Quote:
Originally Posted by demonkevy666
Scaling is definitely Not linear with SMT or CMT designs and, in multithreaded designs, it is OS dependent especially between Vista and 7 where 7's scheduler is much better.
Printable View
Quote:
Originally Posted by demonkevy666
Scaling is definitely Not linear with SMT or CMT designs and, in multithreaded designs, it is OS dependent especially between Vista and 7 where 7's scheduler is much better.
Because it could.Quote:
Originally Posted by demonkevy666
Older Fritzes didn't support >12 threads, or they ran every CPU at 8 instances.
Clocks sound decent so far. I expect IPC to be wildly variable compared to K10 but hopefully it will have a significant increase on average.
what do you mean wildly variable compared to K10?Quote:
Originally Posted by Solus Corvus
I mean two things. First that the types of code in which BD does well will be different than the type of code in which K10 does well. Second that there will be more factors that affect the actual IPC achieved at runtime when compared with K10. I'll gladly elaborate on either/both points if you want.Quote:
Originally Posted by demonkevy666
Solus Corvus I would like to hear both points after all nothing new is known about BD so why not.
Ok. My night is winding down for now so I'll post after I get some sleep.Quote:
Originally Posted by TESKATLIPOKA
Article at donanhimber states that CPU they used is 3.2ghz, but it has 3.6ghz turbo ON ALL 8 cores :) ,and 4.2Ghz at 4 cores.Quote:
Originally Posted by Dresdenboy
So in reality cpu ran 3.2->4.2ghz in those tests, depending on the software.
DH's ES has no Turbo working,or not properly working at least.It's quite obvious when you look at super pi result(lower than what that Czech guy posted by exactly 1Ghz :19s @ 3.2Ghz for DH ES and 14.5s @ 4.2Ghz for "blogger's" ES).
The said blogger had the exact same problem with turbo when he was using GigaByte board, turbo on Asus board was working fine.
But AIDA shows 4.2GHz: http://www.donanimhaber.com/islemci/...onuclari_8.htmQuote:
Originally Posted by informal
However, there are more than one way to throttle SuperPi performance, e.g. memory configuration.
I made a quick search for results to compare:
BD ; 2600K ; 980x
3dmark 2011 P6265 ; sorry, couldnt find score of a stock 2600k with a stock 580 :/ ; P6383 (990x)
http://www.hardwareheaven.com/review...3dmark-11.html
Fritzchess 14197 ; 12834 ; 16046
http://en.inpai.com.cn/doc/enshowcon...44&pageid=7680
Pcmark 3045 ; 4750(ssd) 3232(hdd) ,too much variables to put much faith into this one
http://www.benchmark.pl/testy_i_rece...ona/14341.html
Cinebench 10 24434 ; 23102 ; 25881
http://www.hardwarecanucks.com/forum...review-10.html
SuperPI 19.5s (dubious score as 1090T does better) ; 10s ; 11,6s
http://www.overclockers.com/intel-i7...-bridge-review
x264 136/45 ; 99/36 ; 96/48
http://www.legitreviews.com/article/1501/13/
Comparison Aida64 screens
BD
http://img.donanimhaber.com/thumbnai...5d8a05_600.jpg
1090T
http://i51.tinypic.com/p0y1f.jpg
2600K
http://alienbabeltech.com/main/wp-co...8P67-2600K.png
Deneb
http://img199.imageshack.us/img199/4774/cachememr.png
Cache write speeds are bad, im worried its gonna be the case with final too.Theres some info about "fixes" coming in BD ver 2 (komodo).
The ver 2 means Engineering Sample ver 2 not komodo.Quote:
Originally Posted by XRL8
These are just ES scores,not final.So until you see retail score so bad in memory benchmarks(as in AIDA),don't put so much faith in these results.
As for BD ver2 it's actually Komodo,not ES version 2. Komodo is known as BDver2 in some compiler suit updates as Dresdenboy showed us many months ago. It is a souped up BD(Orochi) with new ISA support and some tweaks(probably similar to Deneb=>Llano ).
So that's OBR who has given screens to donanimhaber ?
That's the thing he says in his blog :
http://obrovsky.blogspot.com/2011/07...are-obeti.html
That guy always full of nonsense, how can he prove the chip which DH have is from him?Quote:
Originally Posted by Olivon
Well taking this into account:Quote:
Originally Posted by informal
It would make much sense that AIDA write scores are lower.Thats a B1, B2 is going to be retail (if not even B1) if they want to keep schedule.Quote:
http://support.amd.com/us/Processor_TechDocs/47414.pdf
The following performance caveats apply when using streaming stores on AMD Family 15h cores.
• When writing out a single stream of data sequentially, performance of AMD Family 15h
processors is comparable to previous generations of AMD processors.
• When writing out two streams of data, AMD Family 15h version 1 processors can be up to three
times slower than previous-generation AMD processors. AMD Family 15h version 2 processor
performance is approximately 1.5 times slower than previous AMD processors.
• When writing out four non-temporal streams, AMD Family 15h version 1 can be up to three
times slower than previous AMD processors. AMD Family 15h version 2 processor performance
is comparable to previous AMD processors.
• Using non-temporal stores but not writing out an entire cacheline may cause performance to be up
to six times slower than previous AMD processors.
I mean, AMD ITSELF states that its going to be slower in some cases :-/ .Not like its going to be huge thing, other scores prove that it isnt affecting HUGELY performance.
It's single threaded when you do the cache and memory bench.Quote:
Originally Posted by Dresdenboy
I know. This was an answer to informal regarding SuperPi (also single threaded) and his assumption of a not functioning turbo.Quote:
Originally Posted by demonkevy666
It could be that Turbo was working,but I still think something was borked in that regard. If indeed it was obr who gave the screens to DH,then the results don't make sense since his previous scores are way better than DH's.
AMD Bulldozer FX-8130P and FX-8110 Processors Available For Pre-Order :hm:
http://www.legitreviews.com/news/11064/
Well i could take a pre order too.Does he state shipping date ?
That's way cheaper here :Quote:
Originally Posted by XRL8
http://s.taobao.com/search?q=amd+fx-...q-amd+fx-8130p
:D
Lol, thats are some samples only for slaes, not retails