cycle efficiency would be at most two cycles and a flush Xor bit data overlay,

Bit Masking before and after pre-fetch presents & also uses data - this method would be fast! (c)Rupert S"

"Obviously in light of buffer exploitation we would suggest that buffers after password entry are cleared, This is not the whole solution because the spy program could be resident..

Buffer exploitation is a common practice in viruses and this type of attack is nothing new..
There is no doubt that buffers are a victim of flooding and exploitation; Over and over!
After all buffer exploitation is a logical consequence of their use on a computer or hardware.

Randomizing buffer allocation, Location and encryption algorithm is the most logical choice on hardware, However! how much effort must be made to protect buffers when an attack on them is predicable and logical? A lot we say.

(c)Rupert S"