Read my previous post on that subject.AMD's design is fine(nothing is perfect,remember that) and superPI is not used for anything meaningful anyway.Quote:
Originally Posted by Glow9
Printable View
Read my previous post on that subject.AMD's design is fine(nothing is perfect,remember that) and superPI is not used for anything meaningful anyway.Quote:
Originally Posted by Glow9
But if AMD is lacking in scores because it's cache sucks or is lacking it doesn't make software biased. Thats like claiming a game that is designed a specific way shouldn't be valid either. The game will be played on either system correct? Well since the game may prefer more cache then you can't blame the software either.Quote:
Originally Posted by roofsniper
If Deneb has 8mb of cache I doubt this will be a problem do you? So because Phenom lacks a design feature we can't use it to benchmark?? I'm going to find this extemely ironic if Deneb shows really good results in super pi then all of you are like oh well it's a good program to benchmark.
The more you post the more clueless you sound.Quote:
Originally Posted by Glow9
Games do not fit inside the cache nore do most useful everyday applications, so superpi is of no use.
I guess you heard the age old saying its not the size but how you use it.Quote:
Originally Posted by Glow9
Its been said many times in this thread already (and countless others), that the reasons why super pi is better on an intel chip. And why its a useful benchmark when used correctly.
@ roofsniper
I think more of the point is that anyone who understands how a "multicore" CPU works with each manufacturer, will also understand the difference on how single threaded and multi threaded software performance are influeced by the differences in architecture .
I really am looking forward deneb retail.Its good to see AMD are refining their current lineup. Reminiscent of the T-Bred A+B (Phenom65nm) > Barton (phenom45nm). > A64 (??32nm ) dont you think ?
Exactly!Quote:
Originally Posted by Final8ty
I've been trying to explain this to him (and to some others) to no avail.
I am yet to see or hear from someone who actually uses application based on Cinebench engine or Povray or "Fritz engine" ; yet to see one good use of SuperPI(which BTW SUCKS in efficiency when it comes to actual pi calculations since there are a bunch of programs that kill SuperPi when it comes to its own purpose .If anyone knows or have some info it's appreciated.
@Glow9
Instead of skipping on some posts in this thread you should read up and stop posting ehm obviously simple questions or make unsubstantiated claims.FYI ,K10 has smaller transistor budget for core die area and less on-die cache than C2Q and yet manages to perform close to or above C2Q level.That's called good engineering if you're looking for appropriate term.
This is what peeps seem to not understand between the 2 chip designs. I think Phenom 65n is holding it's own based on this.Quote:
Originally Posted by informal
1.6V for 4Ghz ...w00t?
like when they says 15% slower...... they for get to mention it's got 50%-75% less cache.Quote:
Originally Posted by charged3800z24
Fist it's not known if CPUz is even reading it correctly,for example take a look at fist CPUz shots from official event where AMD showed the world the C0 stepping back in the day(Cebit 2008):Quote:
Originally Posted by Yukon Trooper
http://pics.computerbase.de/2/0/7/3/8/4.jpg
http://www.roclockers.net/forum/inde...ach=8425;image
Note the vcore that CPUz reads... 1.65V for 2.3Ghz,yeah right :p: .We have seen this exact stepping of the chip in hardspell preview in August,the vcore was 1.25V.So i say CPUz is "spreading FUD" :D
It is not just the cache size but also the type. Deneb will have larger L3 cache but it is intel's large L2 cache that is the big benefit. It would be likely with smaller L2 caches I7 may not preform as well as penryn chips.Quote:
Originally Posted by Glow9
Thanks for clearing that up to those of you not being a jerk, I still think that other chips with larger caches there is a clear benefit. If there wasn't I doubt AMD would have chose to increase Denebs as much as they did. See I agree with Yukon here and looking to the pictures of the heatsink and fan that thing must be running hoooooooooootQuote:
Originally Posted by Yukon Trooper
You agree with what?All he said is "1.6V for 4Ghz ...w00t? " .Quote:
Originally Posted by Glow9
You are not making any sense,apart from making yourself look uneducated with those clueless cache/design comments you made.Now you make another clueless comment without even seeing in what contest the latest 4Ghz shot was made(a preview of maximum clock is not going to be done with stock cooling,would it? great find Columbus :clap: )
Deneb is running much cooler and draws much less power than 65nm Agena.We know this as a fact from August(the earliest C0 stepping,C2 will be better):
http://en.hardspell.com/doc/showcont...58&pageid=3150
Idle:
http://img.inpai.com.cn/2008/8/6/24d...c80f883a83.gif
Deneb draws 21% less power than same clock/same volts Agena!
So idle testing shows you are wring,yet again!Sigh...
Load:
http://img.inpai.com.cn/2008/8/6/2b7...287f1301dc.gif
104/57=1.81 ,so Deneb draws 81% less power than same clock/same volts Agena! That's almost 2x less!!!
Now your claim about Deneb running hot is not only busted,it's shattered...Sigh 2nd time.. :rolleyes:
Quote:
Originally Posted by charged3800z24
Quote:
Originally Posted by informal
http://www.techpowerup.com/img/08-10-11/34a.jpg (techpower)
we see Deneb 4.0ghz at 1.6V's air ? O_o
Deneb is looking better all the time now lol
Glow9 no matter how hard i tried to show you some facts and correct your ignorant views on this topic,you continue to act as a troll.I haven't insulted you with one word while at the same time i posted useful information just to help you stop looking stupid in this thread,but you totally ignored my posts and continued your clueless ramblings.From totally clueless comments on CPU design,to intel troll comment about Deneb running hot after which i responded with hard proof which you dismissed as fanboy post?!This makes you a troll in AMD section who's intention is not to participate and learn here but to poo poo on AMD whenever you see a slightest chance for it.
*Ignore list*
ok fine i admit that amd cpus are poorly designed and intel's cpus are way better designed, for this program. so you happy? that intel can score better scores on a program that is completely meaningless? has nothing to do with real life situations, and the program is only a test. a test that was good in the past but is pointless to use ever since the first dual core cpu came out. and if deneb comes out and gets really high scores on cpuz i will be saying wow thats one hell of a cpu doing so good at a huge disadvantage. i doubt anyone is gonna suddenly say its a good program. that doesn't even make any sense. wouldn't you rather want to win at a disadvantage than to win and say it was even? :shrug:Quote:
Originally Posted by Glow9
I'd like to see AMD win the disadvantaged test and the equal one haha. But even if AMD runs a disadvantaged benchmark we can still compare it to a phenom score in the same program. This is up to whoever has one but the people running Denebs right now I haven't seen that many benchmarks yet. 1.6 is still kinda high for new chips to run 4Ghz would like to see a little lower.Quote:
Originally Posted by roofsniper
I think you're all crazy. :p:
In honor of our bickering over benchmarks, I wrote one called "Generic CPU Benchmark" just for us. Let it be another contender in the "ZOMG is it fair?!" debate. :)
This requires the .NET 2.0 Framework.
For reference, I get about 16750-17000 KCU/s on my 9850 @ 2.5GHz (stock).
I agree the lack of new AMD chips has really done a number on um "certain people". Now they're all defensive I'm pretty sure if AMD released a 4Ghz chip that ran at 3v and was 120C they'd still be saying how awesome it is and it's as good as what Intel is offering. Yeash... I hope AMDs new DDR3 Denebs kick butt to bring a few more sane members over to this end.Quote:
Originally Posted by Yukon Trooper
phenom 9600 stock clocks.
ZOMG Particle i hate you, thanks to the program my pc now has AIDS.
and my score is 11,835 KCU/s on an X2 @ 200 x 14 = 2800mhz
what exactly does that little nifty program test?
im guessing you can only compare scores to cpus with the same number of threads?Quote:
Originally Posted by Manicdan
The program tests a battery of commonly used math operations on a large dataset. It should be a good mix of CPU and RAM stress.
Yes, numbers will increase dramatically with extra cores. It is multithreaded.
ran it a second time and got 21,519. pretty nice as it locks down my comp when its running. doesn't run very long tho.
It's fairly consistent between a dataset of 16MB or 512MB. I can make it longer if you want though.
might work as a stress tester if it can go into a loop.Quote:
Originally Posted by Particle
Screen shot of all cores at 100% woooo? Or how does this proggie work?Quote:
Originally Posted by roofsniper
Glow, did you for the ease just ignore below quoted picture?
We cant be sure how AMD's 45nm process reacts to 1.6Vcore, although Intel's 45nm process from Core 2 line is not happy with 1.6Vcore, yet we see here a 200Mhz higher OC than Deneb using 'only' 0.12Vcore more. So what was your point exactly again:clap: Especially if you think about Bloomfield is already at high/k and metal gates:rolleyes:Quote:
Originally Posted by demonkevy666
He ignored every counter argument(with proofs) against his clueless claims...No wonder why he ignored that one.So i ignore him now.
As for that Core i7 picture,with 1.7V that chip is gonna "celebrate" New Year outside the motherboard socket,i can tell you that much.
It's hard to say still what we'll see on voltages. That i7 is an ES just like the Denebs that we're familiar with.
http://www.tomshardware.co.uk/forum/...eneb-0ghz-time
Oh, BTW, the tester seems to be a contract board tester, and he has to add CPU codes into the BIOS to do the test. So don't bother check voltage. It's his job to figure out the calibration curve.
theres not much of a point in showing nehalem running on 1.72V. intel themselves have already stated anything above 1.6 will damage the chip and even running it at 1.6V for everyday use is probably still gonna hurt the chip. i bet if you hit the volts down to 1.6 you won't even be seeing nehalem above 4ghz.
my phenom 9950be atr 3ghz has only 21551KCU/s??:(
something must be wrong. is this bandwidth sensitive?
what ram are you using? the test appears to be just more than cpu related as i believe particle had a lower score than me on a better cpu.Quote:
Originally Posted by LightSpeed
Hi i got 22379 running 3.107 ghz doing some ACC testing
Yes, because of the large dataset, the program is a mix of not just CPU performance but also RAM performance. It's a more realistic scenario imo since in the real world you're pulling and manipulating data from somewhere--not just playing in the CPU's cache.
what speed ram you guys got? cause i didn't think my corsair dominator 1066 at 5-5-5-24-30 was that great. i don't even think those are stock clocks i believe its underclocked. anyway particle create a new thread for this. might be cool to see what scores people can get since most cpu testing programs out there are affected mostly by cpu clock.
hell it would be nice to see the change in scores on this simple bench going from 2mb L3 to 6mb on the phenom.
the same way "certain" peole were when Intel was reaching the 4ghz mark and was doing it with space heaters, and even at that speed they could not beat a 2.6ghz AMD?"....AMD's Phenom is a good chip and priced right. If Intel didn't have what they have now, I'd be more suprised, since they have like every resource and funding at thier finger tips. AMD has the know how..just not the resourses..Now if AMD was charging 1200 for the fastest Phenom.. I would not own one. So "who-ray" for Intels 12mb cache 1500 dollar 3.2ghz chip that PWNs all......Intel still charged 1000 for there newest chip back before C2D and it couldn't compete with AMD...And those voltages could be wrong.. I have my own screeny of my 9850 OC on my 790gx board..it read 1.776v and it was not correct. I updated CPU-Z and it has read correctly ever since. and, if this board runs more than 1.52 it gets scarey.Quote:
Originally Posted by Glow9
http://3800z24.info/Phenom/3.3ghz.jpg
you should try change the last two timing to 15 and 42.Quote:
Originally Posted by roofsniper
or 15 and 26 if you sticks can run them.
I made a thread for this in the benchmarking software forum.
in bios i believe i set it to 15 and 22 which i believe is the epp spd for these sticks. bios overides that tho and gave me 24-30.Quote:
Originally Posted by demonkevy666
also i found this awhile back for my mobo: TRAS - bios 1002 and up overrides this setting: If tRTP is set to Auto then -2 from what the setting is. Any other tRTP setting and this is 18 no matter what the setting is.
TRC - BIOS 1002 overrides this setting if tRTP is not on Auto. If tRTP is not on Auto this is 26.
i haven't really tried messing with them much i just wanted stock for now mostly because a phenom 9600 is probably amd's worst quad core to overclock besides the 9500. deneb is a different story tho.
I just ran that Bench and got 30,844 @ 3.3ghz /2400mhz NB.
http://3800z24.info/Phenom/9950/32bit/Generic_bench.png
I dunno about you but I'm pretty certain I spoke about a Q9550 not a i7 so whats the point of this? If deneb is only around 13-15% faster per clock than current chips that makes it on par to C2D I dont even thing comparing it to an i7 is doing anything is worth while. I doubt we'll see anything insane until AM3Quote:
Originally Posted by Rammsteiner
Hehe...... :bounce:
Just had to break 31,000 KCU's
http://i30.photobucket.com/albums/c3...1-225722-B.png
I posted in the Bench Thread Particle started too.
It has a full SS with other specs on the run if your interested...
hmmm i just got to thinking seeing that cpuz shot of the bloomfield, that Intels new Core i7s will flop, kinda like the willamette core did. Intel is foraying into the integrated memory controller area now and their quick path data connect, all on an architecture that is heavily dependent on large amounts of L2 cache to perform well. We all know that L3 cache is much slower than L2 and it is a shared cache. L3 helps alot with multi threaded apps that use multiple cores, but with intel dropping their L2 cache on their cores down to 256k, i see that as a big performance hit on regular single threaded apps which is what most programs are. It is a smart time for intel to test the IMC, cause they are ahead, but i think they underestimated AMD again. I think by the time all is said and done, the new 45nm Phenoms will be better than the 45nm Core i7, and on par with the current yorkfield chips.
I see the bigger L3 for the Phenom to be nice to definitely help with multi-threaded apps as well as keep more data in the cache vs having to go to RAM. I still dont see why intel went with such a large L3 and such small L2's. I coudl be wrong, but the dothan core did well with larger L2's, the core architecture was based off of the dothan and loves the uber large L2's. Drop the cache down and it is plain to see that their chips loose quite a bit of performance. I dont see the quickpath doing anything for intel's current arch, maybe when intel does a new arch with the IMC in mind this time they will make a much better performing chip. You still have to remember that intel's quads are still nothing more than 2 dual cores put on the same die, where as Phenom is a naitive quad. We shall see soon if AMD's native quad design pays off, and i have a hunch it will.
Well, considering that some Intel fangirls actually think the 52% gaming increase is true, I don't think the Bloomfield chip itself is at fault.
As with GT200, it's going to be an average upgrade experience.
Nothing even close to revolutionary for average use (for video encoding, graphics cards are the future. Powerdirector with GPGPU encoding is coming soon.).
I mean, K8 vs P4. P4 was better in multimedia benchmarks, but who really gives a ****? Most of us just put our encoding in a queue and do it at night. Even if we do it real-time we'd still do other stuff. It's tangible.
Less tangible stuff? Let's see.... gaming. K10.5 doesn't have the per clock advantage anymore, but it should give a better gaming experience compared to a same-priced Bloomfield.
^
The exact thing K8 did. (Oh, though that was at lower clocks)
daseto: I think you may very well be right, I don't pay a whole lot of attention to what Intel is doing, seeing as how I'm an unapologetic AMD fan... ;)
From what I've read though it seems like Nehalem was designed to compete more in the server market, and the improvements may not translate to the desktop/gamer segment...
Only time will tell, adding an IMC has proven to be a huge benefit to performance but it's a pretty big step and AMD definetly has the jump in that department.
I have pretty high hopes for Deneb, but I really don't expect early releases to hit 4Ghz right off the bat... Early Agena cores were not that impressive,
but now they can actually hold their own against C2Q. They may not win hands down but the price/performance makes them a serious contender....
BTW: I've watched all this sillyness concerning SuperPi.... A dude on TechReport (2-3yrs ago) SMOKED my X2/4800 using a P4. It was the only benchmark he could win.
I won't say it's totally irrelevant but, it's pretty much irrelevant when comparing different uArchs... :p:
Since I cant findit, I'll post my experience here:Quote:
Originally Posted by Particle
X3@3.0 GHz (2.0 GHz NB) DDR2-1000 Ganged [Vista Ultimate x64]
http://img406.imageshack.us/img406/1...0081012gi3.png
Another run with the same settings:
http://img235.imageshack.us/img235/2...0081012lg5.png
this looks to me like a significant marine? And score does looks like to low??!
:shrug:
Now take look this score with these setup:
X3@3.3 GHz (1.8 GHz NB) DDR2-1066 Ganged [Vista Ultimate x32]
http://img399.imageshack.us/img399/8...0081012yp6.png
:confused:
to what to relate this big difference?
i think 64bit vista gives lower scores? my phenom at 2.3nb/3.1cpu/860mem 4-4-4-12 gives 23000KCU/s?
Sorry, I just had to one up you ;)Quote:
Originally Posted by Daveburt714
http://3800z24.info/Phenom/9950/32bi..._Bench_3.4.png
Guys the benchmark maked it way over to the tech powerup forum in the overclock & cooling section you guys should go post your score show them what Phenom can do .
I just checked out the tech powerup forum. That's funny :D Release a CPU/RAM benchmark for fun just for a few guys to play with and leaderboards start cropping up on forums I've never been to. This is neat.
Grats Particle nice job on the benchmark You Are The MAN !! next thing you know your benchmark will be all over the internet forums.
way lower scores under Vista x64 for me as well :p:Quote:
Originally Posted by LightSpeed
I'm averaging ~19,800KB/s at 2.65GHz/2.33GHz 1130RAM 5-5-5
On Win XP x32 I'm getting ~24,700KB/s!
I will try on my win XP 64 and see how it is...Quote:
Originally Posted by Lightman
Edit: Win XP 64bit...I dropped 5000pts when @ 3.1ghz, going from Win XP 32.
http://3800z24.info/Phenom/9950/winx...n64_3.1ghz.png
he made the post in the "benchmarking software" section. lets use that instead of here.
http://www.xtremesystems.org/forums/...d.php?t=204361 link to the page
Anyone been keeping up on info popping up in the coolaler forum topic of this.
I don't know what language it's in so I can't translate it.
If you want to ask him something then see this thread -----> http://www.xtremesystems.org/Forums/...d.php?t=204467Quote:
Originally Posted by Caveman787
I heard only his superpi on 4.1GHz, on 4.2GHz run to windows, but not corectly superpi...
Chinese.......maybe?:DQuote:
Originally Posted by Caveman787