Replies to post #143080 on Intel Corp (INTC)

[Dmcq]

Francois apparently screen capped his geekbench results

Looking at those he says 'up to' and the 25% was on a fanless machine with figures indicating it was probably running one of the M-5Y.. type processors. It would have helped if he said what exactly he was doing and on what machines with what processors. That indicates to me that what he was seeing was thermal throttling so I'm afraid it is probably a bit of an Intel own goal.

[The_Trooper]

https://twitter.com/FPiednoel/status/665246003390345216

So the up to 30% turns out to be 10% for the total score.

If you zoom out you'll find this:

Integer: 3544 vs 3543
FP: 3400 vs 3454
Memory: 5067 vs 3509

The memory score is creating a huge distortion.

https://twitter.com/FPiednoel/status/665249308346724352

This one is more interesting:

Integer: 3075 vs 2456
FP: 2978 vs 2287
Memory: 2706 vs 2325

So on one side mostly no variation on Integer/FP, and ~25% in the other case.

This needs more analysis and preferably not from someone working at Intel.
Or wait for GB4 in the hope it will be better.