• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Nehalem SuperPi score

I'm calling BS... At 2.66ghz... no f'n way... I mean the E8500 at 5ghz is just reaching those superpi speeds... I don't see how this could be.
 
The link above uses this page as a source: http://infomars.fr/forum/index.php?showtopic=1524

and the image is from PkG.1337 @ XS is on xtremesystems? Why didn't they just link to the XS post? :p

I'd like to call shens, but it's not like the source is a complete fraud. I can't find the post if it was posted on XS so I'll stay skeptical.
 
I highly doubt that's real. That means a 100% increase in floating point operations. Intel's pretty damn good, but not that good.
 
10-20% would be a nice improvement, nearly 200% is not realistic!
 
Sunin said:
10-20% would be a nice improvement, nearly 200% is not realistic!
Click to expand...

Its not 200% dude. And, check the picture. Is it me or does the Revision look Photoshopped? Also, 0.08v? Um... would CPU-Z really not read that correctly? I suppose if it is a new chipset and motherboard that has not been released, but seriously... And isn't Nehalem supposed to be 1600MHz FSB?
 
mwarps said:
I highly doubt that's real. That means a 100% increase in floating point operations. Intel's pretty damn good, but not that good.
Click to expand...
Intel has said that the jump from Core2 Duo -> Nehalem is significantly greater than the jump from P4 -> Core2 Duo.

I suspect that benchmark is real, but it is also misleading. Remember, one of the main [supposed] benefits of Nehalem is that it can distribute single-threaded tasks over more than one core using dynamic multithreading. Current Intel CPUs can't do that. Current Intel CPUs can only use one core for that test.

If you ran four instances of SuperPi instead of just one, I bet the net gain would be <20% over a Core2 Quad at the same clock speed.
 
Well it is nearly 200% if at 2.66ghz it can do the equiv Floating point calcs that a 5.2ghz machine can do. I'm no math major, but I'm fairly certain that is near double the performance per clock!

But maybe the problem is simply that it is not ready anythign about the processor correctly. That could be likely and thus maybe it is truly 5ghz vs 5ghz and its just showing that it can do it... who knows... it definitely peaks your interest about the processor though!

Ahh now that is interesting, about the single thread being split by the processor, that certainly could account for the performance gains, which actually the more you think about it is in line with what one might expect. You would not see 400% increase in performance from 4 cores, but somewhere around double the performance due to the mechanics... kinda like SLI doesn't double performance...

Hmm /drool if this puppy truly performs like this I will be more than happy to consider it for my build in 6mos.
 
I think I will need a sub-prime loan to afford one of these things.

 
KenF said:
I suspect that benchmark is real, but it is also misleading. Remember, one of the main benefits of Nehalem is that it can distribute single-threaded tasks over more than one core using dynamic multithreading. Current Intel CPUs can't do that. Current Intel CPUs can only use one core for that test.
Click to expand...

Thats what I was thinking.

SuperPi currently only uses 1 core, if Nehalem can break up single threaded tasks like its supposed to. (like reverse hyperthreading) That would make perfect sense to be 200% faster in a normally single threaded app. And since most apps are single threaded & don't take advantage of the 2nd, or 3rd & 4th cores of Core 2, this would be a HUGE improvement.
 
Software and Game Developers all over the world let out a sigh of relief!

LOL...
 
chrisf6969 said:
Thats what I was thinking.

SuperPi currently only uses 1 core, if Nehalem can break up single threaded tasks like its supposed to. (like reverse hyperthreading) That would make perfect sense to be 200% faster in a normally single threaded app. And since most apps are single threaded & don't take advantage of the 2nd, or 3rd & 4th cores of Core 2, this would be a HUGE improvement.
Click to expand...

Damn, I missed that part about the Nehalem architecture... That is HUGE.
 
chrisf6969 said:
Thats what I was thinking.

SuperPi currently only uses 1 core, if Nehalem can break up single threaded tasks like its supposed to. (like reverse hyperthreading) That would make perfect sense to be 200% faster in a normally single threaded app. And since most apps are single threaded & don't take advantage of the 2nd, or 3rd & 4th cores of Core 2, this would be a HUGE improvement.
Click to expand...

Thank you.
 
Makes me feel much better about using DFI mobo with no penryn support now ;)
After all it won't be needed in one year time ;)
 
Hmm I wonder what a 5 Ghz OC'd Nehalem processor could do... sub 4s SPI? I remember when my 3.2EE did 32s in SPI... hahahaha that is insanely fast at 4s
 
Sunin said:
I'm calling BS... At 2.66ghz... no f'n way... I mean the E8500 at 5ghz is just reaching those superpi speeds... I don't see how this could be.
Click to expand...

The screenshot is probably taken at idle mode. The SuperPI was probably run at x10 multiplier and not x6 in the screenshot, which would make it ~4.4ghz. Screenshot might be fake though
 
surazal said:
The screenshot is probably taken at idle mode. The SuperPI was probably run at x10 multiplier and not x6 in the screenshot, which would make it ~4.4ghz. Screenshot might be fake though
Click to expand...

I'll believe faster that in a particular test Nehalem is 2x as fast , than to believe it's 20% faster runs at 40% higher frequency.C'mon , 4.4GHz on early silicon?? :rolleyes:

IMO , the part is an ES with OC FSB ( or QPI , or whatever ).The original 2.66 ( 266x10 ) was changed to 444x6 , same frequency .
 
If the pic is fake that's a shame, otherwise holy crap this chip is going to be amazing. Also, very interesting about Nehalem being able to distribute the work itself, that'll keep programming much easier AND give a performance boost, good thinking intel.

Also, fwiw, 2x the work is 100% faster, not 200 :p
 
If this dynamic multi-threading is true than hot damn, Nehalem is going to be a freaking beast! Im curious how many cores/HT's can it use to one a single thread? Because I mean it has 4 cores and 4 HT links. I mean just thinking about what this means for the octocore gets me pretty happy in the pants. Cant wait until more info on Nehalem gets released as we near its launch date.
 
DemonDiablo said:
If this dynamic multi-threading is true than hot damn, Nehalem is going to be a freaking beast! Im curious how many cores/HT's can it use to one a single thread? Because I mean it has 4 cores and 4 HT links. I mean just thinking about what this means for the octocore gets me pretty happy in the pants. Cant wait until more info on Nehalem gets released as we near its launch date.
Click to expand...
I suspect four cores will be the maximum for a single thread. On Nehalem, each of the four cores has direct access to the L1 cache on the other three cores. But you don't have that access across separate dies.

You are not going to get a 400% performance increase when using dynamic multithreading to distribute one thread across four cores, due to overhead. But a 100% improvement on single-threaded applications and games -- as suggested by the SuperPi score -- would seem feasible.

I suspect Nehalem will 'obsolete' everything available today for applications and games that use one primary thread. The dual-die, 8-core version should do the same thing for applications written to use two primary threads. Existing Core2 Quad CPUs should remain competitive with the [4-core] Nehalem for any applications that are written to make full use of four cores.

The 'downside' of Nehalem is that developers will have less incentive to update their applications to make full use of more than two cores. That could hurt owners of existing quad-core CPUs from Intel and AMD.
 
My P4 1.4 GHz system just did 1M in 2min 34 sec ;)

Don't know if the above posted Nehalem score is legit (I hope it is), but in any case, we've come a long way...
 
KenF said:
The 'downside' of Nehalem is that developers will have less incentive to update their applications to make full use of more than two cores. That could hurt owners of existing quad-core CPUs from Intel and AMD.
Click to expand...

Yea I was afraid of that possibility myself. Software development is already lagging behind, with the advent of dynamic multithreading it will more than likely lag along even more.
 
I must ask something : where did this BS about reverse HT ( take 2 ) come from ? :rolleyes:

There's no such thing.
 
savantu said:
I must ask something : where did this BS about reverse HT ( take 2 ) come from ? :rolleyes:

There's no such thing.
Click to expand...
There are a half-dozen or so documents about it scattered throughout Intel's site. Here is one.

They call it dynamic multithreading, not reverse multithreading. The processor creates multiple threads from a single thread.
 
KenF said:
There are a half-dozen or so documents about it scattered throughout Intel's site. Here is one.

They call it dynamic multithreading, not reverse multithreading. The processor creates multiple threads from a single thread.
Click to expand...


Beat me to the punch on the links. :eek:
 
KenF said:
There are a half-dozen or so documents about it scattered throughout Intel's site. Here is one.

They call it dynamic multithreading, not reverse multithreading. The processor creates multiple threads from a single thread.
Click to expand...

And what do research papers from 1998 have in common with Nehalem ?

Where does it say that Nehalem implements such a technology ?
 
savantu said:
And what do research papers from 1998 have in common with Nehalem ?

Where does it say that Nehalem implements such a technology ?
Click to expand...
That was not the last presentation, just one I found that had a full PDF. Other presentations were made as recently as 2002-2003.

Intel has also said in the past that a future implementation of SMT would be used to improve single-threaded performance. Based on Intel's statements, I think it is logical to assume that Nehalem is that implementation. Intel has told us that Nehalem can do things which you would need for dynamic SMT -- such as each core having direct access to the other cores' L1 caches.

If you are looking for proof, that doesn't exist because Intel has not announced all details on Nehalem yet. However, they did say over and over at IDF last fall that Nehalem will significantly improve single-threaded performance. They said many times that the improvement from Core2 Duo -> Nehalem is larger than the improvement from P4 -> Core2 Duo.

We should learn more at the Spring IDF.
 
You must log in or register to reply here.
Back
Top