By using this site, you agree to our Privacy Policy and our Terms of Use. Close
BHR-3 said:
drkohler said:

I still can't figure out their memory bandwidth numbers and data path layout by that interview.
A lot of extremely obfuscating descriptions and weird logic in that context (I could also say he is downwright lying at some points but maybe they have invented/implemented completely new technology). In my books, the maximum theoretical bandwidth is 149 (or 178?) GByte/s, but it's been 30 years or so since I designed memory controller circuitry...

...Nick Baker: Yes. That's been measured.

You are referencing the same single interview which boils down to the Leadbetter interview, which floats around the internet in various abbreviated versions. Again, in any versions of this interview, there are VERY questionable numbers and hardware properties raised by MrBaker (and I wonder how they measured their numbers). I do not doubt at all that some numbers can be added (given the right conditions) or that esram can be read and written to concurrently (again given the right conditions). As long as I don't see a satisfactory explanation how they got their numbers, I cast serious doubts on some of them.