Intel's playing catch-up

[Grimmy]

Here's another way of looking at how skewed your tests compare:

Your system:

AMD Phenom 9850 Quad-Core Processor @ 2.51GHz

Timed Linux Kernel Compilation 2.6.25 - 582.22

This system is more comparable - but yet has one test:

Intel Core 2 Quad CPU Q6600 @ 2.39GHz

Timed Linux Kernel Compilation 2.6.25 - 572.86

and the one Q6600 system you compared to yours?

Intel Core 2 Quad CPU Q6600 @ 2.40GHz - System Memory: 2012MB

Timed Linux Kernel Compilation 2.6.25 - 923.64

Tell me, in that one test... why such a big difference?

 

[jimmysmitty]

Ok. Still have no idea why you have to insult people each post.

Either way running with only 1GB will be slower than with 4GB.

But you see to think that running the systems with unequal setups is ok yet you think its not ok to run both systems using the same frequency of memory DDR2 800 because its not optimal....

Either way having different setups is not right. They should have the same memory and GPU in order to eliminate any possible bottlenecks with each CPU therefore revealing the CPUs individual performance.

 

[jimmysmitty]

Grimmy he didn't use that system you listed because it would prove him wrong. That the amount of memory DOES matter.

 

[Grimmy]

:lol:

But but but.. look at the system spec's.. the OS, kernal, memory.

My gawd..

I'm going to bed.

 

[keithlm]

Here's another way of looking at how skewed your tests compare:

So you are going to focus on probably the ONLY benchmark that might actually benefit from having more memory in that entire benchmark set... and ignore all of the other benchmark results.

My gads Intel fans are.... FANATICAL.

(Oh and btw attempting to divert attention is another good forum tactic... but I'm out of here today... time to leave. But I might check later to see what you guys attempt to come up with.)

Oh and Jimmysmitty... 1GB will NOT be slower than 4GB on almost all of those tests.

OH AND DID YOU GUYS NOTICE: I SAID IT WAS THE ONLY FULL SET OF THE "UNIVERSE" TEST TO LINK TO FOR THE Q6600 AT STOCK>>> CAN YOU UNDERSTAND ENGLISH?????

 

[Grimmy]

And you... sir, are a fracken hypocrite.

Nites gents.

 

[keithlm]

And you... sir, are a fracken hypocrite.

Nites gents.

Let's see... you guys have some benchmarks that use less than optimal hardware... I mention it and get pounded into the ground and I'm ridiculed because "it deson't matter".

Today I present data that you don't like... but you grasp at the only thing you can find to complain about... and then when I tell you that it doesn't matter... I get pounded into the ground again.

I see a little pattern here...

Yet... I'm the hypocrite.

 

[jimmysmitty]

^Calm down man. Your going to pop a blood vessle. No need to type all caps and I would hope you are typing english.

But it proved you wrong that yes more memory benefits the PC you are just too stubborn to admit it.

And yes if a real person takes his PC and leaves 1GB in it runs some tests then puts 4GB in and runs the same tests more than likely they will see a huge boots in performance AND responsiveness.

 

[shadowthor]

@Mods PLEASE Close the Thread!!!

 

[yomamafor1]

Oh... you think I care about the outcome of the REVIEW for some reason.

I don't care about the review. I care about the RESULTS. You seem to not understand that.

Review is the compilation of results.

 

[yomamafor1]

So let's see....

1. Using a slower RAM, which has been proven to minimally impact the performance of both Intel and AMD machines = wrong.

2. Using 1/2 the RAM, which may significantly impact the performance of the systems = right.

Yes you're right. I'm an Intel fanboy.

 

[Grimmy]

Let's see... you guys have some benchmarks that use less than optimal hardware... I mention it and get pounded into the ground and I'm ridiculed because "it deson't matter".

Today I present data that you don't like... but you grasp at the only thing you can find to complain about... and then when I tell you that it doesn't matter... I get pounded into the ground again.

I see a little pattern here...

Yet... I'm the hypocrite.

Oh... I can read alright. I'm quite aware that you could only find one that did all the tests, but that doesn't mean single test can't be compared to show the difference you didn't bother to check. Since I took some more time to look at the nice different results for Q6600 that better compare to your systems, even though they don't have the full set of tests.

So, what your saying, it's not fair to pick single tests, since that's all there is, compared to your full set of tests? Do you think there would be a big difference, if those guys did the full set? I think your quite lucky none of those guys did full set tests. If they did, I would imagine you would have tears falling from your face.

This one is memory bandwidth:

Intel Core 2 Quad CPU Q6600 @ 2.40GHz

Motherboard: Gigabyte Co. Ltd., Chipset: Intel 82G33/G31/P35/P31, System Memory: 3958MB, Disk Space: 20GB, Graphics: GeForce 7900 GS/PCI/SSE2 (450/660MHz), Screen Resolution: 1280x1024

OS: Ubuntu 8.04, Kernel: 2.6.24-17-generic (x86_64), X.Org Server: 1.4.0.90, OpenGL: 2.1.2 NVIDIA 169.12, Compiler: N/A

Bandwidth 0.13 / RAM Sequential Read - 4971

Your system:

AMD Phenom 9850 Quad-Core Processor @ 2.51GHz]AMD Phenom 9850 Quad-Core Processor @ 2.51GHz

OS: Ubuntu 8.04, Kernel: 2.6.24-19-generic (x86_64), X.Org Server: 1.4.0.90, OpenGL: 2.1.7769 (fglrx), Compiler: GCC 4.2.3, File-System: ext2/ext3

Bandwidth 0.13 /RAM Sequential Read - 4473

All I did was show benchmarks of your site, and you go ballistic.

Owell.. it was an interesting site. Tanks, I'll keep it bookmarked.

And yes, you are a hypocrite with sand in your eyes.

 

[outlw6669]

Wow.
This thread sure got all the Trolls working overtime...

It is late and I did not bother to read through most of the new posts but I just wanted to point out this.
C4 DDR2 800 will access data stored in RAM at the same speed as C5 DDR2 1000.
Such a minimal bump from 1000Mhz to 1066Mhz is not going to give you any noticable impact on your benchmarks...

A little reaserch on the matter:
http://www.thetechrepository.com/showthread.php?t=160
http://www.thetechrepository.com/showthread.php?t=195

Or, I could be half asleep and have missed something before I posted.
If so....
Whatever.
This is just a troll thread anyways.
If this thread still exists when I get to work, I will see what it was I missed.

BTW
Dude, if you want anyone to take your benchmarks seriously...
Match the systems better.
Jeese.
Trolls these days

Edited:
Pointed the links in the correct direction...

 

[jimmysmitty]

^Interesting chart. I have my Corsair XMS2 PC2 8500 @ DDR2 1333 CAS4 (luckily its stable).

So it runs as fast a DDR2 1000 CAS3.

Sweetness......

 

[Zoron]

and then when I tell you that it doesn't matter.

But that's the rub... it DOES matter, and you damn well know it does. Performance almost always increases when you add more RAM. There's a point that it doesn't make much of a difference to be worth mentioning... but 1GB vs. 4GB certainly does make a difference. Now if it were 8GB vs. 4GB, your point might be correct... but it wasn't.

Now you ask us to believe that the AMD system running RAM at 800MHz is somehow crippled compared to the Intel system also running RAM at 800MHz? Hmmmmm. Somethings rotten... and it sure isn't in Denmark.

 

[jaguarskx]

All too little to late and a failure...AMD will have Deneb, that is backwards compatible with AM2+ Motherboards. Intel requires current Core2 users to replace their whole platform including the use of expensive DDR 3 Memory. Intel will stop overclocking on all but their High-end Chips....Enthusiasts will switch to 45nm AMD black edition CPU's.

Nehalem is a flop! Deneb has already impressed technical experts!

Support the Masters AMD! Evil must not win (Intel)!

AMD4Life!!

--<EDIT>--

It has been exlained to you before, and documented with reliable links by several other posters here that Intel will not "stop overclocking"

If you continue to spread this FUD, you will be banned

I suppose petitioning for an IQ Test and scoring above a certain threshold as a requirement before being able to join THG as a member is too much to ask.

 

[Zoron]

^If AMD says so I will let him know. I just helped him put together a kick ass system.

Its that mobo with a 9850BE, 8GB of Corsair DDR2 1066, 2 HD4870 OC to 780/2000, Thermaltake XaserVI and 2 500GB Seagate 7200.11 SATA 3.0 HDDs. Its going to be awesome.

There's proof jimmysmitty said something positive about AMD. Of course, you'll only focus in on the negative and use it as another nail on your cross. Those poor AMD fans... they're soooooo persecuted.

 

[keithlm]

But that's the rub... it DOES matter, and you damn well know it does. Performance almost always increases when you add more RAM.

You and the others are completely incorrect in your assumptions and opinions.

How do I "damn well know" that it DOES NOT matter? I tested it. I removed a stick of ram and re-ran in single channel mode with half the memory. Guess what happened to the results?

http://global.phoronix-test-suite.com/index.php?k=profile&u=keithlm-10318-23121-9735

Most of them stayed about the same. The only ones that changed a lot actually SPEEDED UP.

So... less memory with this benchmark suite did not really change anything. Completely contrary to the claims of jimmysmitty and grimmy and now you. This confirms what I originally said: For Linux and this benchmark suite the added memory does NOT matter enough to be something that needs to looked at. Definitely not something that needs to be discussed any further.

AND GRIMMY:For these benchmarks you can't just pick and choose what benchmarks you want to review from the single results. Why? Because when the UNIVERSE benchmark is run it requires X windows. If you even move the mouse the timed results can change by a few seconds. The full system compile time benchmark you looked at can be run without X-Windows running at all. (I might run one without X-windows tonight to see how much that changes things.) On the other hand... even if that Q6600 result for that benchmark was anomalous it does not mean the rest were.

NOTE: I can only go down to 2Gb because that is the size of both sets of DDR2-1066 ram I have. But I did watch the memory use. It never went above 800Mb for the entire benchmark suite. Go figure. OOPS... I'm wrong. It went up to 1.3Gb in a few of the compressions tests. But it didn't change the times for those... so it is a non issue anyway. But it might have changed the results if you had only 1Gb of memory. But anyway that is also a non-issue... because that topic of conversation was about the Intel system having 2Gb and me having 4Gb. Guess what. It doesn't matter just like I said.

 

[Zoron]

And the benchmark is the only thing running on both systems? Just because the benchmark is using 800MB, doesn't mean the rest of the memory is being used equally, does it?

I'll give you a partial pass since you did halve the amount of RAM... but now how about doubling it on the Intel system? Oh I know you'll whine and complain and say "it won't make a difference!!"... but just humor us, won't you?

 

[croc]

You and the others are completely incorrect in your assumptions and opinions.

How do I "damn well know" that it DOES NOT matter? I tested it. I removed a stick of ram and re-ran in single channel mode with half the memory. Guess what happened to the results?

http://global.phoronix-test-suite.com/index.php?k=profile&u=keithlm-10318-23121-9735

Most of them stayed about the same. The only ones that changed a lot actually SPEEDED UP.

So... less memory with this benchmark suite did not really change anything. Completely contrary to the claims of jimmysmitty and grimmy and now you. This confirms what I originally said: For Linux and this benchmark suite the added memory does NOT matter enough to be something that needs to looked at. Definitely not something that needs to be discussed any further.

AND GRIMMY:For these benchmarks you can't just pick and choose what benchmarks you want to review from the single results. Why? Because when the UNIVERSE benchmark is run it requires X windows. If you even move the mouse the timed results can change by a few seconds. The full system compile time benchmark you looked at can be run without X-Windows running at all. (I might run one without X-windows tonight to see how much that changes things.) On the other hand... even if that Q6600 result for that benchmark was anomalous it does not mean the rest were.

NOTE: I can only go down to 2Gb because that is the size of both sets of DDR2-1066 ram I have. But I did watch the memory use. It never went above 800Mb for the entire benchmark suite. Go figure. OOPS... I'm wrong. It went up to 1.3Gb in a few of the compressions tests. But it didn't change the times for those... so it is a non issue anyway. But it might have changed the results if you had only 1Gb of memory. But anyway that is also a non-issue... because that topic of conversation was about the Intel system having 2Gb and me having 4Gb. Guess what. It doesn't matter just like I said.

What the hell kind of a lame test suite is this, anyway??? It only runs under linux, won't run under pure unix, and is not comparable to any other suites that are commonly used by review sites. How worthless...

And if your mouse slows down the tests under xwindows, maybe you (or the test suite's maintainer) needs to re-compile with a better x-lib. Mouse movements will not xfer across a x-session unless called by the library or application.

I think all posters should use obscure, non-verifiable data to validate their claims. Then no matter what BS they claim, it can't be repeated or verified.

 

[Zoron]

Good idea croc... but if we were to do that and the results were in Intel's favour, good ol' keithy boy would be kicking up a huge stink. Then, he'll say "Well if you can do it, then so can I!"

 

[MU_Engineer]

So let's see....

1. Using a slower RAM, which has been proven to minimally impact the performance of both Intel and AMD machines = wrong.

2. Using 1/2 the RAM, which may significantly impact the performance of the systems = right.

Yes you're right. I'm an Intel fanboy.

You're not very likely to use more than 2 GB RAM while doing any of those benchmarks, except for possibly Enemy Territory:Quake Wars. You might touch 2 GB RAM usage on ETQW if you run a 64-bit version of a very heavyweight distribution with a ton of crap running in the background. Otherwise, you're usually not likely to break 1 GB usage, let alone 2 GB. So I wouldn't exactly throw out the results for the second Q6600 just because it has half of the RAM of the Phenom system, although it is not good form for benchmarking. Compiling the Linux kernel takes a fairly small amount of RAM as it's all C code- I've done it without swapping on machines with 256 MB RAM, so 2 GB would be more than enough.

I would more suspect that the person running the benchmark on the second Q6600 had something else running at the same time that took a bunch of CPU cycles from the kernel compilation rather than swapping a lot, based on the results of the first Q6600 with 4 GB RAM and the Phenom. I used to have access to a Q6600 machine with 4 GB RAM that I could boot Linux on. I'd gladly run the benchmark on it, but I am back at school and don't have access to that machine, unfortunately.

 

[jimmysmitty]

^In the specific test you are right that more memory may not effect it. But in the real world and in most real applications of that the user may be doing more meaning the system will use more RAM.

My main issue though is that when you have less memory you have a bigger chance to rely on virtual RAM which is vastly slower than physical RAM.

And in a testing enviroment I expect equal setups. Meaning equivalent CPU speeds, same amount of RAM, same RAM speed and type (DDR2/DDR3), same HDD setup and so forth. That way all bottlenecks around it can be eliminated and we can focus on the CPUs actual performance.

Then again having the same system setups would be fair.

 

[Zoron]

But jimmy... it doesn't use more that 800MB!!!!!!!!!!!!!!

Damn AMD hater.

 

[Grimmy]

AND GRIMMY:For these benchmarks you can't just pick and choose what benchmarks you want to review from the single results. Why? Because when the UNIVERSE benchmark is run it requires X windows. If you even move the mouse the timed results can change by a few seconds. The full system compile time benchmark you looked at can be run without X-Windows running at all. (I might run one without X-windows tonight to see how much that changes things.) On the other hand... even if that Q6600 result for that benchmark was anomalous it does not mean the rest were.

My original point was to show how skewed the test were. And so you just proved it again, with your own system with less ram, and even trying to get a better score by not running xwindows.