Why do Skylake CPUs have lower CPUmark than Haswell?

Toggle sidebar Toggle sidebar
modernwar99

modernwar99

Reputable
Jul 9, 2014
1,166
0
5,960
Just noticed on passmark, all the scores for new Skylake CPUs are considerably lower than their Haswell counterpart, yet perform better in real world usage. Why is this? If anything, I would think the smaller sample size of the Skylake K series would result in OCed scores padding the overall score more, but I guess I'm wrong. Does it have to do with the lower clock speeds of Skylake, but higher IPC?

Not really a big deal since it's just one benchmarking tool, but it just seems weird because this didn't happen with Ivy Bridge to Haswell.
 
Solution
synphul
They're not really apples to apples comparisons in some instances. The 4790k would be comparable to the 6700k. While the 4790k does score a bit higher, it also has a higher turbo speed. It's 4ghz/4.4ghz turbo while the 6700k is 4ghz/4.2ghz turbo. The 6700 (non k) does score slightly higher than the comparable 4770 where both are at 3.4ghz. It's also possible that because of the architecture change passmark's testing doesn't accurately utilize the potential of skylake chips.

Some tests are very software dependent and updates can make a night and day difference in scores. When benching using intel's xtu program to achieve hwbot scores, my initial scores were about half of what they should be. Updating windows from a fresh install to sp1...
Sort by date Sort by votes
synphul

synphul

Champion
Mar 11, 2011
9,477
7
54,965
They're not really apples to apples comparisons in some instances. The 4790k would be comparable to the 6700k. While the 4790k does score a bit higher, it also has a higher turbo speed. It's 4ghz/4.4ghz turbo while the 6700k is 4ghz/4.2ghz turbo. The 6700 (non k) does score slightly higher than the comparable 4770 where both are at 3.4ghz. It's also possible that because of the architecture change passmark's testing doesn't accurately utilize the potential of skylake chips.

Some tests are very software dependent and updates can make a night and day difference in scores. When benching using intel's xtu program to achieve hwbot scores, my initial scores were about half of what they should be. Updating windows from a fresh install to sp1 immediately doubled the score on a second test run. I was a bit surprised by such a drastic improvement but it depends on what the software benching algorithms are set to look for and what code they use to run efficiently. In this case it may not be a matter of a windows update but because of skylake being so new passmark itself may not be coded to efficiently test the newer chips.

Personally I would take generic synthetic tests like passmark's results with a grain of salt. They don't tend to accurately represent real world performance. They also show an fx 9590 performing almost on par with a brand new i7 6700k and with a higher score than a 4790 (non k) and 4770k. In the real world this isn't anywhere close to accurate.
 
Upvote 1 Downvote
Solution
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom