THG Athlon FX @ 2.8 GHz

Interesting, very interesting. Thank you for the post, it looks like its doin one heck of a job. I would really like to see it benched against the p-4ee overclocked

The man of steel said that

It's also the best competition that intel can offer RIGHT NOW. However, before jumping to conclusions about the FX (opteron) vs scotty, we really should wait for a scotty.

on a different note...that article is really making me like AMD more and more...i dont care if anyone thinks that OCing in a published benchmark isn't fair or right, it still interests me because I would also try to overclock. Should scotty not quite be up to the task of taking of the FX or even the A64, AMD might have one intel fan converted.

RDRAM = ENEMY

Ok so when is THG gonna pull its head out of its rear end and include both the OC'd P4EE and A-FX in the same review, instead of mixing and matching with one or the other???? Damn fooools!

Arguing on the internet is like running in the special olympics. Even if you win your still retarded!

I quite liked the review, but I got one question allready from the original A64 / fx review. I was delighted to see you had included nuendo in your selection of test software. I was really suprised to notice the P4's where doing so well in it, as it's using the same engine as cubase that has traditionaly performed better with athlons.

I then noticed that you are only testing the time to render the final project, which is generally not the feature that requires alot of power in a sequencer. I know it can be alot harder to do benchmarks with a sequencer in a real world situation, but this is what a large music making community would really need. Most music makers I know would be intrested in seeing how many eq's, reverbs, compressors and softsynths can be run before the cpu starts spiking, and none of them care if rendering the song to a wave file takes 5 or 10 minutes. (as opposed to 3d rendering where the differance in rendering time can mean hours or days)

I feel a bit silly giving critical feedback on my first post, but there is alot of people who would love to see this kind of benchmarks.

What's really interesting to me is that the 2.8GHz limit appears to be imposed by the BIOS.

Note, the multiplier could apparently be taken only from 11 to 14 (2.2GHz to 2.8GHz). I'm not 100% certain on this, but the translation reads as if the BIOS simply didn't have a higher selection. I'd be interested in seeing what happened if the multiplier got taken higher...it looks like the processor could probably take it!

<i>I can love my fellow man...but I'm damned if I'll love yours.</i>

Anyone have a Athlon FX to spare???

go and mess with the bridge and pinout to fix the multiplier to 15 or higher!



--
Would you buy a GPS enabled soap bar?

OMG!

Overclocked AMD chips and no overclocked Intel chips?!?!?!

TOM'S IS AMD BIASED!!!!

I AM LEAVING AND NEVER COMING BACK!!!!

AAAAARRRGGGHGGHHHHH!!!!!



<font color=red> If you design software that is fool-proof, only a fool will want to use it. </font color=red>

Tried to compare the overclocked P4 results with the overclocked Athlon FX results. Many of the benchmarks don't match up though, and both used different videocards and what not.. you can still have a look here though:

<A HREF="http://home.pi.be/~bbaeyen2/P4FX.htm" target="_new">http://home.pi.be/~bbaeyen2/P4FX.htm</A>

In bold you can see performance scaling per clockspeed. I tried with both P4C 3.06 and P4 3.0 GHz since the overclocked P4 uses a FSB somewhere inbetween, but it hardly changed a thing. The FX scales a lot better, especially in games..

ignoring the (important !) videocard issue, it would roughly take a 5 GHz P4 to catch the 2800 MHz opteron in Q3, a 5.7 GHz P4 to match it in Splinter Cell and a 7.3 GHz (!) to match it in UT2k3. take these numbers with a bag of salt, but you get the picture...

= The views stated herein are my personal views, and not necessarily the views of my wife. =

Don't forget the fact that most P4 can't get to 4G. Unless you have some exotic cooling and really lucky.

-Intel PIV 2.6C @ 3.51G -Asus P4P800 -OCZ Copper 2x256 4000EL memory @ 270mhz 2.5-4-4-8 -Geforce2 MX400(i know i know) -SB audigy -120G Maxtor Diamond Plus9 S-ATA150 hdd -450 Enermax PSU

The xeon scaled poorly while the opteron was close to perfect

result of lack of clock speed in vector processing and decreasing lantency in ram.

I dont like french test

You dont consider what they used to get a 27% OC out of the FX "exotic cooling"?
I've got a 2.8C, and it OC's to 3.5ghz on stock cooling & voltages - surely its not that hard to get a 3.2C up to 4ghz & beyond to test against the OCed FX

I'd say the Opteron scaled so well thanks to Hyper Transport and its non-uniform memory addressing.

--
<A HREF="http://www.lochel.com/THGC/album.html" target="_new"><font color=blue><b>This just in, over 56 no-lifers have their pics up on THGC's Photo Album! </b></font color=blue></A>

it quite easy to prove the athlonFX have shown great improvement in 3d rendering where it use vector instruction like SSE-2.SSE1-2 or SSE 100 super advance super 45 dual blabla are all scalar so it just clock speed that made the difference exception to a others bottleneck.As there also much less instruction lantency on a K8 so the ratio instruction latency/clock speed is better that the one of a P4 3.2.If we put a P4 prescott even longer pipeligne they result will be worse but the extra cache and clock speed will put it as a winner until the ratio increase on Athlon FX again.

In short word

a cpu at 10 ghz with 100 stage instruction will take 100 cycle before it got finish.
a cpu at 1 ghz with 10 stage instruction will take 10 cycle before it got finish.So a 1.2 ghz on 11 stage will be the fastest.

In short term this will end soon and P4 will worsen until more cache is add and faster FSB are put some apllie for Athlon if i remember well willimete was scaling well for the first 600 mghz thing have decrease.There the intergrade controleur will help as lantency from ram decrease with increase of clock speed (less that few % but better that nothing)

If in any case the memory bottle hit 1 day or a others.

I dont like french test

> its non-uniform memory addressing.

Hu ? another BS (TM) moment brought to us by Eden ?
You are aware that NUMA only plays a role on SMP systems ? that there are no x86 OS's that properly support it (at least AFAIK) ? Do you even know what NUMA is ??

= The views stated herein are my personal views, and not necessarily the views of my wife. =

Not really. Samba 3.0+Linux still rips the hell out of Win2K3.

And fastest (by far) Itanium on CPU2000 is still an Altix 3000 running Linux.

Best part is, Linux also has a current full-featured AMD64 port.

<i>I can love my fellow man...but I'm damned if I'll love yours.</i>