Originally Posted by Sunagwa
Yes I understand that it is not something that SL just made up.
I wasn't aware that the new x line automatically uses this function but I suppose it makes more sense for that specific market/target workload.
I guess my main focus would be on the "gaming/performance" cpus. 7600k and 7700k. These CPUs never used to be tested using an AVX offset and now it appears to be standard practice.
In any case it's fine we don't have to agree. It would appear we're focused on very different workloads anyways. For me personally, I'm going to just roll the dice in the future as I see less value now that the AVX offsets are standard.
You never have to use the settings we provide. Nothing has really changed, the higher bin you grab the higher quality silicon you get. We tested with just Realbench before, which has some AVX in it. A CPU tested only with Realbench would instantly fail a more difficult AVX load like Prime95. A lot of you are okay with that, and that's perfectly fine, you don't need to run AVX offsets.
In order to provide a better experience in general though, we are providing settings that should be stable in all types of workloads. A customer should not expect the system to ever crash with the new settings during general use, even if more AVX heavy applications start coming out.
If you're looking for a more direct comparison to Realbench as before, I would say you lose about 50MHz jumping from Non-AVX stability to Realbench stability on mainstream chips. That is to say, you can imagine we're showing clocks about 50MHz higher than before.