Tom's Hardware Forums » Motherboards & Memory » Memory » Intel's Core 2 Quadro Kentsfield: Four Cores on a Rampage
 

Intel's Core 2 Quadro Kentsfield: Four Cores on a Rampage




Word :   Username :  
 
 Page : 1 2
Next
Author
 Thread : Intel's Core 2 Quadro Kentsfield: Four Cores on a Rampage
 
bga
Profile: enthusiast
More Information

Last message on previous page:

Quote :

Windows might recognize 4 cores, but does it optimize it's work load across the 4?



Yes, Windows is in fact a quite good multi-tasker, multi-processor operating system. I have tried up to 4 cores, and if you have a lot of treads, Windows distributes them very well, with no problem between the different applications. Sometimes a multi-threaded application can have internal problems, due to race, and dead-lock problems - hardly the fault of Windows. If you want a single application to take advantage of extra cores, the application typically needs to be rewritten. It is normally not a trivial task to divide the work between several treads, unless it is for something relatively simpel as a background printing process. There are very few real multi treaded programs on the market today.
A can remember some tests which documents that Windows scales well up to 8 cores, but then gets decreasing returns for every added core.

Related Product

Register or log in to remove.

cxl
Profile: enthusiast
More Information

Quote :

ic... do they by any chance have a program that I can use to will allow me to assign certain tasks to one of the two processors? Or is it faster for a program to spread across two, like converting from .avi to DVD format...



AFAIK, this is called "thread affinity" and is done by OS automatically.

It is in fact quite important to avoid cache trashing.

Profile: nimble knuckle
More Information

Quote :

ic... do they by any chance have a program that I can use to will allow me to assign certain tasks to one of the two processors? Or is it faster for a program to spread across two, like converting from .avi to DVD format...

I was thinking about being able to conver a movie on one core while being able to play a game on the other core...



Windows is actually really good about managing this for you. If you have a core that is idle and a core that is busy, it will automatically assign all new threads to the core that is idle. It avoids moving a thread to another core if possible, but it will do so if there's another core that is less allocated.

You can use System Monitor in the control panel (and Microsoft Management Console) to monitor number of threads, etc.

Be aware that altough you might have plenty of processor resources, your computer may still be constrained by other bottlenecks, such as the hard drive, memory bandwidth, etc.

Profile: old hand
More Information

wow, with my last 2 posts, i feel like i've stirred up the hornet's nest 8O

thanks for the info though =)

Now, to go home and try the new BIOS and see if i can unlock the multiplier and lower it while raising FSB :D

Perhaps I should get an aftermarket cooler. However, I'm somewhat worried about the weight of some of the coolers, such as the Sunbeam Tuniq Tower 120 or the Scythe Ninja. I read somewhere that the intel stock cooler is only like 480 grams where as these two are almost twice that much! But i take it that's why aftermarket coolers come with their own mounting brackets.

Another thing, what are the odds that my 975X mobo will support Kentsfield processors with a simple BIOS update? I mean, I've been able to overclock and hit a 1400 FSB with the CPU... however, I could not get Prime to run on it :(

Profile: stranger
More Information

I think its obvious who is paying THG's bills. Check this out

Quote :

While the Core 2 Duo registers a maximum power consumption of 65 W with a supply voltage of 1.3 V, it follows that two such chips should consume up to 130 W of power. The test results show that the test platform consumes 167 W when idle (outfitted with an ATI Radeon X300 and a simple 7,200 RPM hard disk). At full load, it eats up 260 W worth of energy. That's still an increase of 100 W! By comparison, the Core 2 Extreme X6800 system consumes 142 W in idle mode and requires 165 W at full load.

OK, Quadro uses 95W more under load than a C2D Extreme. 65W + 95W > 130W. That much power draw is far more than any other Intel chip, right? Well lets see how THG accounts for this...

Quote :

Running at full load, the Core 2 Quadro is at the same level with the old Pentium EE 965. After all, this Pentium version uses 15 W less than the dual-core Pentium EE 840 - and that with clearly enhanced performance thanks to four processors. This is significantly below the TDP limit (Thermal Design Power) of 130 W.

So an EE 965 uses 15W less than an EE 840 (130W). That makes 115W. To summarize THG's math:

Quadro = 115W (EE 965) < 130W (EE 840)

When in fact they've already said that Quadro uses 95W more than a C2D Extreme (65W). So tell me, THG, how does 65W + 95W = 115W?

Couple this with videos 1 and 11 and, gee, I really have to wonder why THG is writing such things...

Profile: journeyman
More Information

How much of a price impact do you all think the introduction of the C2Q will have on the C2D market? Based on past experience, will Intel lower their prices at all, and if so, how quickly?

Profile: old hand
More Information

go back through TH articles on CPUs and I'm sure it'll follow a similar trend to the pricing of the Pentium D's.

Profile: enthusiast
More Information

I'm probably the exception to the rule on this forum as I'm not a hard core gamer, but I play some games.

I couldn't care less about how much power the CPU uses either, as my 805D @ 4.1GHz doesn't bother my 600w PSU and isn't noticeable on my electric bill.

The quad core will allow me to leave my AV running, decrypt a DVD or rip a music CD, and play either Madden 07 or Civilization IV, and all at the same time. This is something my current 805D @ 4.1GHz struggles with. Its fast as.. for gaming alone, or any single program for that matter, but that doesn't solve my everyday multi-tasking problems. The AM2 systems that I've built to date with FX-62 didn't perform any better with multi-tasking. I'll pass on C2D for these reasons and hold out for a quad core.

So for those of you who say "who does all that at the same time?", there's a few of us out there. My wife's multi-tasking makes me look tame. :-)

Another great reason for quad core = Vista. I just intalled a Gigabyte GA-965P-DQ6 to be ready for the quad core when its available (and partially because all Madden NFL games crash on the Asus P5WD2-E Premium with the 975 chipset). I'm running a Leadtek PX7900GTX-TDH Extreme with 2 x 1 gig of Crucial Ballistix 6400 @ 4-3-3-8(3) @ 2.1 volts. Not the best system in the world, but nothing to sneeze at either. Well I run Vista RC1 on a separate SATAII hard drive for testing, and it slows my system to a crawl. (Maybe a little exaggerated, but I can't stand lag.) Of course its the version with all the bells and whistles, but it may well be the most popular version of Vista with its networking and multi-media support. Anyone running this version of Vista will definitely appreciate the advantages of a quad core CPU. I can't speak for other versions of Vista, but my gut feeling is that its an overloaded version of XP for even the basic version.

If all you do is game, then its probably not worth it to upgrade to a quad core until the software catches up, which may be never. But if you multi-task or expect to upgrade to Vista, quad core will definitely have big advantages. Multi-core CPUs are (in the words of Leonardo DiCaprio as Howard Hughes) "the way of the future.. it's the way of the future.. it's the way of the future... :-)

Profile: enthusiast
More Information

With Windows being a little, lets say incompetent to properly handle the 2xDual Core.

Wouldnt it be nice if we saw Toms put out benchmarks on a natively compiled toolchain and kernel on either machine (Core 2 Quad, and X2).

Using an SMP Linux kernel, one could achieve maximum output from these CPUs, As Unix based load balancing has out performed all of Microsoft products in the past. Why not make use of it for benchmarking too?

Espeacially with the VIDEO and AUDIO encoding/decoding.The multitasking could also be easily benchmarked by starting a compilation task with more simultaneous jobs
eg (make -j 4).
Using tools such as "time" can provide dead accurate results.

Such testing would provide bleeding edge performance. (Espeacially with native kernel).
However optimised, LAME or mpeg encoding application may be, if the uderlying OS does not balance load properly, then you will not get real world results.
God knows what Microsoft is emulating with their 64bit Windows, just like how Apple used rosetta for yrs.

http://widefox.pbwiki.com/Kernel%2 [...] %20Windows

If any benchmarking should be done on Windows at all, it should only be for Computer Games.Microsoft platform is not designed for performance. So why does Tomshardware use it for benchmarking???

bga
Profile: enthusiast
More Information

Quote :

With Windows being a little, lets say incompetent to properly handle the 2xDual Core.


Thats is an erroneous assertation. Windows (NT kernel) is very good at multi core/multi CPU systems, at least up to 8 CPU's. It is correct that extremely huge CPU clusters run Linux, Unix or special operating systems, but at least for the Linux supercomputers, they are using kernels which have been modified heavily.
In every benchmark I have seem, Windows is a better multi-processor than Linux, which is again better than MacOS (which is based on a Unix kernel).

Quote :

Using an SMP Linux kernel, one could achieve maximum output from these CPUs, As Unix based load balancing has out performed all of Microsoft products in the past.


Wrong again. Back up you fan-boyism with hard facts or shut up.

Profile: enthusiast
More Information

I left a link for u in my post, if you cant read english then I cant help you.

I personally have administered both Microsoft and Linux based servers in the past. And I can tell you... 1 cpu, 2cpu no difference, in fact IIS was faster than Apache on a Dual Xeon, so we ran it on Windows 2003.

However once we go into Quad Opteron, or Blade servers, the database through put from the 64bit machine running Linux was over 10x more than that of Microsoft SQL on the same machine. The company I worked for had over 3000 users on the database.

And by the way....


Quote :


In every benchmark I have seem, Windows is a better multi-processor than Linux, which is again better than MacOS (which is based on a Unix kernel).


Wrong again. Back up you fan-boyism with hard facts or shut up.
:) Two can play at that game. :)

Profile: enthusiast
More Information

Sharpen up on your reading skills then try this..
http://www.tpc.org/tpcc/results/tpcc_perf_results.asp

as the size of the db increases, Microsoft is nowhere to be seen

http://www.tpc.org/default.asp

http://www.tpc.org/tpch/results/tpch_perf_results.asp

Remember Benchmarks are not only for gamers. Working professional computer engineers also want to know whats fast and whats not.

My suggestions is just to test the HARDWARE. not the software.
If Linux yields better results than windows on the tests i mentioned earlier, then only show those results. Don't need to even mention the OS used. Please, i am not trying to see who can shout louder.

bga
Profile: enthusiast
More Information

Quote :

I personally have administered both Microsoft and Linux based servers in the past.


And I have 20 years of heavy duty IT experience, and working with Windows Server all the way back to OS/2 LAN Manager 1.0. So I have to pull rank on you there.

Quote :

And I can tell you... 1 cpu, 2cpu no difference,


Then you must have configured your server wrong. Everybody competent gets much better webserver performance with 2 CPU's.

Quote :

in fact IIS was faster than Apache on a Dual Xeon, so we ran it on Windows 2003.


Well, that just proves my point! I'm not a Microsoft fanboy, so I don't claim that IIS is better than Apache in every setting. In high performance setting with PHP, Perl or any scripted backend Apache is much better than IIS. But that has NOTHING to do, with Windows being a bad multi-tasker/multi-processor.

Quote :

However once we go into Quad Opteron, or Blade servers, the database through put from the 64bit machine running Linux was over 10x more than that of Microsoft SQL on the same machine. The company I worked for had over 3000 users on the database.


If you got a 10x difference, then you should hire a Windows professional. Again, companies much bigger than 3000 users run SQL server with better performance than MySQL or whatever you ran on your system. Again, applications such as databases, and enduser applications using those databases have got much more influence on the total system performance than the operatingsystem. By the way, did you use 64bit windows and 64 bit SQL server on your setup? If not, you are comparing a RAM limited 32bit Windows to a 64 bit Linux machine which can have the whole database running in RAM. Not a very good benchmark!

Quote :

Sharpen up on your reading skills then try this..
http://www.tpc.org/tpcc/results/tpcc_perf_results.asp
I left a link for u in my post, if you cant read english then I cant help you.


Well obviously you can't read! :roll: In the link you sent me as "documentation", Windows wins over Linux! Only Aix on a very high end IBM machine is faster than Windows on a HP Windows system.

Here for some links which documents Windows very good multiprocessing capabilities:

See here for 3DSMax scaling.

http://www.anandtech.com/IT/showdoc.aspx?i=2793
(Woodcrest test on Windows 2003 64 bit - Performance scales nicely with CPU power).

http://h71028.www7.hp.com/ERC/down [...] 208-way%22
(Windows scales nicely on BIG hardware.

IBM says Windows scales well:
http://www-03.ibm.com/servers/eser [...] eries.html

Now little Linux fanboy - go play somewhere else....

Profile: enthusiast
More Information

Hahaha, your so not getting the point here.

when i said "1 cpu, 2cpu no difference," I meant the performance difference between a SMP NT and Linux kernel . Obviouslt the dual was faster than the single, but there was no difference between the performace of Windows and Linux with dual cpu.
The performance starts to show once the number of cpus are increased.

Quote :

And I have 20 years of heavy duty IT experience, and working with Windows Server all the way back to OS/2 LAN Manager 1.0. So I have to pull rank on you there.


Haha, soo childish. :P :P :wink:
What made you think I dont have over 15yrs of High Level design and management experience of Enterprise systems, I have worked for the "Education Council" in Botswana (South Africa) over 35 000 users. I have worked in Australia as a Design Engineer for IPX management. And now I am in Canada working as a systems engineer for large scae Wireless networks.
Next time think twice before pulling rank.

I went through all the links u posted, none of which are doing you any good, as they have NO COMPARISONS on SMP on LINUX vs WINDOWS.

Windows scales well, I AGREE with you.
What Im saying is linux scales BETTER.

the AIX was a clear example. Dont forget what Unix was designed for "Time Sharing"

Again I will post the same link, look at it http://www.tpc.org/tpch/results/tpch_perf_results.asp
at the 100GB WINDOWS is BOSS.
but anything more than that, !!Microsoft NOWHERE to be SEEN!!
at the 300GB level, Red Hat Enterprise Linux AS 3 Tops out.With *nixs taking the lead then on.

But anyways we all know that Unix/BSD and Linux are far more superior to any Microsoft product, otherwise why would your Bank use it? or your local hospital? or the Airport? or the American Goverment? and the British Goverment? and even the Danish goverment? just to name a few.

http://news.zdnet.co.uk/software/l [...] 606,00.htm

Now look what you made me do. You made me proove to you time and time again Linux is better than windows.
But this was not my intension. Honestly.

All I wanted was, if Tomshardware could look into using linux for some of their benchmarks (!!!NOT AS A COMPARISON between WINDOWS AND LINUX)

if for exmple the LAME encoding was faster on Linux than on windows, then show the RESULT ONLY. Not which OS was used, and hence maintain the same OS for all hardware for the LAME testing. The best performer in a test is considered the Benchmark (back to 2nd yr engineering). All other tests are done against the Benchmark.

Do you understand now?

Quote :

Now little Linux fanboy - go play somewhere else....


That was not necessary, I am beginning to doubt your "20yrs experience" now.

bga
Profile: enthusiast
More Information
n°1252790