-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
So will the 6 core AMD will outperform i7?
- Thread starter easyrider
- Start date
Soldato
Sometimes I wonder if games that claim there using more then 2 cores really are, in Supreme commander even after the tweek there was still raised eyebrows.
Same applies for many games, obviously theres some ie unreal engine based games where it jumps from 150fps to 200+, Still find it iffy perhaps its just poorly coded games half the time.
What would be the best route though go Corei5/i7 and upgrade to a cheaper 6 core come fall/2011, or go AMD and then upgrade to a 6 core come Summer time ?
Same applies for many games, obviously theres some ie unreal engine based games where it jumps from 150fps to 200+, Still find it iffy perhaps its just poorly coded games half the time.
What would be the best route though go Corei5/i7 and upgrade to a cheaper 6 core come fall/2011, or go AMD and then upgrade to a 6 core come Summer time ?
Soldato
1 years bit far off I think 
Im tempted too stick too intel, but the prospect of AMDs 6 affordable 6 core coming this summer makes it tempting too go AMD route
Im tempted too stick too intel, but the prospect of AMDs 6 affordable 6 core coming this summer makes it tempting too go AMD route
I'm very confused with the way AMD is "counting cores" on its Bulldozer CPU. The way I read the article on anandtech was that a 8 core CPU would be made up of 4 Bulldozer modules. That makes me thing its a Quad core with hyperthreading?
Am I completely misreading this?
Am I completely misreading this?
Caporegime
A tweaked Phenom II's clock for clock better than a Core 2. (I.E NB/HT tweaking)
Soldato
I don't think it is tbh...
Caporegime
Tested it against a mates Q9550, however, we only tried 3DSMax 9 with a render test.
In other app's, maybe the Core 2's better, but the difference is still minimal.
Associate
- Joined
- 16 Feb 2010
- Posts
- 64
Many games may use up to X cores but you wont see it unless the applications thread management system decides it's quicker to spool another thread and partition the work load than it is to wait in line and execute it normally.
Spooling up threads and partitioning work isn't free and in many cases it's not needed unless the current queue of work is long enough to justify it...
So your i7/i5/Phenom II/etc is often so fast its gonna plow through the current task queue only using 1 or 2 or 3 cores.
Also not all work can be done in parallel, there will be plenty of stuff your game is doing that just cant easily be sent of to be done on another thread because of lock and race conditions, etc.
The exception may be these console ports where the target system (e.g PS3, 360) clock speeds and core counts are know up front and work partitioning and thread spooling happens at hard coded points.
I've just written a procedural level generator for a game, I spent an hour+ multi-threading it to work on any number of cores... But the levels it will be generating are so small (relatively speaking) that the version that was running on all 4 of my Phenom x4's cores took longer making threads than it did to just make the whole level on a single core.
This is an example where the end user might moan "But it's not using all my cores"... to which the reply is "... and it's all the better for not doing so"
Spooling up threads and partitioning work isn't free and in many cases it's not needed unless the current queue of work is long enough to justify it...
So your i7/i5/Phenom II/etc is often so fast its gonna plow through the current task queue only using 1 or 2 or 3 cores.
Also not all work can be done in parallel, there will be plenty of stuff your game is doing that just cant easily be sent of to be done on another thread because of lock and race conditions, etc.
The exception may be these console ports where the target system (e.g PS3, 360) clock speeds and core counts are know up front and work partitioning and thread spooling happens at hard coded points.
I've just written a procedural level generator for a game, I spent an hour+ multi-threading it to work on any number of cores... But the levels it will be generating are so small (relatively speaking) that the version that was running on all 4 of my Phenom x4's cores took longer making threads than it did to just make the whole level on a single core.
This is an example where the end user might moan "But it's not using all my cores"... to which the reply is "... and it's all the better for not doing so"
Soldato
That's definately untrue fella
Benchmarking, Encoding, compressing\uncompressing and raw performance in games. you'll find no backup on that claim atall
Last edited:
Soldato
I'm pretty sure that at best the Phenom 2 is clock for clock the same as a Yorkfield.
Caporegime
Oh boy you guys make it too easy
Soldato
Encoding is a popular task for multicore CPUs, so I just ran a x264 benchmark here with my Phenom 2 clocked to 3.8GHz.
My RAM and CPU-NB were "tweaked", ram @ 1600MHz 7,7,6,20....CPU-NB @ 2.8GHz
The benchmark can be obtained here: http://www.techarp.com/showarticle.aspx?artno=520
And got the following average results:
The techarp database doesn't have a Yorkfield clocked to 3.8GHz but they have one at 3.64 and another at 4.1GHz. By interpolation (assuming linear scaling) this means at 3.8GHz the results of a Yorkfield would be:
That means the Phenom 2 is 3% faster at run1 and 3% slower at run2 in comparison with the Yorkfield....which I think we can conclude means that they are pretty much the same clock for clock at encoding.
My RAM and CPU-NB were "tweaked", ram @ 1600MHz 7,7,6,20....CPU-NB @ 2.8GHz
The benchmark can be obtained here: http://www.techarp.com/showarticle.aspx?artno=520
And got the following average results:
Run1 = 82.65
Run2 = 24.795
Run2 = 24.795
The techarp database doesn't have a Yorkfield clocked to 3.8GHz but they have one at 3.64 and another at 4.1GHz. By interpolation (assuming linear scaling) this means at 3.8GHz the results of a Yorkfield would be:
Run1 = 80.3
Run2 = 25.35
Run2 = 25.35
That means the Phenom 2 is 3% faster at run1 and 3% slower at run2 in comparison with the Yorkfield....which I think we can conclude means that they are pretty much the same clock for clock at encoding.
Last edited:
Most developers are just splitting the workload that would ordinarily be done on a single processor into multiple threads, which means that you have a game using 4 cores with 25% usage each instead of 1 core with 100% usage.
It's the reason that slower Phenom/C2Q processors are able to match i7 in gaming, because all the processors are able to get the job done with plenty of cycles to spare.
Last edited:
Soldato
Oh boy you guys make it too easy
I'm confused....Are you trolling? or are you saying everyone is wrong but you?
No one is denying the performance is similar, but phenomII does not out perform core2 @ like for like clocks
Caporegime
Oh, lol..
I was trolling lol.
Regardless, in the 3dsmax 9 rendering, my 965 at 3.8GHZ/2.8GHZ NB/2.4GHZ HT outperformed clock for clock, a 3.8GHZ Q9550. I'm saying it's different app for app.
Caporegime