News MSI Accidentally Publishes Specs of new Raptor Lake-S Refresh CPUs

[bit_user]

So can reading what somebody responded to instead of just taking it out of context.

I always read posts I respond to. If you think I missed a part, it's on you to explain what you think I missed.

Similarly, if you believe a point was taken out of context, it's on you to explain what context got removed and why it's important.

Merely complaining adds no value. Every post should attempt to add value to the discussion, or else it doesn't belong here.

 

[PCWarrior]

Power/thermal constraints and memory limitations can reduce that 3.4% particularly if the node is the same as said in the article.

I am perfectly aware that you don’t get perfect linear scaling with frequency due to other bottlenecks but I am not sure why you felt the need to point that out. I never made a claim to the contrary in my original post. All I said is that by calculating you expect 3.4% so the 3% that MSI says fits. Also I said that this 3% being on average would also mean having applications that get less than 3% or even no increase at all (that would mean they are completely bottlenecked/constrained by something else) while there are other workloads that get more than 3%. On that note my argument was that in multithreaded the percentage increase in frequency and the resulting performance uplift could likely be higher. And such a thing would be enabled by a larger increase in the all-core frequency (e.g. increasing from 4.9GHz to 5.3-5.4GHz when all cores including all e-cores are loaded).

Totally agree. But from an Intel's product development costs POV I really cannot understand.

No, it makes huge sense from a marketing perspective. Besides the new press coverage, the buzz about it and likely Intel reclaiming the top spot in everything, you have the buyer’s psychology. A buyer is hesitant to buy something that might be replaced soon or that soon one can buy equivalent performance for less. This is what happened with Nvidia’s 3000 series during the 4-5 months before the launch of the 4000 series. A next gen launch (even an incremental one like Intel’s 14h gen) resets the clock. Buying 14th gen means that you don’t buy something that will soon be considered obsolete. Also knowing that 14th gen is not much of an upgrade versus 13th gen means that many will feel good in buying 13th gen and even do so before 14th gen launches as 13th gen won’t be obsolete either. So yeah it makes perfect sense from an Intel's product development.

 

[bit_user]

On that note my argument was that in multithreaded the percentage increase in frequency and the resulting performance uplift could likely be higher. And such a thing would be enabled by a larger increase in the all-core frequency (e.g. increasing from 4.9GHz to 5.3-5.4GHz when all cores including all e-cores are loaded).

Don't forget the increase in DDR5 speed. That's a likely bottleneck, for all-core workloads. Going from DDR5-5600 to DDR5-6400 is a 14.3% increase in throughput.

Also knowing that 14th gen is not much of an upgrade versus 13th gen means that many will feel good in buying 13th gen and even do so before 14th gen launches as 13th gen won’t be obsolete either. So yeah it makes perfect sense from an Intel's product development.

It would've been nice to see them do a little more, like upgrading the ring interconnect to a mesh, or something like that.

Kaby Lake didn't impress me - it felt too much like buying a factory-overclocked Skylake. This feels similar.