K8 vs K10 on 790FX - Results

[OBR]

on my weird Gigabyte mobo OC with K8 - 400MHz HT, with Phenom only 230MHz ... thats strange

[Tony]

Not strange if you know whats going on...working with motherboard vendors now to add NB multi adjustment to bios, this may allow upwards 230HTT.

[KTE]

Not strange if you know whats going on...working with motherboard vendors now to add NB multi adjustment to bios, this may allow upwards 230HTT.

It doesn't with me. I've had HT and NB to 1x and 5x or even to 10x and 11x and the results are the same. Not 230HT, that's very low, but 266-268HT.

[JohannesRS]

Really great job! Congrats!

Can I ask for some sciencemark and f@h benchies?

[VulgarHandle]

mAJORD, great work, thanks for all the hard work and patience with the trolls

wanted to ask a big request, can you disable 2 of the cores and run the same tests to see how it truly compares to dual-cores, or even single-cores..., and if you really want to try something, try disabling different combinations of cores to see if there are cores that work better with others. More important would be at least disabling 2 cores. If you do so, you'll have my gratitude, as well as gratitude from many here.

[Vapor]

Also agree this is great work mAJORD....not every day someone takes hardware and objectively runs so much stuff on two setups

And the differing settings don't bother me so much...the 1-2% isn't a big deal (though exact settings isn't a bad thing ).

But I'd also agree that forcing the K10 into 2-core mode would be a very good thing to do....and it will open up the options for various benchmarks. 3D06 CPU score, WinRAR MC, Cinebench MC, etc, will all be fair game

The easiest way, IMO...is to go into start -> run -> msconfig (hit enter) -> BOOT.INI tab -> Advanced Options -> check /NUMPROC= -> 2 -> close and restart

Allows any program to open any number of processes it wants and still limits them to 2 cores being used. Also doesn't allow GPU drivers to go on the other cores either...something that may effect 3D06.

Can't wait for more tests, esp video encoding and SSE stuff

[Teccxpert]

I'm watching this thread with great interest... looking at quad core upgrade... not sure to dump E6850 & X38 mobo...

[Blacklash]

The thing that will probably move me toward Penryn is not the clock for clock speed difference and rather the 1k overclocking advantage it will likely have on air. (3.0GHz vs 4.0Ghz) I won't even get into more elaborate cooling.

Still, I might build another rig with a 9600 BE anyway.

[mAJORD]

Hey Guys, Sorry I haven't updated this. some IRL stuff has come up , including having to fix my car!

I've got results for more stuff just need to put them up

SmartAss, hopefully having a little LAN here so should be able to get a bunch of games on it off a friend. So that'll be nice.

[adamsleath]

will phenom x2's oc well do you think?

if so, that should start to displace the k8 x2's conclusively i wooda thought.

tanxs for the results.

[Lightman]

I did some tests for you guys simulating Phenom X2 2200MHz with DDR-800 4-4-4:



CPU Queen 4395 vs 4262 of X2 4200+ 2.2GHz


CPU PhotoWorxx 26375 vs 17902 of X2 4200+ 2.2GHz


CPU ZLib 29487KB/s vs 26238KB/s of X2 4200+ 2.2GHz


FPU Julia 4660 vs 3742 of X2 4200+ 2.2GHz


FPU Mandel 2308 vs 2114 of X2 4200+ 2.2GHz


FPU SinJulia 5671 vs 5872 of X2 4200+ 2.2GHz


Cinebench 2003 x64 + SuperPi

[hollo]

oh, vs the built-in results for s939 x2 4200 with ddr400

julia tests single precision floating point (32 bit), mandel tests double precision floating point (64 bit), sinjulia tests extended precision floating point (80 bit) afaik... k10 loses to s939 in extended precision (exponential and trigonometric functions according to my everest help file)?
i bet extended precision trigonometric and exponential functions don't get used very often :p

nb speed 1.8 ghz?

[Lightman]

oh, vs the built-in results for s939 x2 4200 with ddr400

julia tests single precision floating point (32 bit), mandel tests double precision floating point (64 bit), sinjulia tests extended precision floating point (80 bit) afaik... k10 loses to s939 in extended precision (exponential and trigonometric functions according to my everest help file)?
i bet extended precision trigonometric and exponential functions don't get used very often :p

nb speed 1.8 ghz?

Yes, NB 1.8GHz is default for Phenom 9500!

Slower SinJulia results might be caused by x64 OS.... By default you can't use 80bit FP, only 64/128bit... I will re-run this test under Win XP MCE to confirm that...

[awdrifter]

With these new boards that can do over 1000HTT, would we see any performance increase on an K8 by using say 2000HTT instead of the stock 1000HTT? Can someone do a quick test? Thanks. Great thread btw.

[BadNizze]

With these new boards that can do over 1000HTT, would we see any performance increase on an K8 by using say 2000HTT instead of the stock 1000HTT? Can someone do a quick test? Thanks. Great thread btw.

My DFI 570 ultra did 1600 HTT easy, prime stable. I did run 8*400 for a while but did not find any preformace leaps over my 6400+´s stock 16*200.

[mAJORD]

Hey Guys I've updated the first Post with new benchies

Lightman: Thanks for your Additions, do you want me to paste that into the main thread under your name with a link to post?

[informal]

Thanks mAJORD .You have done an amazing job .

edit:i saw you 've seen Lightman's post about GB "bios good news"

For everyone else,Lightman managed to break 240HTT limit for his Phenom and GB board.Post here

[JohannesRS]

Thanks for the sicencemark results!

Found it strange to see that the results points toward equivalent performance in everything, except blas performance. Strange behaviour, any clues?

Can I ask for an F@H bench please? Thanks a lot!

And really good work, that seems to be the best job until now on such a subject. Do you plan to redo everything with B3 if it starts to show strong differences in performance from B2?

[awdrifter]

My DFI 570 ultra did 1600 HTT easy, prime stable. I did run 8*400 for a while but did not find any preformace leaps over my 6400+´s stock 16*200.

Cool, thanks for the test.

[Astennu]

Very nice tests !!

To bad that Phenom's clockspeeds cant get that high like the 3200 of the K8. Hopefully AMD can fix dit and push phenom above 3200 within 120 watt TDP.

[nemrod]

Hey Guys I've updated the first Post with new benchies

Thanks!

I believe there is 2 small mistakes. For Systool, the results are invert:
In fact from your screenshoots:
Phenom: 237sec
Athlon: 263sec
So K10 + 10.9% is still valid


But the other about x264 Benchmark:
I don't see time on your screen, but fps seems to shows that time you give after is good.

Phenom:
PASS1:89.00
PASS2:17.58

[106.58]

Athlon:
PASS1:75.13
PASS2:17.97

[93.1]

So it should be K10 -14.4% overall (in this test K10 underperform K8)

x264 Benchmark

Phenom:
PASS1:89.00
PASS2:17.58

[106.58]

Athlon:
PASS1:75.13
PASS2:17.97

[93.1]


K10 +14.4% overall


Systool
Phenom: 263sec
Athlon: 237sec

K10 + 10.9%

[informal]

@nemrod

How can K10 underperform K8 when it has a lot more fps in x264 test?

[KTE]

Interesting, great work, thanks.

I did a X2/C2D/X4 comparison at similar MHz (2640MHz) but I was too short of time to replicate much and the Phenom was giving me NB problems to continue on. Plus I couldnlt get 2k6 to run on the C2D build no matter what.

I might post my results in my K9A2 thread later, but I don't see the point since you have already done most of it.

Problem is, if you try and get a high NB fre. like X2 has you start seeing many similarities in perf.

[nemrod]

@nemrod

How can K10 underperform K8 when it has a lot more fps in x264 test?

I was thinking it was time for compression. For fps, I was looking at the second pass. But just an error


Anyway, this benchmark seems dependent of the number of core:
http://www.techarp.com/x264_Benchmar...ts_amd_big.gif