Interesting note on Phenoms

[Kaotik]

Like everyone already knows, Phenom X4 looses on average on most tests to similarily clocked Core2Quads.

At HardwareCanucks test, there was something odd that caught my and some others eyes - the minimum framerates. While the Phenom lost on average FPS most of the time, the minimum FPS'es were faster on every case, sometimes just a tiny bit, sometimes a lot more.

Anyone got any educated guesses on what's going on?
http://www.hardwarecanucks.com/foru...7-amd-phenom-x4-9600-performance-preview.html

 

[K.I.L.E.R]

It's settled, Phenom is for me. Minimum FPS counts more than anything.

 

[3dilettante]

A lot of things can cause a downward spike in FPS.

Given the really low numbers they have for minimums, I'd hazard a guess it's not necessarily the CPU or GPU stumbling.

There are a few system differences not accounted for, such as the motherboard chipset and driver differences.

An overlooked suboptimal driver for the chipset or other slow part of the system might be enough to make that rare dip a few fames lower than it would be otherwise.

There could be some inherent advantage that Phenom might have over Core2.
Perhaps there are a few times where FSB contention slightly impacts performance.

There's not much context, such as whether these dips happened often or just once or twice, and the amounts involved are so small that it is difficult to say it's significant unless they ran the tests 10 times.

 

[digitalwanderer]

Mebbe it's because Intel is just bolting together two dual cores instead of being a true quadcore architecture? :|









/me runs like hell!

 

[Bludd]

Maybe the higher min. FPS is due to the integrated memory controller and maybe also HyperTransport being better than FSB?

 

[BRiT]

Nice find.

It would be interesting to see a histogram of the frame-rate. A single high or single low is nearly worthless.

 

[Accord1999]

Other reviews show the opposite:

http://www.tecchannel.de/pc_mobile/prozessoren/1739204/index17.html
http://www.legitreviews.com/article/597/8/

 

[L233]

Nice find.

It would be interesting to see a histogram of the frame-rate. A single high or single low is nearly worthless.

Yeah, I thought the same thing. Min/max values are useless without a graph, they might be isolated occurances. Maybe an average of the 10 lowest values or something like that would be more useful in absence of a histogram.

 

[Blazkowicz]

Mebbe it's because Intel is just bolting together two dual cores instead of being a true quadcore architecture? :|









/me runs like hell!

no, Intel likes high rev-ing, low torque gasoline engines with more than 100 hp per liter whiles AMD prefers big turbodiesel engines that cope well with the heavy loads.

 

[swaaye]

Except those low-torque, high rev-ring Intel engines have much better IPC most of the time. AMD is big on letting their engines breathe though, lol. Big, open air intakes and wide exhaust to their hot, rough-running, inefficient diesels lol.

 

[DJ12]

It seems odd that it doesn't even beat the 6400+.

Seems like something is wrong, maybe the games aren't using all the cores?

 

[3dilettante]

Most games won't.
Unless the software is explicitely programmed to run on enough threads to be used by all cores, there is minimal benefit to the cores left out, besides the possibility that a spare core can handle any random OS or system work that would get in the way of the game.

There's nothing wrong with this.
It's not sensible for games that were mostly written before multicore took off to assume the presence of multiple cores that didn't exist at the time.

Without the advantage of multiple utilized cores, it comes down to single-threaded execution.
Phenom's worse memory latency, smaller L2 cache, and lower clock speeds can cancel out its other advantages in a number of areas.