confusion with i7 vs i5 results - passmark

Toggle sidebar Toggle sidebar
G

ginjaninja

Distinguished
Jul 16, 2010
33
0
18,530
I'm looking at passmark's benchmarks and seeing some strange results. In single threaded performance, the 2600k gets 1915 and the 3570k gets 2011.

With that in mind, I'd expect the 2600k to get about twice the 'total' score as the 3570k. But this isn't the case, the 2600k only does about 19% better (7120 vs 8473).

Is someone able to explain this? Which one do I believe?
 
Solution
B
For single threaded apps, i7 vs i5 should make little difference since the main difference is the Hyper-Threading. Likewise, i3's are not spectacularly faster than Pentiums with just 1 core active - they come into their own with more threads running than physical cores (4T/2C for i3 & 8T/4C for i7).

Ivy i5 scores higher than Sandy i7 in 1T mainly because the Ivy platform is about 10% more efficient than Sandy. So 2011 (3570k) vs 1915 (2600k) is not unusual for 1T. Comparing "like for like" (same gen), a 3770k i7 may well get 2000-2100, and a 2500k only 1900.

"so this 'per core performance' that passmark does, is that per physical core, not per thread?"

I'm not sure how they calculate it internally, but "the 2600k only does...
Sort by date Sort by votes
B

BSim500

Honorable
Apr 6, 2013
269
0
10,960
i7's won't get double the score of an i5 (the effect of Hyper-Threading is around 10-25%, not 100%, and Ivy Bridge's are around 10% faster than Sandy Bridge all else being equal). Hyper-Threading also won't show much difference in single-threaded benchmarks (by definition). And in my experience, Passmark scores can vary wildly even amongst the same CPU. Same is true of "Geekbench" online scores too. I was looking online the other day at i3's and for some chips, the difference was upto 40% "spread" (highest vs lowest) for the same CPU & clock speed.
 
Upvote 0 Downvote
G

ginjaninja

Distinguished
Jul 16, 2010
33
0
18,530


Ahhhhh so this 'per core performance' that passmark does, is that per physical core, not per thread?

are the i7s then slower for some single threaded tasks? it would seem their per thread performance is much lower than the i5s when all cores are loaded
 
Upvote 0 Downvote
B

BSim500

Honorable
Apr 6, 2013
269
0
10,960
For single threaded apps, i7 vs i5 should make little difference since the main difference is the Hyper-Threading. Likewise, i3's are not spectacularly faster than Pentiums with just 1 core active - they come into their own with more threads running than physical cores (4T/2C for i3 & 8T/4C for i7).

Ivy i5 scores higher than Sandy i7 in 1T mainly because the Ivy platform is about 10% more efficient than Sandy. So 2011 (3570k) vs 1915 (2600k) is not unusual for 1T. Comparing "like for like" (same gen), a 3770k i7 may well get 2000-2100, and a 2500k only 1900.

"so this 'per core performance' that passmark does, is that per physical core, not per thread?"

I'm not sure how they calculate it internally, but "the 2600k only does about 19% better (7120 vs 8473)." is perfectly normal. Your i7 probably does 30% better in multi-threaded benchmarks vs the i5 2500k - but only about 20% difference of 2600k vs 3570k due to Ivy being 10% more efficient per clock.

"are the i7s then slower for some single threaded tasks?"

They shouldn't be across the same generation (Sandy vs Sandy). The main difference you're seeing is Ivy vs Sandy. Yes an older i7 Sandy may well be slightly slower than a newer i5-3570k Ivy at same clock rate 1T only. And an i7 3770k may be slower than a newer i5-4670k at same clock rate 1T only. In fact, some of today's Ivy/Haswell i3's get higher 1T scores than 1st gen i7-920 (skt 1366) even adjusted for same clock rates.
 
Upvote 0 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