Barcelona OC

In relation to Phenoms coming....


Join Date: Sep 2002
Posts: 5,028
Quote:Originally Posted by BeardyMan
Are you expecting them soon or???
Click to expand...

Join Date: Sep 2002
Posts: 5,028

Originally Posted by s7e9h3n

Yes...very soon...along with B2 Barcelonas
Click to expand...
__________________



:D :cool: Says he will be testing the chips on a new 790 mobo he has.
 
Very interesting how performance scaled with a 400MHz OC. Seems there's something under K10's hood that likes higher clocks and gives better than linear scaling.

untitled.jpg
 
Yes, XtremeSystems has always been a great place for practical real world info. And the signal to noise ratio is far less than most places. They don't tolerate trolls and their FUD trying to muddy the waters of information. And just to reiterate, those Barcelonas are only B1 steppings, with B2 right around the corner. Quite impressive.
 
I'm not familiar... who's this s7e9h3n fellow? He seems awfully popular there.
 
Same exact super pi as me, except im at 2.7 and thats at 2.4. Though I dont know how to account for the memory differences. Mine is DDR500 2.5-3-3-7 (highest I can get at 1t). Discounting all memory differences though its about 13% faster clock per clock than my 939 in superpi.
 
Jakalwarrior said:
Same exact super pi as me, except im at 2.7 and thats at 2.4. Though I dont know how to account for the memory differences. Mine is DDR500 2.5-3-3-7 (highest I can get at 1t). Discounting all memory differences though its about 13% faster clock per clock than my 939 in superpi.
Click to expand...

Don't forget that SuperPi only uses one core AFAIK, and isn't a very good benchmark metric to begin with, so hopefully in real world apps (like gaming, video...) we will see about a 20-30% difference over the old K8.
 
Super Pi is indeed single threaded. I just don't get how this program suddenly goes from stress tester to benchmark overnight, especially since it does both poorly. I can think of dozens of benchmarks that are more indicative or real-world performance. Heck, there are a few programs better/faster at calculating Pi.
 
That doesnt matter. People just use it because it shows processor scaling quickly and easily clock for clock on just 1 core (most of the stuff we run is single threaded so its very important).
Personally I lean towards Sisoft Sandras CPU benchmark though.
 
Aside from the phenom being at 50mhz less on ram (And looser timings here and there) and the test being bad for pure cpu testing... If we assume thats real thats probably about right.
 
That's a major slap to the face that the Phenom X4 loses to the X6850. I was hoping for better. :(

Then again it's only an engineering sample so things could change. *crosses fingers*
 
Manny Calavera said:
http://www.fudzilla.com/index.php?option=com_content&task=view&id=3882&Itemid=1
Click to expand...


A motherboard with a preproduction bios, crappy ram timings for the Phenom with a 2T command rating on top of that??? You call that a valid comparison? :confused:

They need to get some better ram, jack up the ram settings to the tested Intel equivalents, use a motherboard with fully working bios (or a board with the newly announced AMD 770 chipset) and then MAYBE we'll have some legitimate testing here. Once again all I see is shoddy testing of pre-production hardware, i.e.engineering samples... *yawn*

Taking into account the AMD setup was cripped, I'd say it fared well, all things considered. But to take this seriously as an actual idea of what to expect of Phenom once released? Haha! NEXT!
 
Mostly it just makes me wonder how they can claim to know anything about computers and use a benchmark where ram differences matter to compare CPUs while the ram settings aren't equal. Be scientific about your testing or go home. Im not saying that the AMD will win a fair fight. Just saying thats a crappy test.
 
This was an attempt to compare overall CPU efficiency, right? So much for logic and the scientific method. Those benchmarks have more holes than Swiss cheese. And it doesn't take Sherlock Holmes to see the glaring inconsistencies or Richard Petri to notice the lack of a controlled experiment.
 
No it was not 100% fare,but still.

It shows that its not the Care2 beater they said it would be,and certainly not the Penryn beater eithier.

Lowering the ram timings to T1 instead of T2 will not make a any real difference.
 
Manny Calavera said:
No it was not 100% fare,but still.

It shows that its not the Care2 beater they said it would be,and certainly not the Penryn beater eithier.

Lowering the ram timings to T1 instead of T2 will not make a any real difference.
Click to expand...

I thought 1T would make a pretty big diff in applications like Cinebench :p?
 
Manny Calavera said:
No it was not 100% fare,but still.

It shows that its not the Care2 beater they said it would be,and certainly not the Penryn beater eithier.

Lowering the ram timings to T1 instead of T2 will not make a any real difference.
Click to expand...

It's not the fact it's not fair, it's the fact the TESTING IS FLAWED Manny, it's that simple. Until a final production sample of Phenom is tested at comparable Intel equivalents settings with fully working hardware, that comparison is WORTHLESS and doesn't amount to Jack Schitt. I'm sure I'm far and away from being the only person who sees it that way either. Like I said before, put both of them on even ground hardware and setup wise with some REAL scientific testing and then get back to us.
 
We aren't ripping you. Just the people who did the benchmarking ;) Even a badly done benchmark beats the wild speculation that normally fills the forums.
 
That's cool and all, but it doesn't do much good( I can see a little) if intel still keeps a better price/performance ratio
 
nonameo said:
That's cool and all, but it doesn't do much good( I can see a little) if intel still keeps a better price/performance ratio
Click to expand...

Uhhhh YES, it does alot of good for AMD. If this software can enable AMD quadcores to run much more cooler and efficiently by adjusting power or clock frequencies for specific tasks than Intel's offerings while still holding a good price performance ratio, it's a win/win situation, period. This software is putting AMD's new superior multi-plane power and clock distribution system to very good use, I know system admins and power users will love it.
 
Lou Ceifer said:
Uhhhh YES, it does alot of good for AMD. If this software can enable AMD quadcores to run much more cooler and efficiently by adjusting power or clock frequencies for specific tasks than Intel's offerings while still holding a good price performance ratio, it's a win/win situation, period. This software is putting AMD's new superior multi-plane power and clock distribution system to very good use, I know system admins and power users will love it.
Click to expand...
This is like the whole HTT vs FSB argument. It doesn't matter if what AMD is doing is more elegant if Intel can get superior results from their less elegant solution.

I eagerly await hard numbers on both performance and power consumption.
 
LstOfTheBrunnenG said:
This is like the whole HTT vs FSB argument. It doesn't matter if what AMD is doing is more elegant if Intel can get superior results from their less elegant solution.

I eagerly await hard numbers on both performance and power consumption.
Click to expand...

Bad analogy/argument for the server market. Power consumption is crucial in data centers, i.e. servers. Even if AMD is not quite up to Intel's performance or is slightly below, it's power conservation hardware/software will easily help them earn sales if AMD prices them right to Intel's offerings if they run far hotter or consume more power. Another issue is scalability and right now, Intel is still losing that battle at 4 sockets and above due to it's dated FSB, Barcelona and HT3.0 will only improve AMD's server offerings, which is what K10 was really designed for.

On the desktop market, your comments would have an valid argument as AMD and Intel's design philosophies differ with their current respective architechtures: AMD designs their chips primarily from a server standpoint for performance, Intel, from the desktop market's perspective for performance.
 
You guys do know that Independent Voltage doesn't work on separate cores right?

Even if the cores are different clocks, all the cores have the same voltage as the highest core.

If core 3 is 3.3Ghz but is using 1.4V, then all 4 cores will be 1.4V even if they are at a lower speed.

It's just looks "cool" but isn't practical.

Whats the point of cutting down clocks when the power state is still Dependant on the highest core.
 
n19htmare said:
It's just looks "cool" but isn't practical.

Whats the point of cutting down clocks when the power state is still Dependant on the highest core.
Click to expand...
Temperature control and overall stability are two reasons.
 
n19htmare said:
You guys do know that Independent Voltage doesn't work on separate cores right?

Even if the cores are different clocks, all the cores have the same voltage as the highest core.

If core 3 is 3.3Ghz but is using 1.4V, then all 4 cores will be 1.4V even if they are at a lower speed.

It's just looks "cool" but isn't practical.

Whats the point of cutting down clocks when the power state is still Dependant on the highest core.
Click to expand...

Not true. Each core has it's own voltage plane. That is one of the biggest differences between AM2, and AM2+..... Besides I think this will really help overclockers out as well. If you have 4 cores, and one of them ias capable of hitting 3.4GHZ, and another is only capable of hitting 2.9GHZ, you can clock them separately without being held back by the slower core.
 
duby229 said:
Not true. Each core has it's own voltage plane. That is one of the biggest differences between AM2, and AM2+..... Besides I think this will really help overclockers out as well. If you have 4 cores, and one of them ias capable of hitting 3.4GHZ, and another is only capable of hitting 2.9GHZ, you can clock them separately without being held back by the slower core.
Click to expand...
the SPP / IPP splits power between the IMC / 4Cores, not core to core from what I read,
 
duby229 said:
Not true. Each core has it's own voltage plane. That is one of the biggest differences between AM2, and AM2+..... Besides I think this will really help overclockers out as well. If you have 4 cores, and one of them ias capable of hitting 3.4GHZ, and another is only capable of hitting 2.9GHZ, you can clock them separately without being held back by the slower core.
Click to expand...

You're wrong.
http://multicore.amd.com/us-en/AMD-Multi-Core/Quad-Core-Advantage/Power-Efficiency.aspx
 
http://www.hardwaresecrets.com/article/480/4

There's a good article on the features. It does indeed only have two voltages (cores + imc); however it does also allow for independent clocks on each core, which can effectively bring the power consumption of each core independently down as well. Also the ability to shut off functional units (not just the entire core) can be huge in the way of power savings.
 
You must log in or register to reply here.
Back
Top