Barcelona Opteron 2350(B1) arrived

[mrcmtl]

For dual-channel mode Spi 1M time was 38.968, and it is clear that overclocked results were taken with memory in dual-channel mode.
So, 38.968 / 32.672 = 1.1927 = 19.3% it is.

can't you see the Single word written on cpu-z at the ram tab?

[Sparky]

How come no one shouted out OMGZ !!! when steven said B2 Barcelonas...are people THAT disconnected when it comes to amd 0_o...

B2 Barcelonas = penryn fighting power people.

Perkam

I'm just not getting my hopes up, I'd love B2 to be much better, but if I don't get my hopes up then there isn't the chance of me crashing to the ground in case my hopes were in vain.

can't you see the Single word written on cpu-z at the ram tab?

It was set to dual channel in the bios, CPU-Z seems to be reporting it wrong (as it is the HT link as well).

[perkam]

I'd love B2 to be much better

Then let me tell you in advance, it will

Perkam

[freeloader]

It is 7 second faster in super pi with just 400mhz up
Awsome

I noticed that too. Quite a good showing for a small clock speed increase.

[PhilDoc]

For dual-channel mode Spi 1M time was 38.968, and it is clear that overclocked results were taken with memory in dual-channel mode.
So, 38.968 / 32.672 = 1.1927 = 19.3% it is.

Yep, I finally saw it. I was just looking at the screen shots. Not totally sure he's actually running dual channel as cpu-z show single. It could be that cpu-z needs an upgrade for Barcelona. The difference in times may be just fluctuation in the spi test. We'll find out soon I suppose.

[tictac]

3.88 speeup on cinebench.. is it good enough compare to c2q?

[KTE]

Actually it should be ((39.750-32.672)/32.675 = 21.7%.

That's 108% scaling. Its incorrect because you're not working out percentages here, but percentage change. The formula is as follows

Maths:
http://www.gcseguide.co.uk/percentages.htm
http://mathforum.org/library/drmath/view/63055.html

New - Old / Old x 100 = Percentage Change

KTE, your math is wrong. Also, the SuperPi 1M time is 38.968s for 2GHz. Read the first post.

For single channel it's 39.750s for dual channel its what you said: http://www.xtremesystems.org/forums/...22&postcount=1

For simplicity, we swap the top figures around or it would become a minus change this time because of importance is decreasing values, not increasing (which is practically untrue i.e. to us 7 sec dropped is not 7 disadvantage but 7 advantage). The base always stays the same as its "new" relative to "old". What matters is the change.

38.968s - 32.672s (change) / 38.968s (relative to old) x 100 (percentage) = 16.157%

If you use that first general formula it would be -16.157% change which is perfectly true from the original time. But to us that's a +16.157% performance gain.

Now we can use a simple percentage formula:

16.157 / 20 x 100 = 80.8% scaling.

[Cronos]

3.88 speeup on cinebench.. is it good enough compare to c2q?

It is unreachable for С2Q. The main reason is Core2 shared cache.
This is also one of the main reasons for excellent single-threaded performance of Core2 cpu's, which -in turn - is unreachable for comparatively clocked K10, regardless of stepping.
The net result is that C2Q is excellent CPU with very high SMP performance. It's
"untrue" nature does not hurt its performance to any noticeable extent. No reason for Intel to hurry up with "native" solution -at least, not from performance stand point.

[metro.cl]

Hammerhead is an AM2+ platform, not socketF. I'm waiting for a Phenom to run on the board. I could run an AM2, but what's the point?

Yep, you actually want a Wahoo board for dual f1207 sockets

[kyosen]

FELIX kindly pointed out to me about K10's memory controller
According to AMD's document, BIOS and Kernel　Developer’s Guide,
K10's has 2 DCTs(DCT: Dram ConTroller!?), and it can work different two mode, i.e.,
(1) to behave as a single dual-channel DCT; this is called ganged mode;
(2) to behave as two single-channel DCTs; this is called unganged mode.

My understanding is:
Ganged mode reads or writes with 2 DCTs at the same time.
Unganged mode also can reads with 1 DCT and writes with another 1 DCT at the same time,
if the system requests as so.
Unganged mode is more flexible than ganged mode,
and ganged mode is equal to "conventional" dual channel mode.

Then, I noticed the BIOS setting, "Unganged Mode support".


So I tried SuperPI 4M in 3 case as below, and results:
with 1 memory module: 3m27.141s
with 2 memory modules, ganged mode: 3m24.438s
with 2 memory modules, unganged mode: 3m22.953s
and we can see CPU-Z and FELIX's test program properly detects
"conventional" dual channel mode
Default BIOS setting is "Unganged Mode = Enabled", and at least,
unganged mode looks faster than ganged mode for SuperPI.




And these are latecy test results with Everest Ultimate v4.20:
at 2.0G=200x10: 76.0ns
at 2.2G=220x10: 69.0ns
at 2.4G=240x10: 63.4ns
The latest Everest detects Barcelona properly




Thanks to everyone for many comments and suggestions
...but don't be bullying poor small AMD, hipro5 (yeah, just kidding )

[accord99]

38.968s - 32.672s (change) / 38.968s (relative to old) x 100 (percentage) = 16.157%

What if the faster score is 20s? With your equation, it's only a 49% increase in performance.

Or look at it this way, the 2GHz does 1M digits in 38.968s = 25,662 digits/s; the 2.4GHz does 1M digits in 32.672s = 30607 digit/s. The 2.4GHz does 19.2% more digits/s.

[kyosen]

Please calm down about scaling for SuperPI time...
I'll try it again, with K8 Opteron 2212 and K10, with same OS

[Cronos]

We are all calm. Somebody just needs to learn their math lessons, thats all

It is much more interesting to see multi-threaded scores of dual-cpu K10 system,
as it is there AMD is going to compete and win against Intel -not on single-socket platform.

[zakelwe]

ny chance of a pifast bench? That was always more favourable to AMD than SuperPi and also seemed a bit more relevant to general useage.

Regards

Andy

[JumpingJack]

It is 7 second faster in super pi with just 400mhz up
Awsome

Actually, it is about 20% faster for a 20% faster clock... this is expected... if it didn't something is very wrong.

[RedBull78]

Wow, not bad actually. I get 32sec @ 2.9ghz on opty 165. So that means a 3ghz barc. should do around 25 sec, not bad at all.

Interesting though, how a 6000+ A64 is par with a C2d 6600 model in real time performance but in SuperPi the 6600 smokes it. But in this case a 2.4ghz Phenom should smoke the hell out of the 6600 out the water in real time applications.

[RedBull78]

Actually, it is about 20% faster for a 20% faster clock... this is expected... if it didn't something is very wrong.

It depends how you benchmark each application, most of these here are singel threaded.

[STEvil]

how about kribibench?

http://www.xtremesystems.org/forums/...89#post2476489

[KTE]

Great results kyosen. Keep 'em coming.

[mongoled]

Great results kyosen. Keep 'em coming.

Agreed, now we want higher clocks
Higher clocks = Answer to the important questions

[tictac]

can you do same comparison on the memory mode with 64bit cinebench

[mAJORD]

These are quite promising, Thanks aswell Kyosen

[hipro5]

Thanks to everyone for many comments and suggestions
...but don't be bullying poor small AMD, hipro5 (yeah, just kidding )

I won't for sure.....

[savantu]

The L3 latency is 25 cycles , that's a lot.

[Lastviking]

Some fast testing for showing K8 improvement over mhz

1 year old WinXp install for 24/7 os (thats why the times are slow and also 4x512mb and not tweaked)

Dfi Venus
Opty 165
4x512mb bh5

Used clockgen in windows to lower and higher the clocks"not same memmory speed and fsb, like the k10 scores"

54,907s 1600mhz 177fsb
43,641 2000 222
36,266 2400 267 ~7,375 improvement
34,843 2500 277
33,453 2600 289
32,203 2700 300
31,078 2800 311
30,000 2900 322
28,938 3000 333
28,047 3100 344 ~4,156 improvement from 2700mhz