Nehalem-EP......BLOOMFIELD

[JCornell]

Was about to comment that, but im not going down to ur lvl.

XS is idd going down the drain, well maybe not but the quality just aint here anymore.
Im going back to the WC section, its a little better there than the news section anyways.

Hey, don't ... I'll taking this fall ...

[Donnie27]

U dont get the point. All im saying is that u can compare the 2 systems.

Im not saying u should buy the "old" amd system, id go with nehalem myself.
Shure the amd system may beat the nehalem in 8 threaded apps, but nehalem crushes it in singel, double and even quad id guess.
But that is not my point.

You don't get the point either because I didn't say you couldn't or even shouldn't run the test. I advocated running it and many more tests for better or more complete info. I said the results shouldn't be taken out of Context. Now if I'm saying don't do it, how do you get results? I said too much is unknown to draw conclusions from limited information. If he said "Faster on this test". That'd be different as well, get it?

[Movieman]

Wow. I must say it's sad that these (XS) forums have turned to **** . Two and half years ago they were a hell of a lot better.

I joined here in November 2005.. Just about "Two and half years ago"..
There were app 33,000 listed members then.
Now over 80,000.
Like all things in life, the more involved the tougher to control.
Also the more diverse the opinions.
You ignore the part that doesn't interest you and if some peoples posts bother you, you stick them on your ignore list.
Very simple way to deal with it.

[eXa]

You don't get the point either because I didn't say you couldn't or even shouldn't run the test. I advocated running it and many more tests for better or more complete info. I said the results shouldn't be taken out of Context. Now if I'm saying don't do it, how do you get results? I said too much is unknown to draw conclusions from limited information. If he said "Faster on this test". That'd be different as well, get it?

I havent talked about results at all. What bugs me is the attitude of many people here.

If 2 systems can do the same job (be it cinebench, winrar superpi, server stuff or whatever) u can compare the 2. It really is that simple.
If its fair or not is not relevant either

[Movieman]

I havent talked about results at all. What bugs me is the attitude of many people here.

If 2 systems can do the same job (be it cinebench, winrar superpi, server stuff or whatever) u can compare the 2. It really is that simple.
If its fair or not is not relevant either

Yes, you can compare them but what is the relavance of that comparision?
If I run a benchmark on my old single core FX51 and then run that same benchmark on my 8 core Harpertown what am I learning from that time spent?
Something that I already know, the Harper is faster.
You only get relavance when you compare one apple to one apple, not one to eight or four to eight.. I think that was the whole issue here.

[Hornet331]

just as movieman said... whats the point of such comparisons?

but maybe its my fault for putting all systems into one graph, maybe if someone can provide phenom benches @ 3ghz i could add them in direct comparison to bloomfield and core2.

[JCornell]

just as movieman said... whats the point of such comparisons?

but maybe its my fault for putting all systems into one graph, maybe if someone can provide phenom benches @ 3ghz i could add them in direct comparison to bloomfield and core2.

No, No ... You did a great job ... I personally was lazy to sum it all into excel charts ...

[Aberration]

Yes, you can compare them but what is the relavance of that comparision?
If I run a benchmark on my old single core FX51 and then run that same benchmark on my 8 core Harpertown what am I learning from that time spent?
Something that I already know, the Harper is faster.
You only get relavance when you compare one apple to one apple, not one to eight or four to eight.. I think that was the whole issue here.

Dont we constantly compare 1GPU systems to multiGPU systems?

[Shintai]

Dont we constantly compare 1GPU systems to multiGPU systems?

And the price range is what?

Or you want to see a E2160 compared with a 4P Tigerton system?

[Movieman]

Dont we constantly compare 1GPU systems to multiGPU systems?

HMM, interesting point, how to reply.
Now remember that for what I do a high end vid card is not something I use BUT with my limited knowledge I can say this in rebuttal:
On the ORB don't they distinguish between scores submitted between single card and SLI or Crossfire equiped machines as it is not reasonable to compare one to the other..
The defense rests..

[Aberration]

I wouldnt know, I dont look at ORB. But I know you have seen all the screens of the benchmarks done for the various cards over the last few years. Even for CPU's we see comparisions of Quad to Dual. Is that not the same?

And Shintai, in most benchmarks you can see 3850's compared to a 4870CF or GTX280 SLI, who cares about price.

I think relative gains in performance is good to know. Sure I think it wouldnt be useful to compare an FX51, but that doesnt mean it is irrelevant. What if the FX51 is the only thing you have now, and you want to guage the reletive performace increase you would see to the Bloomfield and maybe decide that for the $$$ a E8400 would be good enough?

I just dont see the problem with comparing a Bloomfield Quad with HT to a 4x4 Phenom. I just dont see that being the same as some of the comparisons that people are liking it to.

[eXa]

Yes, you can compare them but what is the relavance of that comparision?

As long as the apps are multithreaded its very relevant. If the 4core with HT performed identically to the 8core barcelonas that would that would be very interresting and relevant. No?

what that test did show is that the 8core barcelonas wasnt twice as fast as the nehalem. Ofc that was expected since each nehalem core is much faster than each opteron core. But the test show how much difference it is between the 2. (in that test atleast)

If I run a benchmark on my old single core FX51 and then run that same benchmark on my 8 core Harpertown what am I learning from that time spent?
Something that I already know, the Harper is faster.

If u go with that, why bother testing anything, u already know its faster...

You only get relavance when you compare one apple to one apple, not one to eight or four to eight.. I think that was the whole issue here.

Matter of definition. Comparing one computer to another computer, thats apples to apples in my eyes...

[Donnie27]

Wow. I must say it's sad that these (XS) forums have turned to **** . Two and half years ago they were a hell of a lot better.

2.5 to 3 years ago you had the same kinds of fights. AMD fans were flaming Intel guys about Faked Conroe tests. A year before that Lithan, Petnorth and gang lived in the Intel section bashing any one with anything good to say about Intel for any reason, not just processors. The sad part was there would be 200 folks in the AMD section and 19 in the Intel section and 4 of those 19 were AMD Fans bashing the other 15. I remember Fugger making Lithan take back comments he made to 4 folks! Hell, look up Duvie LOL?

http://www.xtremesystems.org/forums/...+Lithan&page=3

[Donnie27]

I havent talked about results at all. What bugs me is the attitude of many people here.

If 2 systems can do the same job (be it cinebench, winrar superpi, server stuff or whatever) u can compare the 2. It really is that simple.
If its fair or not is not relevant either

You get no argument from me there at all. Look at what I said, I said add more to give it a better prospective. ALL of the numbers would be interesting. Not pull any one *line out.

I have NOT even thought of buying two B2 Phenoms knowing they have a problem that will make them even slower? Look at the list I gave? His comparison is there. I added two more to it. I'd interested in seeing the best and worse case scenario.

Here?

To get a full set of numbers, run;
Intel vs AMD in that order.
One core on each,
4 cores on each with HT disabled,
4 cores plus HT vs 4 cores,
4 cores plus HT vs 8 cores as he's saying,
8 cores with HT disabled vs 8 cores,
And finally 8 Cores vs. 8 Cores with HT enabled.

Blue = Intel, Black AMD. I asked him what was wrong with that? Most Geeks here I know would love that info. It is not about who comes out on top. I bet folks would talk about how HT changed or didn't the results, more than what or who won, don't you?

[Shintai]

Matter of definition. Comparing one computer to another computer, thats apples to apples in my eyes...

Look, this Puma platform with an X2 3800 CPU. A shame it gets obliterated against this Skulltrail with quad SLi and 2 heavily OCed QX9770.

Apples to apples....right?

All you and LIKMARK want is to make K10 look good against Nehalem. In whatever way you can. And now calling it apples to apples?
Just a shame it was 2 quads vs 1 quad....

[Hornet331]

Most Geeks here I know would love that info. It is not about who comes out on top. I bet folks would talk about how HT changed or didn't the results, more than what or who won, don't you?

count me in for that.

[informal]

LIKMARK's intent was to compare how efficent is SMT on Nehalem,that's the way i look at it.So he wanted to know how 8 threads in Bloomfield system compare with his 8 thread(2x RevBA).
I already showed you in my previous posts that LIKMARK's Barcelona system is around 25% per clock slower than jan's RevB3 system in Ceinebench10.
So if anything ,LIKMARK's system is actually holding its own against the new Nehalem system.The fact that it's 2xQC vs 1 QC(8 thread) chip doesn't make it apples to apples comparison,but it's valid as any other similar comparison when it comes to comparing previous gen. vs next generation chips.Remember ,while AMD is at its first attempt with Quad Cores and still on DDR2,intel is on it's third generation(65nm C2Q,45nm C2Q,45nm Nehalem).So basically you compare first TLB patch plagued RevBA system with yet unreleased 3rd generation intel quad core system.
The more valid comparison would be single Deneb versus single Bloomfield ,all at the same frequency.That's what we will have in Q4.Even better,a Deneb running a same DDR3 memory(AM3) as Bloomfield so we can have as close as possible memory subsystems,even tho Nehalem is better in that respect thanks to 3 channel IMC.

[RaZz!]

guys, just do the comparison benchmarks, and laugh/whine for a minute and leave it at that - just to end this crap here.

you're running in circles - there's so much useful information in this thread that's totally overshadowed by such useless discussions.

[Donnie27]

LIKMARK's intent was to compare how efficent is SMT on Nehalem,that's the way i look at it.So he wanted to know how 8 threads in Bloomfield system compare with his 8 thread(2x RevBA).
I already showed you in my previous posts that LIKMARK's Barcelona system is around 25% per clock slower than jan's RevB3 system in Ceinebench10.
So if anything ,LIKMARK's system is actually holding its own against the new Nehalem system.The fact that it's 2xQC vs 1 QC(8 thread) chip doesn't make it apples to apples comparison,but it's valid as any other similar comparison when it comes to comparing previous gen. vs next generation chips.Remember ,while AMD is at its first attempt with Quad Cores and still on DDR2,intel is on it's third generation(65nm C2Q,45nm C2Q,45nm Nehalem).So basically you compare first TLB patch plagued RevBA system with yet unreleased 3rd generation intel quad core system.
The more valid comparison would be single Deneb versus single Bloomfield ,all at the same frequency.That's what we will have in Q4.Even better,a Deneb running a same DDR3 memory(AM3) as Bloomfield so we can have as close as possible memory subsystems,even tho Nehalem is better in that respect thanks to 3 channel IMC.

Sure if that Xeon vs Opteron. Good point about the RAM. Also 45nm Phonem for Desktop vs. Bloomfield

Then I'd love to see a wide range of test apps as well. Not one or two that one's better at.

[informal]

Yes,a whole range of test is needed,covering all bases(professional->like cad/cam,compiling,web server etc. and non-professional like games ,encoding,archiving etc.).

[FlanK3r]

winrar bench my X3 8650@2750MHz

this about 1500, Nehalem on 2.66Ghz 2500KB/s !!!OMG, nice "speed"..
How much is on 2.99Ghz? Try it anyone

[XS Janus]

For just a casual observer looking for new posts on Nehalem benches this threat is out of control.

Maybe there should be a sticky thread, not for bickering and flaming but just with graphs in the first post being updated and screenshots being posted by someone who can be trusted to do a fair job of it.
This is chaos, and already the graphs in this thread cannot be found by 90% of the viewer of this thread.

Just a thought...
:/

[Boschwanza]

winrar bench my X3 8650@2750MHz
this about 1500, Nehalem on 2.66Ghz 2500KB/s !!!OMG, nice "speed"..
How much is on 2.99Ghz? Try it anyone

A tweaked Phenom with high NB Clocks and DDR2 1100 makes around 2200kb/s at 2.75 GHz. B/c this bench is very Memory related Deneb would even score higher with DDR3.

[foch3 -USA-]

You can compare anything you want, its the conclusion you reach that's important.

[KeZzZu]

/ot

Well i don't know about you guys, but AMD section losed again one valuable memeber, thanks alot. Getting harder to get any iformation about amd systems.

ot off/

Thanks JC for result's, hoping prices to drop down on DDR 3 so i can buy one of those for testing :P and how's the triple channel performance vs single channel, i think i have missed that :S