Intel Core i7 On Linux [Archive]

phoronix

04-07-2009, 07:50 AM

Phoronix: Intel Core i7 On Linux

Back in November Intel had formally launched the Core i7 series, but Linux benchmarks were not to be found. However, in the weeks that followed, results began to emerge through the Phoronix Test Suite and Phoronix Global. Over time there were many more Linux test results from the Phoronix Test Suite community, but now we have posted some of our own Intel Core i7 numbers from Ubuntu Linux.

http://www.phoronix.com/vr.php?view=13681

RealNC

04-07-2009, 08:51 AM

Something's missing here... I've no idea what those numbers mean. How much faster is this than my 3.33GHz Core 2 for example?

korpenkraxar

04-07-2009, 09:00 AM

Something's missing here... I've no idea what those numbers mean. How much faster is this than my 3.33GHz Core 2 for example?

Or a Phenom or Opteron for about the same $$$.

Michael

04-07-2009, 09:00 AM

Something's missing here... I've no idea what those numbers mean. How much faster is this than my 3.33GHz Core 2 for example?

Run phoronix-test-suite benchmark michael-23540-28957-28437 and you will know precisely what it means compared to your Core 2 3.33GHz system.

RealNC

04-07-2009, 09:01 AM

Except that we're not testing on the same setup. I would need to setup an ubuntu :P

mattst88

04-07-2009, 09:11 AM

Run phoronix-test-suite benchmark michael-23540-28957-28437 and you will know precisely what it means compared to your Core 2 3.33GHz system.

That's fine and well, but for the sake of the article -- there's nothing to compare the i7 to. Surely you see our point?

ayumu

04-07-2009, 09:27 AM

Wouldn't have hurt to include q9650 or q9550 to have something to compare with.

I sort of totally expected at least that much.

Michael

04-07-2009, 10:20 AM

Wouldn't have hurt to include q9650 or q9550 to have something to compare with.

I sort of totally expected at least that much.

If only Intel actually sent out processors to Phoronix...

gtrawoger

04-07-2009, 10:45 AM

People, you are missing the idea. What Michael wants you to do is post YOUR results from YOUR setup.

That way, when someone post his from a Q9XXX, you will know. You can't expect Michael to have all those processors.

As soon as I get Phoronix Test Suite installed, I will post.

Tares

04-07-2009, 11:24 AM

Well E8400 would be nice to add it in the charts, as for c2d family.

GreatWalrus

04-07-2009, 12:09 PM

Thanks so much for reviewing this processor, I know it is not a cheap one to purchase and go benchmark.

It really doesn't take much time or intelligence to go to Phoronix global (http://global.phoronix-test-suite.com) and search (http://global.phoronix-test-suite.com/?k=search) for a system with similar specifications to compare processors. Unless the processor benchmarks are really close, it isn't going to matter a whole lot of some of the hardware like ram or motherboard is different.

By the way, did you take any temperature readings or remember what they were like at the different clock increments? I would be interested to know :)

If I ever get 5 hours (:p), I will benchmark my system against this since I might be getting an i7 in the future.

Fixxer_Linux

04-07-2009, 12:25 PM

The shame in that story is that no manufacturer / retailer or any big company has been able to send a sample review to Phoronix.

It's a real shame, knowing that there is not so much Linux site hardware review and that I guess that Phoronix has many visitors...

GreatWalrus

04-07-2009, 12:33 PM

Kano

04-07-2009, 01:15 PM

Well some tests did not work with PTS 1.8.0 final - also Nexuiz is now a different test.

http://global.phoronix-test-suite.com/?k=profile&u=kano-628-6932-9485

Very interesting is the Parallel BZIP2 Compression test which is much faster on my test system. The kernel is basically the same, just the basesystem is Debian 5.0.

cruiseoveride

04-07-2009, 02:01 PM

FAIL

This was such a pointless benchmark.

Its quite obvious what the results were going to be when you put 3 i7s against each other.

Without Phenoms and C2D/Quads this article is pointless

GreatWalrus

04-07-2009, 02:31 PM

FAIL

This was such a pointless benchmark.

Its quite obvious what the results were going to be when you put 3 i7s against each other.

Without Phenoms and C2D/Quads this article is pointless
Here we go again, cruiseoveride. I still have not seen a positive post from you.

This article is supposed to be informative and engaging of readers to go test their system against the i7, something that takes advantage of the newly release PTS 1.8, which was big news.

Seriously, amaze me and appreciate something for once.

poofyyoda

04-07-2009, 04:27 PM

mycroes

04-07-2009, 04:39 PM

Although I understand everyone can benchmark his own machine, we can't all buy a machine to benchmark against the core i7. I could of course benchmark my C2D E8400 (and I will when I have more than just a few hours of free time), but I can't benchmark it against any of AMD's recent offerings because I simply don't have anything in AMD's recent offerings. Also phoronix could benchmark the E8400 with DDR3 memory, while I can't (iirc both an E8400 has shown up in tests as well as a socket 775 motherboard with DDR3 memory).

Anyway, it would be nice to see more of a comparison IN the article, but if that's not gonna happen then we'll have to do with what's offered right now.
Regards,

Michael

superppl

04-07-2009, 05:49 PM

we will be back shortly with more benchmarks, including virtualization benchmarks, but we just wanted to get this information and numbers out there now for people interested in seeing how their system stacks up
That should have been in bold. Seriously, do you guys even read the article, or just blast through the graphs?

Anyway, I haven't bought a computer in a few years, and I just recently discovered Nvidia's coverup with the 8 series gpus, and seeing how Intel and Nvidia are nowhere near as generous as AMD when it comes to Phoronix, I think I know what my next purchase will be. :)
Also, Intel can take their chips and shove it. I refuse to buy anything Intel until they either come out with a reasonable gpu or drop out of the market entirely. :p

cruiseoveride

04-07-2009, 06:14 PM

Here we go again, cruiseoveride. I still have not seen a positive post from you.

This article is supposed to be informative and engaging of readers to go test their system against the i7, something that takes advantage of the newly release PTS 1.8, which was big news.

Seriously, amaze me and appreciate something for once.

http://www.youtube.com/watch?v=ulVDM0a49Lw

I try. And I try. But i dont get no.. all this useless information... its meant to drive my imagination?

TheK

04-07-2009, 08:00 PM

Very interesting is the Parallel BZIP2 Compression test which is much faster on my test system. The kernel is basically the same, just the basesystem is Debian 5.0.

Seams to be one of the cases, where the i7 really sucks. Another extreme example of this is the Windows game "Bioshock", where the 3 GHz i7 gets beaten by an 2,33 GHz C2Q. Overall for gaming the 220€ Phenom II 940 isn't slower than the 560€ Core i7 940...

MaestroMaus

04-08-2009, 01:35 AM

Typo on page 1

We were able to push this processor any further than 3.60GHz with this motherboard and the stock cooling before running into stability problems.

It should be We were not...

lordmozilla

04-08-2009, 07:29 AM

a direct comparison to a Q9xxx/Q6xxx quad core cpu's would have been great. Not to mention a Phenom II thrown in there.

Also is there a way of disabling HT and trying again? I would love to see the performance gains from this new generation of hyperthreading.

nightalon

04-08-2009, 04:32 PM

It would be nice to see a low-end and a high-end CPU also compared on the same test bed in the dual-core and quad-core categories.

Additionally, one could try to make comparisons with Windows and Mac to see which operating systems take advantage of new features like added cores and sse3-sse4.1 best.

To mention nothing about AMD.

I know this is tough since many sites (Tom's Hardware comes to mind) benchmark processors in Windows, but comparatively few do on Linux!

downhillgames

04-08-2009, 06:49 PM

Rarely do I post here anymore.

People, you are missing the idea. What Michael wants you to do is post YOUR results from YOUR setup.

That way, when someone post his from a Q9XXX, you will know. You can't expect Michael to have all those processors.

As soon as I get Phoronix Test Suite installed, I will post.
I agree in the sense of the community generating meaningful results itself; results which can then be compared against each other. But, I think most people here, myself included, were hoping for something in the article to compare it against. After all, not everyone has the same OS configuration, nor Phoronix Test Suite, nor Linux for that matter. What about the audience looking for i7 Linux performance who 1) have never used it, and 2) don't have an i7 but may be considering buying one & installing Linux on that rig (or any single part of that)?

Don't get me wrong. I love the Phoronix Test Suite for what it is, and I'm very grateful to Michael for providing this critical piece of software for free. Also, I have no reason to believe Michael isn't acting with his best intentions. However, I'm with everyone else on this issue: It's hard to make a judgment call with just one chip that's been overclocked.

(tl;dr) Put shortly: Perhaps a follow-up article with community-submitted benches of comparable systems would be something Michael could consider.

just my 2c.

(btw, didn't mean to single you out, gtrawoger. nothing personal!)

gtrawoger

04-08-2009, 10:39 PM

meh ... no offense taken. I just thought I'd point out the intent of Michael's providing the command for the test. I actually thought it would be a pretty neat idea to have the community provide some benchmarks from their rigs and have larger pool of stats to pull from.

Now, all we'd need now is a way to make something out of that data.

downhillgames

04-09-2009, 11:32 AM

RobbieAB

04-09-2009, 02:04 PM

Also is there a way of disabling HT and trying again? I would love to see the performance gains from this new generation of hyperthreading.

Assuming there is one...

The problem with Hyperthreading on multi-core is the scheduler needs to know which cores are virtual and which are not. As the whole point of hyperthreading is to hide such information...

Virtual cores sounds great in theory, but don't always work out. It is possible to get a scenario where the schedulers mess up and load 4 threads on 2 cores, and the two virtual cores, and leave the other two cores unloaded. This is not a problem once you hit 8+ threads (strictly speaking 7+), or when you only have one seriously heavy load thread, but any where in between there is scope to get it wrong. On a single core system, this is, of course, totally irrelevant as it isn't actually possible to assign processes to the "wrong" core.

bnolsen

04-12-2009, 11:15 PM

Assuming there is one...

The problem with Hyperthreading on multi-core is the scheduler needs to know which cores are virtual and which are not. As the whole point of hyperthreading is to hide such information...

The linux kernel is aware of the difference between real and virtual cores and tries to schedule with those in mind. That was added back when the P4 went hyperthreaded.

The core i7 is more super scalar than the P4 and should be able to scale much better.

bnolsen

04-12-2009, 11:24 PM

So, I installed Wallbuntu 9.04 Beta 1 on some spare space I had on this HDD here.
http://img18.imageshack.us/img18/1015/screenshot15.th.png (http://img18.imageshack.us/img18/1015/screenshot15.png)

The benchmark results may be found here: http://global.phoronix-test-suite.com/?k=profile&u=downhill-6937-12014-280

Ugh, original phenom.

Below is a full battery of tests from a phenom2 x3 which seems to generally outperform your original phenom.

http://global.phoronix-test-suite.com/?k=profile&u=nico-26876-29519-12230

If anything, the global test results database is a PITA to navigate. I tend to agree with most folks on here. Benchmarking a single platform is about useless. An existing frame of reference...*some* frame of reference would be nice.

RobbieAB

04-13-2009, 12:02 AM

The linux kernel is aware of the difference between real and virtual cores and tries to schedule with those in mind. That was added back when the P4 went hyperthreaded.

The core i7 is more super scalar than the P4 and should be able to scale much better.

So the Linux kernel, compiled with the correct options, can schedule correctly? Nice, so long as more complexity in the kernel doesn't cause any slowdowns...

IIRC, Hyper-Threading in the P4s was estimated to cause a 30% slowdown under full load. I will confess to never having verified that figure, but my general experience with it was the big gain from hyper-threading lay in the interactivity area, and not in "throughput" which is what most benchmarks test.

gtrawoger

04-13-2009, 08:32 AM

So, I installed Wallbuntu 9.04 Beta 1 on some spare space I had on this HDD here.
http://img18.imageshack.us/img18/1015/screenshot15.th.png (http://img18.imageshack.us/img18/1015/screenshot15.png)

The benchmark results may be found here: http://global.phoronix-test-suite.com/?k=profile&u=downhill-6937-12014-280

Here is my test-result. Are most of the tests in here multi-threated? Maybe that's why the Core i7 beats the pants off my 8400 @ 4Ghz.

http://global.phoronix-test-suite.com/index.php?k=profile&u=gtrawoger-5512-2646-27047

gtrawoger

04-13-2009, 10:49 AM

Here is a benchmark of an X2 5200. http://global.phoronix-test-suite.com/index.php?k=profile&u=gtrawoger-17876-16696-29149

I also agree, and nice summary or total view of all submitted benchmark would be nice.

bnolsen

04-13-2009, 12:48 PM

So the Linux kernel, compiled with the correct options, can schedule correctly? Nice, so long as more complexity in the kernel doesn't cause any slowdowns...

IIRC, Hyper-Threading in the P4s was estimated to cause a 30% slowdown under full load. I will confess to never having verified that figure, but my general experience with it was the big gain from hyper-threading lay in the interactivity area, and not in "throughput" which is what most benchmarks test.

The core i7 is definitely NOT a P4. You could say this is hyperthreading 2.0. Seems intel got it right this time.

The virtual cores don't double your performance, but it seems now that it almost never subtracts from performance. I suspect it has to do with how many more execution units the core i7 has compared with any other x86_64 compatible cpu available today.

The atom also hyperthreads decently.

RobbieAB

04-13-2009, 01:30 PM

The core i7 is definitely NOT a P4. You could say this is hyperthreading 2.0. Seems intel got it right this time.

The virtual cores don't double your performance, but it seems now that it almost never subtracts from performance. I suspect it has to do with how many more execution units the core i7 has compared with any other x86_64 compatible cpu available today.

The atom also hyperthreads decently.

I am aware of the fact that i7 is not P4. ;) However, the basic point still stands: Hyperthreading increases complexity in managing multiple processes. 99% of the time it doesn't matter, but I would be very surprised if the heavily tuned science apps don't show slowdown when hyper-threading is enabled, simply because most of them are tuned to saturate the CPU, with minimal breaks in the pipeline, a class of optimisation I have NEVER heard discussed since I left Uni, mostly because outside of that specialised field people don't bother with it much. These are the kind of apps where a P4 @ 3.6GHz could actually compete with the early C2Ds simply because they COULD avoid the pipeline breaks that killed the P4 in general usage.

Hyper-threading is about interactivity, not throughput. It sacrifices complexity to try and drop latency. For desktops it will often manifest as a throughput boost, but that is related to the fact that desktop systems just aren't tuned that heavily, and frequently hit pipeline breaks, and not due to any kind of hyperthreading magic.

Core i7 is killing pretty much all the competition because it is a good core, designed to do a job, and running a generation ahead of it's competitors, who also seem to have decided that pushing the CPU itself is a mugs game: Platform performance is where it is at, not raw CPU power.

Kano

04-13-2009, 01:50 PM

Well i7 is not for mass market, i5 will challenge AMD much more...

bnolsen

04-13-2009, 05:20 PM

I am aware of the fact that i7 is not P4. ;) However, the basic point still stands: Hyperthreading increases complexity in managing multiple processes. 99% of the time it doesn't matter, but I would be very surprised if the heavily tuned science apps don't show slowdown when hyper-threading is enabled, simply because most of them are tuned to saturate the CPU, with minimal breaks in the pipeline, a class of optimisation I have NEVER heard discussed since I left Uni, mostly because outside of that specialised field people don't bother with it much. These are the kind of apps where a P4 @ 3.6GHz could actually compete with the early C2Ds simply because they COULD avoid the pipeline breaks that killed the P4 in general usage.

I will tell you if this is true or not later this week. I happen to work on a highly threaded high throughput aerial/close range photogrammetry package that does auto feature matching, photogrammetric bundle adjustment, radiometric computations and ortho rectification using both line scanner and frame imagery. A customer is building a dual quad core i7 with 24G of ram that I'll run some tests on.

RobbieAB

04-13-2009, 05:29 PM

I will tell you if this is true or not later this week. I happen to work on a highly threaded high throughput aerial/close range photogrammetry package that does auto feature matching, photogrammetric bundle adjustment, radiometric computations and ortho rectification using both line scanner and frame imagery. A customer is building a dual quad core i7 with 24G of ram that I'll run some tests on.

And I will withhold judgement until I've seen the code. ;)

How much of that task is logic (IF statements) and how much is straight computation? How many instructions can be run without the output of the last one? It sounds a nice "hard" problem, but the actual structure of the algorithms are crucial for this kind of work.

As I said, the only people I have heard playing those games tend to be doing things like self consistent field solutions to Schrodingers Equation for many body systems, the kind of task that uses large clusters, and ties them up for weeks at a go. The kind of task that uses Fortran for critical numeric sections, because it is easier for the compiler to optimise than C (yes, this IS true, just mostly irrelevant for 99% of the worlds usage).

bnolsen

04-13-2009, 05:33 PM

RobbieAB

04-13-2009, 05:38 PM

The code is very branchy and has lots of input IO.
Here's a screenshot of the program running on windows multicore.
This is just after the application started so the "packets" are still somewhat synchronized at this point and fighting over resources.

http://pushbroom.org/windows.gif

Um... branchy code means LOTS of pipeline breaks, which is the kind of task where hyper-threading will help.

I was thinking in terms of 10,000+ square matrix mashing as the kind of task which avoids pipeline breaks. As I said, very much science cluster work.

bnolsen

04-29-2009, 12:21 PM

Um... branchy code means LOTS of pipeline breaks, which is the kind of task where hyper-threading will help.

I was thinking in terms of 10,000+ square matrix mashing as the kind of task which avoids pipeline breaks. As I said, very much science cluster work.

Just got the machine installed on Monday. Yesterday was able to start running real tests.

Configuration: 14GB ram, E5504's @ 2.0GHz (no hyperthreading....grr)

So far testing on our most mixed processing load shows performance only within ~10-15% of a 2.33GHz harpertown (E5410). This is still very cpu heavy but requires lots of streaming IO, both input and output.