Why the Core i7 may not be a good enthusiast processor

I think that they are. From the comparison of the Core i5 540M with the P8700 (both are clocked at 2.53GHz), you can see that there is no difference at idle or in various sustained load conditions. You are right, if you let it run at maximum power with maximum load, the battery life will be shorter -- but the total amount of work it will do in that time is greater so you gain rather than lose. They're more or less at level of the P-series C2Ds... the obvious question is (and the article asks and answers), why aren't they better?

It looks like all of the Core i5s and i7s (both desktop and laptop) overclock to some extent even when all cores are in use. The problem is that the details aren't published anywhere -- for the desktop CPUs, the various review sites had the info, but for laptops Intel just quotes an upper limit and most people assume that Turbo Mode only works with one core.

 

Actually it is published. The clock multipliers were listed on NBR a while back for each situation (ie. 720QM would have different numbers for 4/3/2/1 core(s) being used). Even when all cores are used, Turbo Boost can be utilized.

 

i totally agree with the op.
i read in anandtech that the real purpose of the i7 cpus was to bring overclocking to the masses AND prevent overclockers from transforming midlevel cpus into highend performers.
one of the things they did to the i7 is physically remove the frontside bus, which is a data pipeline between the cpu and the motherboard. increasing this pipline makes data transfer from cpu ( and other components that rely on the northbridge) and motherboard faster. theoretically, removing this pipeline should make the computer faster. however you do lose the ability to control cpu speeds via fsb since it is not there anymore.
in desktop cpu's and bios'es they have something called BCLK. i think this is what the op is referring to when he means increasing the fsb(which is not present anymore) in the i7's. i dont have a desktop and im no enthusiast myself, but i know increasing the BClk has almost the same effect as the old fsb- that is, it allows the cpu to be overclocked. and afaik, desktop bios'es also allow for turning off the turbo boost feature.
so i think its just a matter of time till we get i7 overclocking in notebooks. just like during the fsb days, the first solutions were bios implementations followed by applications that allowed changing overclocking settings on the fly , within the os.

 

Shirleyfu posted some benchmarks on the i7-620M a week or so ago that showed that the extra physical cores really aren't helping the i7-720QM all that much even in multi-threaded apps.

http://forum.notebookreview.com/showthread.php?t=455347

If you're heavily taxing more than 4 cores a lot then the 720QM would probably be a better bet but for gaming (especially if you play WoW) the 620M would also be my best choice at that price point. If you're willing to spend more however then a faster Clarksfield would beat the Arrandale.

With the C2Q it was better to go for the quads in most cases because there was no hyper-threading to deal with bad console ports and like you've been pointing out it in this thread it was usually possible to overclock all 4 cores to speeds as fast as the dual-cores.

But then again even if/when we gain a method to manually overclock the i7-620M could still be the better choice because of it's 32nm die.

 

this would probably hold true until applications are optimized to use four or more cores, which could be very soon.
during the switch from single to dual era, many people were skeptical - including myself, as how a 2.2ghz dual core could be so much better than a 4ghz pentium. we just had to wait for the software to come out, and we never looked back to using single core cpu's again.
the jump from 2core to 4+core optimized software could be an easier transition for developers than it was from the single to dual core applications. plus win7 itself is already multithreaded, so 4+ core optimized 'everyday' programs could be with us very soon.

 

While applications might not be fully optimized for multiple cores at the moment, it was shown in current applications that supported multiple cores that the difference between 620M and 720QM wasn't significant. They were both on par for these programs, while the higher clock speed of the 620M was preferred in single or lightly threaded applications. Going a step further, even if you went with the 520M or 540M (according to notebookcheck), the difference in performance wouldn't be much (as compared to the power consumption and price increase of the 620M and/or 720QM).

 

I'm going to chalk one up for the developers. If you're using the right language with the proper abilities (I myself am using C++ and OpenMP for my university work), making a multi-threaded program is often quite easy.

 

I code extensively using C++/Fortran and OpenMP/MPI myself. While some programs are easily to parallelize (e.g., those which use algorithms which are "embarrassingly parallel"), many are not. In practice, *very* few algorithms scale linearly (or better). Essentially, non-parallelizable portions of programs limit the gains from multi-threading.

We just have to acknowledge that Intel and co. switch to multi-core primarily due to PHYSICAL constraints that limit frequency scaling. They simply don't have any others Aces up their sleeve.

The average consumer experiences far better performance with x cores running at frequency y as compared to 2x cores running at frequency y/2. Intel will launch a hexacore for the desktop fairly soon. but we're not even at a point today where most consumers are better off with a triple or quad core...

 

Mmmm, graphene transistors

 

You guys are not getting it. I completely agree with the OP. Your CPU will be running high clocks only when you have an application that only uses 2 cores. Applications being more and more multithreaded will make use of all your cores so your once in a while 2.8GHz automatic OC or whatever is useless. Also from an OC point of view, you like your CPU to keep the same frequency all the time, jumping between 1.6GHz and 2.8GHz is annoying. I'm sure that I won't upgrade to i7 until I have to upgrade laptops due to MXM compatibility. My QX9300 at 3.15GHz 24/7 beats an overclocked i7 920XM because you can only overclock the Turbo frequency on the mobile i7. I'm not saying that the i7 is a bad serie, but it just doesn't compare to the high end C2D Quad Cores in my opinion.

 

This has been a very interesting thread. I am shopping for a laptop for music production and only been considering core i7 machines. Would I be better off getting one of these c2d quad machines that can be OC'd? If so, which model?

 

i got myself an i5. it dances around my old core2quad, as it overclocks as needed. i wouldn't manually tweak a music production system, ever. the i5 have high clockspeeds and can get very high. and they behave as smooth as quadcores.

good bang for the buck imho.

 

i don't know if this has been said, but the turbo boost only works based on the the total processor usage AND the vcore/thermal properties at whatever clock speed it's currently at. your CPU will not raise the mutliplier to x21 and melt itself.

if you want to overclock the i7, go into the BIOS and turn off speed step and turn off the turbo boost if that makes you happy.

 

i use a core i7 and i have noted that while video editing the program makes use of all 8 cores at 60 to 80%. afaik music production, like video editing software, highly utilizes a computers multithreading capabilities since they are very processor intensive. and i have read in the past that music production software are highly optimized because of this.
i think you will see a major difference between an i7 and a o/c'ed quad cpu. i believe atm the applications that benefit from i7 technology are video editing, music production, gaming, engineering s/w, and possibly animation s/w.

 

8 threads, 4 cores.

 

I remain unconvinced by your central point. For one, clock speed is not a good metric for measuring performance between different CPUs. Secondly, you're comparing an overclocked processor of one model against a stock newer model and claiming that the latter is somehow inferior because it doesn't run at the same clock speed, a factor that, as I pointed out before, is not a good metric for measuring performance.

Another thing to keep in mind is that the i7 is the first multicore processor to introduce new power-saving technology that not only throttles clock speed, but also individual cores. Simultaneously, loss in performance versus power savings is a very small ratio due to Turbo Boost.This is a fact that you neglect to mention in your post.

Perhaps you just don't like the fact that the i7 takes control from your hands?

The design goal behind Turbo Boost isn't performance, it is tremendous power savings at a small cost of performance. This feature is indeed pointless with respect to playing games and all that nonsense, but it is very valuable when you're not playing games, which is almost all the time.

 

I completely get it, read my previous post. I wouldn't say thats a fair comparison, since the 920 is the first generation and your QX9300 was the last generation. Give em some time, these C2Qs have been out for 2 years, i7 has been out for maybe a year, if that. Also, clock for clock i7 is faster, these are not Pentium 4s.

You can't just add up all the cores and say " I HAVE 6467384974590763463GHZ OMGOZZZZ" because each core is set at a certain speed, each core is say 2Ghz, they cannot pile on top of each other for added power.

 

tell me that you are not really adding up the cores like that... please tell me i misunderstood. i re-read it and everything...

 

4 physical cores and 4 virtual cores. all of them running at 60-80+%. id say that would make a difference.

 

The only thing I dont like about the i7 is the throttling. When you are doing a single threaded app that makes it step to 2.8 and "shut off" the other 3 cores you are essentially using a single core HT chip. That is no good for me, if I buy a quad core CPU I want 4 cores all the time.

 

Even when the application you're using wouldn't take advantage of more than a single HT core?

 

Yes, because there are other processes then the one you are focusing on.

 

i was saying the i7's do not have 8 cores, they have 4 cores that are hyperthreaded. they built to support 8 threads, not 8 cores of processing power.

anyways, the i7 does not shut off the other cores. it overclocks them based upon usage and voltage.

 

At maximum clock speed it will only use one core though, I know it doesn't shut them off (why I used quotation marks).

 

as if the cpu wouldn't be intelligent and fast enough to then handle your other processes, in a singlecore or multicore solution, what ever fits it best to shift the workload.

it's not like it turns of the other cores and can't use them later again. that is a thing happening in submillisecond or millisecond regions (i don't know exact numbers). nothing you'd ever notice.

 

The Clarksdale's may not shut them off, I'm not sure. But I do know that the Arrandale's do have the clock gating, and will shut down unused cores, and the thing is, they shut them down and turn them on QUICKLY. That's the big thing... if you have other background processes that need a core, they won't tear down your OC'd core with the single thread... they'll just minimally power up the single core that they need for whatever housekeeping is going on, and then go back to sleep. Besides, when your single thread is blasting, there's still a second hyperthread core (also overclocked) that can handle some of that housekeeping stuff.

 

I am just for comparison of total available processing power when in a worst case scenario (full load on all 4 cores) this worse case scenario is not hard to run into.

Just encoding a video with the x264 codec will cause all 4 cores to max out for the duration of the encoding (can be several hours) this encoding is incredibly linear. From the same series of cpu's (core2duo vs core2quad) it takes exactly have the time for the quad core to do the job if it has the same clock speed.

So its quite ok to say I have 8ghz of power for a stock Q9000 and 4ghz of power for a 2ghz dual core as the encoding is carried out by that much cpu power. This scenario is only when using all 4 cores 100% of course.

So now I am saying I have 2.7ghz x 4 cores of power on the Q9000 to do the job and only 1.6ghz x 4 for the i7.

Yes I know you can not compare the clock speed of one generation of cpu to another directly and I said this many times already but I am saying that even with the better clock per clock performance the i7 has that it cant overcome the huge clock difference the old core2quad has.

I already tested it. I have the wprime benchmarks left over from the W90 when I was reviewing and testing it. A very linear cpu benchmark the best score I got was just under 14 seconds with 4 threads. With only 4 threads the i7 only gets 21 seconds at stock speed.

If you feed it 8 threads (kinda like cheating for a comparison, its like using phisix for 3dmark) it can get all the way down to 15.5 seconds. But even with that advantage of hyperthreading it can not beat the processing power from the old core2quad when its overclocked.

If you say its not fair to compare an overclocked cpu to a stock one, then you missed the entire point of this thread.... because that is what I am saying is the problem. That you CAN overclock the old cpu's for MORE performance for LESS money and the i7 has no place for an enthusiast that likes to overclock and would so the new i7 is actually downgrade.

 

it would only be a downgrade if you would 100% of the time be able to fill all your cores with work. which you can't. at the moment you can't, it's an upgrade.

 

do note that the i7 is hyperthreaded, so it would be 1.6 x 4 x 2 for the i7... in terms of how you're calculating performance.

 

While hyperthreading is great and all it is not identical to having 8 cores.

 

1.6*4 + .2*1.6*4 then. Hyperthreading is about 20% as fast as a "real" core in H.264 encoding, so... 7.68. Compared to a 2.7*4 = 10.8. But then again, you start comparing different core architectures, and things aren't completely equal. Like comparing bananas and plantains.

 

Anyone who overclocks on a laptop is mad. Most parts are not user replaceable, and the chassis is designed around the thermal characteristics of the installed and supported cpu and gpu.

On a desktop, you can upgrade the cooling, the case and fans. With a laptop it is a closed system essentially. Sure you can upgrade the CPU, and in some cases the gpu. However the case is not designed ( unless the upgrade is to something the OEM sells as a configuration ) to cool these parts! If you fry your GPU for example, you are pretty much boned ( unless it is a rare mxm II type ) I know you can add extra cooling pads, cooling fans etc. But then that defeats the object of being a laptop and portable! Not only that but overclocking will decrease battery life to a point of almost non existence and you end up relying on having it plugged in all the time.

My advice is if you overclock, get a desktop and a fast as you can afford laptop for portability. BTW I do overclock on my desktop, but I spent a ton on fans, cpu coolers and casing to enable it to be practically silent but cool running.

 

no.

look around you. this entire forum is full of people who are doing it and doing it well.

 

I know on the new CW models from Sony, upon a new windows 7 installation, intel requires you to install the turbo-boost driver. What would happen if that driver wasn't installed. Sorry if this has already been asked, but I just flipped through and didn't see anything about it.

 

I like the way you essentially omitted his entire argument.

 

actually it does. wether you call it turning it off, entering a low power state, or parking the cpu, or shutting down the cores - they kinda mean the same thing, all a matter of semantics. and have been discussed a lot in many thread/posts.
but thats kinda going way off topic, just thought you should now.

 

MSI and Asus have OC features on their gaming laptops. As long as you know what you are doing, there's no "major" problem.

 

You've missed the point. Such laptops are already designed with manufacturer OCing in mind. The vast majority of laptops don't have such acommodations.

 

Until the i7s are overclcokable in laptops, we cannot know the answer to this question. In my desktop, I am clocked at 2.8 Ghz, it OCs to 3.1Ghz on full 4 core load, between 3.4-3.5Ghz on 2 cores and I believe I have see it almost at 4Ghz on one core load situations, which are more common then you would think.

By only overclocking by 130Mhz, on 4 core load I am actually at 3.1Ghz. As I said these comparisons are not fair as you cannot, that of know of, OC an i7 notebook. It was not until the later generation Core2 and Quad series when Ocing became more prominant, give it some time these are the first i7 chips.

 

I don't see the logic in this at all.

Have people been burned so much by how "Bad" Pentium 4's Hyperthreading was that the Nehalem's Hyperthreading has to be compared in the absolute worst light possible?

It would only be cheating if the Core 2 Quad actually gained from using 8 threads, but you'll find out(surprisingly to YOU of course) that it won't be faster than when feeding 4 threads.

However, on the Core i7 you can feed 8 threads and make it faster, and you won't pay for the price of 8 cores nor the power usage.

Both Arrandale and Clarkdale can shut cores off(or power gating if you prefer to call it that instead). Whoever said otherwise is a crock. The whole reason Turbo Boost exists relies on the fact it can turn off cores.

 

As long as you can keep the temps in control, it doesn't matter what "accommodations" (or lack of) the manufacturer has in place.

 

If your primary application is single-threaded, you are better off with 2.8 GHz of power on one core and the other three off (though I don't think they go all the way off on the Intels), as it will go much quicker than with 4 cores at 2.0 GHz (three of which it can't use). Background processes really don't take that much power. I tested a very CPU-heavy game on my laptop with both 2.2 GHz dual-core and 2.4 GHz single-core (Dynamic Acceleration, one core disabled in the BIOS) configurations. It did better in the single-core setup. The difference wasn't quite the 9% clockspeed difference, but it was faster. If you've really got that many background processes running, Turbo Boost will switch to the two-core mode, still get your application done quicker than without Turbo Boosting, and have another core for all your background tasks.

And the great thing about HyperThreading is that even with only one core, you can have two tasks running. Thus if your primary app has to wait for data from disk or something of that sort, your background tasks can get some work done. True, it wasn't that great in the Pentium 4, but it's better than nothing, and doesn't take much silicon to implement (Intel estimated 20-25% performance boost for 5% additional silicon cost in the Pentium 4 - whereas doubling the cores roughly doubles the silicon cost). If you add more threads per CPU, as IBM and Sun have done, you get more benefit - sometimes much more.

On the original topic, it sounds like there's an issue with not being able to disable Turbo Boost. I see the problem, as clearly you can't have your CPU suddenly jump to 5.25 GHz. But I'm surprised there's no option in the BIOS to disable Turbo Boost - I know my BIOS allows disabling of Dynamic Acceleration (but then it doesn't allow overclocking the FSB, either). I'm also curious if software such as CPUGenie allows or will at some point allow disabling Turbo Boost. RightMark CPU Clock utility lets you disable Turbo Boost (or Super Low Frequency Mode), but doesn't work on i5/i7. If CPUGenie allowed that, then overclocking all four cores could safely be done once Windows was started (not before, as then your CPU might spike to 5.25 GHz while booting).

Sorry if this has already been said; I skipped pages two and three.

 

Cheers Bog! ajreynol just because people do, does not mean it is wise. I never said you could not do it. You can of course overclock, but you failed to address the section about reduced battery life / non user replaceable parts.

Exactly, there are only a few high end configs that support overclocking. There is a lot of stress and heat put on the system, and without proper cooling ( most laptops have hardly any ) it will burn up your system and kill parts and battery life.

 

No, you can overclock anything. people overclock netbooks (do you think they were designed for it).

My toshiba business notebook, it wasn't designed for overclocking, but there is no problem running a 30% OC.

You just need to find the limit of the system.

About battery life, ever heard of intel speedstep?

 

You are still not understanding. Where did I say you could not overclock? Exactly I did not say it all! Also don't patronize me about speedstep! When running games ( which is the main reason for overclocking ) your cpu is under full load most of the time, rendering speedstep useless. You do not need to overclock for desktop performance unless you are doing encoding or 3d work, which then takes you back to battery performance...I also will repeat for the final time, parts not not so easily replaceable in laptops as a desktop. If you burn out the gpu or motherboard you are going to in most cases need a whole new laptop.. With a desktop, you replace the part that is broke, which is much cheaper and cost effective.

 

1. Who plays games on battery?

2. This thread is about "enthusiasts" who should know how to get a new motherboard (BTW, $200-$300 will get you a motherboard on ebay for pretty much any notebook).

And I have never had a part fail on me from overclocking, so I say it's unlikely and worth the risk.

 

1) people that have a laptop. all others bought the wrong gaming machine from a cost/performance standpoint but them being enthusiasts, they can't agree on it for bragging reasons

 

Well still, battery life isn't an argument, otherwise you should never buy fast CPUs because you will have bad battery life, it has nothing to do with overclocking...

Also, you shouldn't game on battery power because it causes your battery to drain fast, which ruins the battery (fast charge/discharge is not good).

 

maybe you want a fast PORTABLE system, which is what notebook cpus get designed for? as said, anyone else could get a desktop for half the price, having higher performance.

if i want a high performance notebook, it should still last quite good on battery. an i7 delivers much more bang for the buck, then. fast performance and good batterylife, at least compared to an ordinary quadcore.

and no, fast draining a battery doesn't really ruin it, if you buy something built with quality in mind.

 

http://www.batteryuniversity.com/partone-15.htm