SIX is Sexy! (Unboxed)

[ridney]

I am done benching with 250 HTT turbo on all auto settings, single and multi threaded runs the turbo is really amazing, i'm really impressed with this new technology. Congratulations AMD, you've done a good job and made us proud

Single thread











multi-thread








that's it guys, i'll be back tomorrow to seriously overclock this beast

[Manicdan]

only 19k in CB10? can you do a run without taking a bunch of screenshots, see if ur score is really higher?

edit, well i guess 19k for 3.5ghz is sexy enough, 20k is probably going to be a joke even on the stock cooler

[informal]

3.5Ghz and 19K+ . And Turbo Can Overclock Real Easy works like a charm,4.12Ghz and perfect scaling to give 4350pts . The tech is obviously immune to Windows 7/Vista's poor thread scheduling ,this is really amazing stuff we are seeing here .

[radaja]

how do you get cpuz to show each core?i tried this and it shows the same core0 for all instances

[ridney]

manicdan, i was trying to show that it's possible to up HTT to 250 with Turbo CORE still working on auto settings. mind you i have just done 258 HTT and still going strong (turbo'd to 4.257ghz ) i won't stop upping the HTT until this thing reaches the limit. 20k is pretty easy tomorrow i will try the maximum

[AMD-C]

only 19k in CB10? can you do a run without taking a bunch of screenshots, see if ur score is really higher?

edit, well i guess 19k for 3.5ghz is sexy enough, 20k is probably going to be a joke even on the stock cooler

19K is right about where it should be. Additional cores aren't going to be perfectly linear. You generally see about 6% reduction going from 2 to 4. Seeing a further 13% going from 4 to 6 isn't terrible. Nehalems seem to lose about 8% going 2 to 4 and a further 10% going 4 to 6 (Of course, the combination of Turbo and HT makes and kind of consistent/sensible comparison, even within the family, very hard.)

[Manicdan]

manicdan, i was trying to show that it's possible to up HTT to 250 with Turbo CORE still working on auto settings. mind you i have just done 258 HTT and still going strong (turbo'd to 4.257ghz ) i won't stop upping the HTT until this thing reaches the limit. 20k is pretty easy tomorrow i will try the maximum

thats the best damn part, just oc the HTT a little, dont touch nothing else, boom 25% more performance, while still probably consuming less power than what people get with CnQ just being off.

i just bought a second 4850, 85$ shipped, 15$ rebate, and i play WoW heavily, which is sofa king retarded being only duel core optimized, that im cpu limited even at 3.X ghz, so its time for some 4.X ghz fun. and at 250$, i think this could last me till BD is moved to 22nm and is 8 cores for like 50W

[ridney]

this is a very dirty run, i promise tomorrow i will turn off turbo and let all six cores do the talking

260HTT is possible Turbo on auto settings

[ridney]

POV-Ray 3.7

The 1055T with a hefty NB OC should be particularly impressive in POV-Ray as the benchmark is very sensitive to memory access. This is why Nehalems get a 20%+ boost from HT in it. On a 965 @ 4Ghz, taking the NB from 2000 to 2600 and timings from 9-9-9-24 to 8-8-8-20 result in a 20%+ boost. I'd expect a 1055T @ 4.2Ghz + 3Ghz NB to perform very close (6000ish) to if not actually exceed the 980X (6400).

x264 HD Benchmark

You'll need Avisynth to go with x264 HD.

3DS Max SPEC

thanks for heads-up on this amd-c, i will do my best to get near that score

Woo hoo man,nice update .Your single thread score is identical as if you have OCed your chip to 3.3Ghz and assigned only one core in affinity settings then rerun the test ,so the Turbo CORE works flawlessly,a perfect scaling from 2.8->3.3Ghz !!! And it's out of the box feature,nice going AMD

that's great informal! turbo is really amazing and i'm really impressed with amd on this one

This is why I have 2 laptops and I'll be putting my 965 into a secondary desktop once I get my hands on an X6 =)

Great results. Can't wait to see more.

i am waiting for bobcat, i'm itching on a core i5 but must resist

Nice, looks like AMD wasn't lieing about about how quickly it will toggle turbo when threads are shifting from core to core. Thanks for the screen shots and the effort.

sure charged!

wooot that was a detailed view. thanks

You might wanna try TPMonitor instead of opening multiple instances of cpuz

tried but wasn't working correctly, thanks jc

how do you get cpuz to show each core?i tried this and it shows the same core0 for all instances

right click on the box and select a specified core#

Nice work ridney. Thanks for taking the time to run all those tests then posting up the results. Anyother time I would call that a bunch of work but in this case it being a new toy I figure maybe it was more like a bunch of fun.

Thanks oldtimer! it really takes time to test things and we feel like it's work but we enjoy these things, especially that it's a new technology so in way it's a lot of fun. cheers!

[jcmarfilph]

20k should be fairly easy if Turbo Core is off.

[josh1980]

I love that the processor box says "True six core design" on the side. That statement is true, just like when AMD released their first "True quad core design", competing with Intel's 2xCore Duo processors glued together. Of course, about 2 years later the AMD top executives admitted that creating a "true quad core design" was a big mistake and they should have just shoved 2 dual core processors together like Intel.

[informal]

A true guad core was created with servers in mind and not desktops(you couldn't scale the hypothetical MCM to >4P due to HT links limitation).

[radaja]

ridney,nice results so far.i cant wait to get one.six cores is going to be fun and affordable

right click on the box and select a specified core#

thanks for showing me how to do this.this is a great feature that i had no clue about

[ajaidev]

A true guad core was created with servers in mind and not desktops(you couldn't scale the hypothetical MCM to >4P due to HT links limitation).

I am sure they will when the process is small enough other wise they will have a thermal bottle neck...

[FlanK3r]

ridney: nice result in Cinebench, can u go a bit higher, exmaple with 3700-3800 MHz?

[Titan7171]

ridney: nice result in Cinebench, can u go a bit higher, exmaple with 3700-3800 MHz?

hey save some testing for us

[Opteron146]

a perfect scaling from 2.8->3.3Ghz !!! And it's out of the box feature,nice going AMD

Are u sure ?
Anandtech posted some Cinebench scores here:
http://www.anandtech.com/show/2819/5
A 920 scored 3244 there, not far away from the above 3523.
A 955 with 3.2 GHz scored already 3675, thus the Turbo scaling cant be perfect. Or is there another difference ? OS, 32/64bit, sth. else ?

[informal]

Are u sure ?
Anandtech posted some Cinebench scores here:
http://www.anandtech.com/show/2819/5
A 920 scored 3244 there, not far away from the above 3523.
A 955 with 3.2 GHz scored already 3675, thus the Turbo scaling cant be perfect. Or is there another difference ? OS, 32/64bit, sth. else ?

I checked with my Agena @2.8Ghz ,it scores almost the same in single thread test as Deneb @ 2.8Ghz(Cinebench is one of those tests that don't quite show the difference of improved IPC between Agena and Deneb).

[FlanK3r]

titan:dont fear, i have plan for testing :-) (example:1M,32M,wprime 32+1024M, Cinebench R10 32bit/64bit, Cinebench R11.5 32bit/64bit, maxxpi, winrar bench, x264 bench, practice encoding at AVS and at wme, practice photo editing, fritz bench, 3D05, 3D06, Vantage, LinX, maybe PRIME too, max validation, max NB. All benchmarks with default/default without turbo/OC+turbo/max OC and a bit comparsion with 960T a 965 BE )

[Opteron146]

I checked with my Agena @2.8Ghz ,it scores almost the same in single thread test as Deneb @ 2.8Ghz(Cinebench is one of those tests that don't quite show the difference of improved IPC between Agena and Deneb).

How much does it score @3.3Ghz ? (if you can run it ;-) )
I found a single thread score of an oc'ed X3 720 here:

3822 points with 64bit, i.e. 300 points more than the 1055T with Turbo, which should run @3.3GHz, too.

[jcmarfilph]

@Opteron146:

you have to take into consideration the HTT and HT Link speed. That is why it scored higher than 1055T.

And will all those craps in the screen. I am sure it will go higher than that.

[Opteron146]

@Opteron146:

you have to take into consideration the HTT and HT Link speed. That is why it scored higher than 1055T.

And will all those craps in the screen. I am sure it will go higher than that.

AH thanks .. actually HTT and HT Link speed should not have a big effect, but I just saw the Uncore/NB/L3 clock of the above 3.3 GHz score: 2.5 GHz ... that could give a nice boost ;-)

Is somebody out there who can verify the influence of L3 clocks ?

However, there are still the andantech results, which should be done @standard clocks, and a normal 3.2 GHz 955 scored higher ...

[FlanK3r]

Are u sure ?
Anandtech posted some Cinebench scores here:
http://www.anandtech.com/show/2819/5
A 920 scored 3244 there, not far away from the above 3523.
A 955 with 3.2 GHz scored already 3675, thus the Turbo scaling cant be perfect. Or is there another difference ? OS, 32/64bit, sth. else ?

RAMs...AM3 score is better than AM2+ in most aplliacations, example here:

[informal]

I can't hit 3.3Ghz ,i can hit 3.2Ghz though(2.2Ghz NB,400Mhz DDR2 cas4) :


bare in mind that 720BE has super beefed up NB and RAM ,that alone can account for 4% IPC difference between my run and the run you linked to..

[jcmarfilph]

AH thanks .. actually HTT and HT Link speed should not have a big effect, but I just saw the Uncore/NB/L3 clock of the above 3.3 GHz score: 2.5 GHz ... that could give a nice boost ;-)

Is somebody out there who can verify the influence of L3 clocks ?

However, there are still the andantech results, which should be done @standard clocks, and a normal 3.2 GHz 955 scored higher ...

CnQ was enabled too so that could affect the result.

And correct me if i am wrong, Turbo Core is just spiking on when needed so a static 3.3Ghz > dynamic 2.8Ghz(Turbo Core ON and CnQ ON)