Anandtech i5 preview

Printable View

Show 100 post(s) from this thread on one page

05-30-2009, 11:53 AM
Movieman

Quote:

Originally Posted by bingo13

Your second pass is actually the first pass for the updated x.264 bench, notice the version numbers on each of your tests. ;)

Ok. so in relevance to that chart posted a few posts above are they taking the averages or are they taking the highest numbers?
05-30-2009, 11:57 AM
Hornet331

Quote:

Originally Posted by jmke

doesn't matter one iota that it is or isn't server hardware; they target consumers, I see in your sig you got Core i7 920, the entry level, the kind budget that would buy Core i5 if it had come out first; so now that same you has to choice between two platforms, whereas this is not good a policy; one platform / generation for consumers is best;

If the two platform would have beend released nearly at the same time frame i still would have goen for s1366, soley for the reason that i can upgrade to gulftown.

And the reason why you don't see a higher model in my sig is, that on aircooling theres no reason to go for a higher model, thought i have a feeling that for gulftown another EE cpu might be mine. :D
05-30-2009, 11:59 AM
jmke

Quote:

If the two platform would have beend released nearly at the same time frame i still would have goen for s1366, soley for the reason that i can upgrade to gulftown.

thank you for proving my point that having a two different platforms is not good for the consumer and having one that allows you to install entry level up to high end is best. :)
05-30-2009, 12:05 PM
Hornet331

Quote:

Originally Posted by jmke

thank you for proving my point that having a two different platforms is not good for the consumer and having one that allows you to install entry level up to high end is best. :)

to bad gulftown is a server processor and isn't aimed to be marketed to consumers. :p:

(well depends on what is defined as consumer... i for myself don't see enthusisats as consumers, so S1366 was never a consumer plattform for me to beginn with. :D )
05-30-2009, 12:17 PM
Movieman

Quote:

Originally Posted by Hornet331

to bad gulftown is a server processor and isn't aimed to be marketed to consumers. :p:

(well depends on what is defined as consumer... i for myself don't see enthusisats as consumers, so S1366 was never a consumer plattform for me to beginn with. :D )

I've always had issues with the term server.
That's more a usage of a system than the hardware itself.
no reason at all why you can't use a dual Gainestown with 12-24gig of memory as your daily driver:D
It opens email so fast your eyes can't keep up!:rofl:
05-30-2009, 12:30 PM
Donnie27

Quote:

Originally Posted by highoctane

Considering a vast majority of folks don't even consider upgrading their systems other than the enthusiasts of course, platform choice isn't really going to be major issue. General users buying prebuilt rigs are going to buy their i5, i7 or AMx based rig and run it until the wheels fall off so to speak without ever opening the case.

The difference I see with i7 & i5 vs s754 & s939 is that Intel released the high end platform from the start whereas AMD released the higher end platform later which left early adopter s754 users who might have wanted more performance with a limited upgrade path.

Why do folks keep missing this excellent point?:shrug:
05-30-2009, 12:32 PM
Hornet331

If i could spend more money on hardware i would get a few of those puppies: :D
http://www.dell.com/content/products...555&l=en&s=biz
05-30-2009, 02:40 PM
PetNorth

Quote:

Originally Posted by JumpingJack

That's a good point, I always forget about Trubo (adds extra complexity)... but you are missing something from the Anand article, the engineering sample was locked at 1 step for the engineering sample.

I'm not sure about that at all: http://www.xtremesystems.org/forums/...7&postcount=88
05-30-2009, 03:05 PM
JumpingJack

Quote:

Originally Posted by PetNorth

I'm not sure about that at all: http://www.xtremesystems.org/forums/...7&postcount=88

Anand has the CPU, he clearly noted it would not bump past 1x. Regardless, I can reproduce his numbers pretty easily with my hardware but I will talk to him and see if that was indeed true.
05-30-2009, 03:29 PM
informal

Quote:

Originally Posted by JumpingJack

Anand has the CPU, he clearly noted it would not bump past 1x. Regardless, I can reproduce his numbers pretty easily with my hardware but I will talk to him and see if that was indeed true.

Well those C10 numbers show i5 @2.66 being approx. one bin faster than i7@ 2.66Ghz. The only reason could be turbo boost being higher than on i7 part.i7 920 should do ~2793Mhz ,if not more, in single threaded apps.
05-30-2009, 03:36 PM
Shintai

Quote:

Originally Posted by informal

Well those C10 numbers show i5 @2.66 being approx. one bin faster than i7@ 2.66Ghz. The only reason could be turbo boost being higher than on i7 part.i7 920 should do ~2793Mhz ,if not more, in single threaded apps.

Maybe the Uncore speed. I guess the uncore is OCed on the Lynnfield. There is a bunch of unknown values. Did Turbomode kick in or not. Did it exceed the TDP value define in BIOS etc etc.
05-30-2009, 03:41 PM
JumpingJack

Quote:

Originally Posted by informal

Well those C10 numbers show i5 @2.66 being approx. one bin faster than i7@ 2.66Ghz. The only reason could be turbo boost being higher than on i7 part.i7 920 should do ~2793Mhz ,if not more, in single threaded apps.

Anand has mixed his cinbench results before, Scientia wrote a complete article inferring the incorrect conclusion because he was comparing 32-bit to 64-bit runs, I shot him an email I will see if he can provide an explanation.

32-bit single threaded, for example, I only see +1x bump
http://forums.xcpus.com/gallery/d/96...enchR10-32.jpg

This is because windows scheduler is distributing the load across all cores (real and virtual).
05-30-2009, 03:47 PM
Shintai

Thats quite possible. But again, in the Cinebench there is a 2.5% difference. Could just be a benching/scheduler glitch. Cinebench doesnt seem to like HT in singlethreaded mode.
05-30-2009, 03:48 PM
informal

Quote:

Originally Posted by JumpingJack

Anand has mixed his cinbench results before, Scientia wrote a complete article inferring the incorrect conclusion because he was comparing 32-bit to 64-bit runs, I shot him an email I will see if he can provide an explanation.

32-bit single threaded, for example, I only see +1x bump
http://forums.xcpus.com/gallery/d/96...enchR10-32.jpg

This is because windows scheduler is distributing the load across all cores (real and virtual).

Yeah I hear you.It still up to AT to clarify what exactly was going on with BIOS settings on that lynnfield system.
As for i7 965,shouldn't the 920 have a bit higher Turbo boost kick compared to 965(which gets only "1" bin)?I ask since I think I read at i7 launch that 920 was to have the biggest boost from Turbo being the lowest clocked model(thus having more "headroom" than 2 higher end models,TDP wise).
05-30-2009, 03:50 PM
Shintai

Quote:

Originally Posted by informal

Yeah I hear you.It still up to AT to clarify what exactly was going on with BIOS settings on that lynnfield system.
As for i7 965,shouldn't the 920 have a bit higher Turbo boost kick compared to 965(which gets only "1" bin)?I ask since I think I read at i7 launch that 920 was to have the biggest boost from Turbo being the lowest clocked model(thus having more "headroom" than 2 higher end models,TDP wise).

No, they all have a predefined value in a linear scale. I actually think the retail 2.66Ghz got less turbomode headroom than the 2.8 and 2.93Ghz for Lynnfield. For i7 its all the same. 1x quad, 2x single/dual in C0 stepping.

All artificial limited ofcourse.

There could also be turbomode difference between C0 and D0 stepping. However I dont know and havent tested. But its worth noticing that Xeons (D0) got higher turbomode modes.

And the 920 by AT seems to be an old C0 ES sample for sure (Who knows how thats limited or not). Why dont they test with retail CPUs? :(

However its all starting to be a mess. Both from AMD and Intel.
Uncore speed differences. 200Mhz for example between AM2+ and AM3. And yet another 200Mhz between those and Opterons. Lynnfield also got a 133Mhz (I think) slower uncore than i7.
Then there is all the turbomode issue. A hot testlab could show poor results for example. Next is Hyperthreading etc.

And it certainly wont be easier in the future :p:
05-30-2009, 04:01 PM
informal

Yeah I know. Well this is just an earfly preview,by the time it launches hopefully there will be less unkown variables(mature boards/bios,defined models that have/don't have SMT,fixed Turbo multies etc.)
05-30-2009, 04:07 PM
highoctane

Quote:

Originally Posted by Shintai

For i7 its all the same. 1x quad, 2x single/dual in C0 stepping.

I'm fairly certain that i7 gets 1x for quad/dual & 2x for single C0 stepping, I've never seen 2x outside of one active core on i7 personally.
05-30-2009, 04:09 PM
mAJORD

it's a good preview anyway, but given the variables, not really much point to it (tbh).

The margin of error from having poor functioning turbo mode, unknown uncore freq's etc outweigh anyones estimation of performance anyway! ( in other words, we know roughtly how it will perform given it's still nehalem, and we know dual channel DDR3 has little effect)

I agree with you Shintai, it is a mess, and a lot of reviewers unfortunatly DONT get their head around all the minor differences, and end up doing 'unfair' comparisons or incorrectly simulate other models, causing flamewards on forums :p , or accusations of reviewer brand Bias (when infact they may just have missed the detail)
05-30-2009, 04:25 PM
ap4lifetn

the faster BCLK on the i5 affects performance does it not? or did anandtech overclock the BCLK on the i7 to 166 too and dropped the multiplier? I didn't read the review so maybe someone can fill me in >.<
05-30-2009, 04:32 PM
PetNorth

Quote:

Originally Posted by JumpingJack

Anand has the CPU, he clearly noted it would not bump past 1x. Regardless, I can reproduce his numbers pretty easily with my hardware but I will talk to him and see if that was indeed true.

Can you run with your i7 @2.93 the Cinebench 10 x64 1 CPU?
(Anand uses it in this preview).
05-30-2009, 05:40 PM
JumpingJack

Quote:

Originally Posted by PetNorth

Can you run with your i7 @2.93 the Cinebench 10 x64 1 CPU?
(Anand uses it in this preview).

Yeah, sure will.... give me a bit of time though, I have to spend a bit more time with my son.

Jack
05-30-2009, 05:41 PM
JumpingJack

Quote:

Originally Posted by informal

Yeah I hear you.It still up to AT to clarify what exactly was going on with BIOS settings on that lynnfield system.
As for i7 965,shouldn't the 920 have a bit higher Turbo boost kick compared to 965(which gets only "1" bin)?I ask since I think I read at i7 launch that 920 was to have the biggest boost from Turbo being the lowest clocked model(thus having more "headroom" than 2 higher end models,TDP wise).

It is one of my biggest complaints, not direct at AT, but at the HW sites in general. More or less, much of their test setup information is very lacking, it is hard to make heads or tails of things sometimes.
05-30-2009, 06:10 PM
Sapo84

Quote:

Originally Posted by informal

Well those C10 numbers show i5 @2.66 being approx. one bin faster than i7@ 2.66Ghz. The only reason could be turbo boost being higher than on i7 part.i7 920 should do ~2793Mhz ,if not more, in single threaded apps.

Thanks to the BCLK @ 166Mhz the core i5 was probably running @2822Mhz when the turbo mode was active, and that could explain why it was faster in some benchmarks.
05-30-2009, 07:36 PM
bingo13

Quote:

Originally Posted by Movieman

Ok. so in relevance to that chart posted a few posts above are they taking the averages or are they taking the highest numbers?

Averages...
05-30-2009, 07:39 PM
bingo13

Quote:

Originally Posted by saaya

what?

wha? you tested with the first lynnfield ES chips intel ever sent out, what did you expect anand?
boards not beeing ready, im not surprised, but cpus? they are fine...

Not the first ones by any means.... and the CPU still has a couple of revisions left in it before it is production ready.
05-30-2009, 11:09 PM
saaya

i never saw any problems with them :shrug:
05-31-2009, 01:25 AM
PetNorth

Quote:

Originally Posted by JumpingJack

Yeah, sure will.... give me a bit of time though, I have to spend a bit more time with my son.

Jack

Thanks Jack! no hurry :up:
05-31-2009, 06:47 AM
JumpingJack

Quote:

Originally Posted by PetNorth

Thanks Jack! no hurry :up:

I am working on it.... it will be sometime today.
05-31-2009, 10:34 PM
JumpingJack

Quote:

Originally Posted by PetNorth

Thanks Jack! no hurry :up:

Ok, there is still more work to be done here ... but I have some initial results for you.

Rig Setup:
Asus Rampage II Extreme (BIOS 1306), Core i7 965, 12 Gigs of DDR3-1333 (C9), 2 GTX295s in SLI, 1 9600GT (drives second monitors), 6 HDs (3 2x1 Raid0), Thermalright 120 1366 edition, TT 1200W PSU, Vista Ultimate 64 SP2 with oodles of crud loaded at startup (you should see my tray, very messy not good).

Tests:
I ran tests at 2.66, 2.80, 2.93 GHz (using the core multiplier) -- as such, Asus BIOS disables turbo mode, screen shots are located in the gallery link attached. Each run was observed via CPUID to run at the set multiplier. Due to running at DDR3-1333 the NB frequency was set at 2.66 GHz (has to be 2x the effective mem speed, this I don't think will match Anand's setup). I ran 2 tests at each clock speed, one with HT on and one with HT off. In most cases I ran the 1CPU test twice (not all, but most cases). So most test ran 1CPU/XCPU/1CPU and in cases where I ran twice I report the second run value below. EDIT NOTE -- I turned off the second monitor such that when I did screen captures I did not get 3840x1200 images :) ...

I report the data below as such:
CPU Clock/NB Clock/HT on or off -- 1CPU/XCPU, e.g. 2.66/2.66/On is 2.66 GHz core clock, 2.66 GHz NB clock, and HT is on.

2.66/2.66/Off -- 3709/13165
2.66/2.66/On--- 3501/14735
2.80/2.66/Off -- 3907/13386
2.80/2.66/On -- 3620/15453
2.93/2.66/Off -- 4124/14278
2.93/2.66/On -- 3805/15953

Link to screens (note image names follow same convention as above) -- please check me for typo's:
http://forums.xcpus.com/gallery/v/Ju...sc_001/i7CB10/

So it is indeed possible that Anand's i7 920 turbo'ed by +2, though frankly I have never observed this on the 965 (of course I am using a different stepping than he is). Also, my NB frequency is not likely matching Anand's since I ran at DDR3-1333 and the NB must be 2x or higher the RAM speed. Finally -- Notice the number of processes running -- > 100 (the rig is currently way bloated and it was to much troubles for this initial run to go through and disable all the unnecessary services).

In my experience with i7, I can get +2 turbo on single or dual threaded apps when the app schedules itself exclusive to one core (or two as the case may be). CB10 uses the Vista scheduler (which just sux for the most part), as a result threads are distributed over all cores (HT or no HT), which I believe (not entirely sure) is enough that the turbo algorithm does not kick up into the higher mode.

Nonetheless, his Lynnfield data is still discrepant on this test in my opinion, so I will see what I can do to find some time to flesh this out more. Including a fresh install of Vista without all the bloat.

I do plan to work on this some more, and study just how much impact the NB speed would have -- and with a clean build (no bloat) but it will take a few days.

Jack
05-31-2009, 10:36 PM
JumpingJack

Quote:

Originally Posted by Sapo84

Thanks to the BCLK @ 166Mhz the core i5 was probably running @2822Mhz when the turbo mode was active, and that could explain why it was faster in some benchmarks.

I think it has more to do with it than that -- using Bclk and nothing else OCs the memory and NB, the latter being more important for performance in these simple benches.

Jack
05-31-2009, 10:42 PM
saaya

yepp, bclock doesnt matter for perf, but if you oc the bclock you overclock uncore and mem, since anand didnt detail whether he lowered those multipliers, it explains why his numbers are slightly higher than later 2.66 stock speed retail i5 chips.

thx for all the scores jumpingjack! :toast:
06-01-2009, 01:15 AM
PetNorth

Quote:

Originally Posted by JumpingJack

Ok, there is still more work to be done here ... but I have some initial results for you.

Rig Setup:
Asus Rampage II Extreme (BIOS 1306), Core i7 965, 12 Gigs of DDR3-1333 (C9), 2 GTX295s in SLI, 1 9600GT (drives second monitors), 6 HDs (3 2x1 Raid0), Thermalright 120 1366 edition, TT 1200W PSU, Vista Ultimate 64 SP2 with oodles of crud loaded at startup (you should see my tray, very messy not good).

Tests:
I ran tests at 2.66, 2.80, 2.93 GHz (using the core multiplier) -- as such, Asus BIOS disables turbo mode, screen shots are located in the gallery link attached. Each run was observed via CPUID to run at the set multiplier. Due to running at DDR3-1333 the NB frequency was set at 2.66 GHz (has to be 2x the effective mem speed, this I don't think will match Anand's setup). I ran 2 tests at each clock speed, one with HT on and one with HT off. In most cases I ran the 1CPU test twice (not all, but most cases). So most test ran 1CPU/XCPU/1CPU and in cases where I ran twice I report the second run value below. EDIT NOTE -- I turned off the second monitor such that when I did screen captures I did not get 3840x1200 images :) ...

I report the data below as such:
CPU Clock/NB Clock/HT on or off -- 1CPU/XCPU, e.g. 2.66/2.66/On is 2.66 GHz core clock, 2.66 GHz NB clock, and HT is on.

2.66/2.66/Off -- 3709/13165
2.66/2.66/On--- 3501/14735
2.80/2.66/Off -- 3907/13386
2.80/2.66/On -- 3620/15453
2.93/2.66/Off -- 4124/14278
2.93/2.66/On -- 3805/15953

Link to screens (note image names follow same convention as above) -- please check me for typo's:
http://forums.xcpus.com/gallery/v/Ju...sc_001/i7CB10/

So it is indeed possible that Anand's i7 920 turbo'ed by +2, though frankly I have never observed this on the 965 (of course I am using a different stepping than he is). Also, my NB frequency is not likely matching Anand's since I ran at DDR3-1333 and the NB must be 2x or higher the RAM speed. Finally -- Notice the number of processes running -- > 100 (the rig is currently way bloated and it was to much troubles for this initial run to go through and disable all the unnecessary services).

In my experience with i7, I can get +2 turbo on single or dual threaded apps when the app schedules itself exclusive to one core (or two as the case may be). CB10 uses the Vista scheduler (which just sux for the most part), as a result threads are distributed over all cores (HT or no HT), which I believe (not entirely sure) is enough that the turbo algorithm does not kick up into the higher mode.

Nonetheless, his Lynnfield data is still discrepant on this test in my opinion, so I will see what I can do to find some time to flesh this out more. Including a fresh install of Vista without all the bloat.

I do plan to work on this some more, and study just how much impact the NB speed would have -- and with a clean build (no bloat) but it will take a few days.

Jack

Thanks Jack, really very informative :yepp:

So, comparing your tests with this Anand Lynnfield preview, his i7 920 is running one thread benches (and probably two threads benches) at ~2.93. So, no way that this Lynnfiled is limited to 2.8 with Turbo ON. Like I suspected, Anand statement doesn't seem accurate at all. It's running at least at 2.93, but probably higher.

This Turbo thing confuses a lot :(
06-01-2009, 05:00 AM
informal

Thanks Jack.According to your data,the lynnfield rig AT tested is running higher Turbo with clocks over 2.93Ghz in many cases(CB10 is one as things stand atm).

Show 100 post(s) from this thread on one page