Sandy Bridge i5 vs i7

arad85

arad85

Suspended / Banned
Messages
9,438
Name
Andy
Edit My Images
Yes
Hmmm..

An off comment on another forum got me to thinking.

As many people know, I have an i7-2600K I love and bought it over an i5-2500K as I recode video quite often and I figured the extra threading would help. So.. I thought I'd do a little experiment. Do some tasks with hyperthreading turned on and then use the BIOS to switch off hyperthreading and run the same code again.

Without boring with details, my findings are that for compute intensive tasks (i.e. where the cores are working near enough flat out) you get a 20% performance improvement. Where the cores are not flat out, performance can actually decrease as the system must spend more time managing more threads than doing work.

I think I'm going to change my recommendation from i7-2x00K to i5-2x00K for desktops. For mobile, the thing that matters is that the i5's are actually 2 cores and all but the lowest i7 mobile chipset is a real quad core. Sticking with i7-2xx0QMs for mobile...

An enlightening - if fairly boring - couple of hours. Now, where's the humble pie smilie....
 
The i5-2500k is reported to perform as well as the i7-2600k for most day to day tasks and gaming. Plus you can clock it to 5000 with the right cooling.

Still wouldn't trade the balls out grunt of the i7 though personally for when I need it.
 
I 2-pass reencoded a 25 minute HD TV program. In 4 core only mode it took 0:24:00 dead. In hyperthreading mode it took 0:22:41. A whole 5% improvement.....
 
arad85 said:
I 2-pass reencoded a 25 minute HD TV program. In 4 core only mode it took 0:24:00 dead. In hyperthreading mode it took 0:22:41. A whole 5% improvement.....
Click to expand...

Pfft, 2 mins is 2 mins :lol:

There's no denying the 2500 is a good option. What's the price difference between that and a 2600?
 
Pfft.. am I your secretary now...

£173 vs £253.... for i5-2500K vs i7-2600K
 
Last edited:
PS. and it was 1 minute 19 seconds.... :p
 
i was lucky that I bought mine i7 for 233 not long ago ;D
 
According to my Amazon wish list it is £235 currently.

Yes I know I should stop wishing and just order it....

Back on topic... I looked at a CPU benchmark chart site and the i5 scores 6700 and the i7 scores 9100 which seems a lot more than 5%. Are those comparisons not realistic then?
 
i imagine the tests use hyperthreading to its full potential whereas currently the amount of apps that utilise it are few? im sure andy will be along to go into much more detail though..
 
Yes, I think the benchmarks must use hyperthreading to the full - and up to now, my findings have scaled well with those benchmarks.

My particular benchmark is a 2 pass recode of HD video. In pass 1 the encoder just does a rough encode looking at which frames to allocate the most bandwidth to when encoding. The second pass it uses this information to encode frames with either a lot of bandwidth or a small amount of bandwidth. The first pass is actually slower with hyperthreading enabled - the system must spend more time managing the threads than actually doing work in them.

I've just done a test changing the number of threads used to encode 5 minutes of video on the first pass as described above and this is the result

Code: 
Threads  fps
   1      30
   2      65
   4      73
   6      73
   8      72

Just tried 8 threads from the SSD and it's exactly the same - so it's not I/O bound.

Video encoding is something that can be multi-threaded very well and computationally intense, so should scale well but doesn't in this case and I'm not sure why at the moment.
 
You must log in or register to reply here.
Back
Top