Why the Core i7 may not be a good enthusiast processor

Enthusiast – A person who likes to gain a higher than normal knowledge about a product. A person who
likes to tinker with, experiment with, and to extract the maximum amount of performance/ value out of
equipment.

Given that explanation of a enthusiast you must wonder why I would make a bold claim as to why the
latest and greatest cpu offering from intel may not be a good choice.

Well I do have a reason at least I have evidence that supports a strong possibility of potential problems
the i7 may present to enthusiast like myself.

First a brief history of a common practice an enthusiast would always take advantage of with his
computer, Overclocking. Overclocking is single handily the greatest method to increase the
performance and value of you computer for free. Knowing how to pick the right parts for a machine and
then overclocking them can save you tons of money and increase the overall performance of your
system by great amounts.

Of all the computer components the one that offers you the greatest gains as far as saving money is
concerned is the CPU. CPU manufactures produce basically 1 model of cpu at a time and within that
model is a few different revisions and variations. I’ll use the core2duo for example.

If the model is the core2duo they created a base for it say 3.2ghz is the top model. Every core2duo be it
2.0ghz, 2.4ghz, 2.8ghz, or 3.2ghz is exactly the same and capable of doing the exact same thing (given
the minor variations you may find like cache size) there are only two big differences between each
version of this one model. First and foremost the price. Usually the lowest model is about $200 and the
highest model (usually an extreme model standing for extremely expensive) goes for $1000. The second
major variation is obviously the speed of the cpu, that you receive it at…. 2.0ghz is not 3.2ghz but that’s
because of a setting of the cpu the multiplier. Each level of the cpu is changed by +/- 1 on the multiplier.
In the example of the cpus listed above say the system FSB was 400mhz the 2gz cpu has a internal
multiplier of 5x so it takes 400mhzx5 = 2000mhz aka 2.ghz. The 2.4ghz cpu is 400mhzx6 = 2.4ghz.

So in theory if you could magically change that multiplier on the cpu the 2.0ghz cpu is exactly identical to
the 3.2ghz cpu, it just cost $800 less. This is basically true. There is a quality control process called
binning where they test the max limits of each cpu, they always test it for the limits of the top cpu
model first so they will test it for 3.2ghz and if it passes it can become a 3.2ghz cpu. Its not very
common for a cpu to not make the cut for the highest model but it can happen so you may find you
don’t have a magic cpu that can overclock to the same level as the highest offering, this is extremely
rare. What is more common is they have too many of the top model cpu’s and they need to sell cpus to
all market segments, from rich to poor so what they do is once they make the quota for the top model
cpu they have to start marking the same cpu’s as the lower level cpu’s and just change that internal
multiplier so that it runs slower (and maybe lower cache size or something very minor).

Yes it’s a cruel game making people pay way more money for the same hardware, the cost the cpu is the
same for the company to make no matter what level they sell it as, but that’s the game.

So this is where enthusiast started to play the game back. I said it was possible in theory to transform a
cheaper cpu into a more expensive one if you could just change that multiplier, well unfortunately we
cant (in most cases ) but there were two parts to that formula for cpu speed the FSB and the CPU
multiplier. It only makes sense if you cant change one that you should try to change the other one.

This is exactly what we can do, and have been doing for years. When you overclock a cpu you just
overclock the systems front side buss. If you managed to overclock the default 400mhz FSB to 600mhz
that same x5 $200 cpu will change from 400x5=2ghz to 600x5=3ghz. Poof! Its magic you just made a
$200 cpu into a $600 cpu. Often times in many cases even the cheapest cpu will overclock BEYOND the
stock limits of the most expensive cpu model on the market. The cpu manufacturer does not crank out
products at there max limit, they make them much lower so that they can later release a faster version
for more money.

Now for a real life example related to notebooks since this is NBR and this is the topic at hand. The still
very great Core2Quad was the latest and greatest cpu until just recently with the i7 release. The Q9000
ran at 2.0ghz stock and cost about $200 MSRP, the most powerful and expensive quadcore was the
QX9300 that ran at 2.53ghz and ran for about $1000 MSRP.

Using a simple free tool that let me overclock the system FSB from inside windows my humble Q9000
would overclock to 2.7ghz easy and was totally stable. For $200 I beat out the $1000 cpu. Whats even
more impressive is that to this day I think that my limit on that overclock was actually the RAM and not
the cpu since increasing the FSB also boosts your RAM speed. This follows logic since owners of the
same laptop that sprung the money for the higher model cpu’s only got a very slightly higher overclock
and it was when the RAM got to about the same speed as mine.

Now flash forward to now, here is the i7 and the bottom model cpu is now the 720QM running at 1.6ghz
and boasting the “awesome” new turbo boost technology. Turbo boost seems like a revelation in cpu
technology and for some people it may be. I think those people are the standard users that never even
heard the word overclocking. I do believe the core reason for this article is this new technology. I think
while it came in the form of a blessing this turbo boost is actually a curse in disguise.

So what is turbo boost? Its intels answer to boost performance on your computer for applications that
are not optimized for a quad core cpu. If you have a demanding program that can only use one core of
the cpu instead of being limited to only 1.6ghz and having the other 3 -1.6ghz cores go to waste. In real
time the cpu will overclock that single core for by almost 70%.

The 720QM normal operation is 131FSBx12 = 1600mh aka 1.6ghz however when turbo boost kicks in it
can raise that multiplier to as high as x21 bringing the core speed all the way up to 2.8ghz.
So it is indeed true if you were paying some old game that ran on just one thread and 1.6ghz was not
enough it can turbo boost all the way up to 2.8ghz and handle the job when the old core2quad series
would not have been able to do it. This is what intel wants you to see, this is what the normal user
would see. I however as an enthusiast see the huge problem here.

This only happens when you have less than all 4 cores under high load. The reality of the fact is that
anything these days that is highly cpu dependent is multi threaded and will use all 4 cores. So the
technology in itself is not as valuable as it seems. Any of those old programs and games that only use 1
thread were running just fine on a Pentium 4 with just 1 core. Todays cpus have advanced so far that
just 1 core on a 2.0ghz Q9000 is just as fast or faster than a 3.4ghz Pentium. The same evolution took
place between the core2quad and the i7. The stock 1.6ghz of the i7 is very close to the power of the
2ghz core2quad.

So whats the problem? The problem is what we started talking about. I didn’t run my Q9000 @ 2ghz I
ran it at 2.7ghz all the time, literally. One reason I love quad core cpus is just how much power you get
out of each little increase in core speed. A core2quad @ 2ghz = 8ghz of total processing power for stuff
that is multi threaded. You would need a core2duo @ 4ghz to get the same performance. This means
overclocking has 2x the effect on the quad as it does the dual.

Now take a 2.7ghz Q9000 and face it up against its new replacement the 720QM 1.6ghzx4 = 6.4ghz its
just not happening. I ran the benchmarks and the i7 did remarkably well for its clock speed it got pretty
close the the Q9000 performance numbers but not close enough its still behind it so last generations
product is able to run faster and cost you less than this generation.
The solution is quite simple, just like always once you overclock the new cpu it will go above and beyond
the performance of the old one.

I was super confident before my purchase that when I saw turbo boost of 2.8ghz that this cpu would and
could do 3gz+ on all 4 cores super easy. Infact I know it can but now we can really discuss the problem
with all of that explaining done.

When you overclock you cant change the cpu multi only the FSB. If I wanted to overclock my new i7
from 1.6ghz to say 2.2ghz where it should be a good match for my old Q9000 I would have to increase
the FSB from 161 to 200. Not a big raise on the FSB itself but the problem is the multiplier we could not
change is still there and turbo boost is still there. Sure 200x12=2.2ghz but the multiplier shoots up to
x21 every time you load anything turbo boost always kicks in and seems to be built into the cpu itself.
So now that small increase in speed so that I can match my old Q9000 for encoding a video on all 4 cores
is going to result in this happening. 200x21=4.2ghz this is… too high chances are even this small
overclock to bring the i7 to the same level as the Q9000 was is going to result in a blue screen and
system crash. You can totally forget about a 3ghz overclock like I had planned on. 250mzh fsb x 12
default multi = 3ghz. When turbo boost raises the multi to 21 250x21=5.25ghz. No way is the machine
going to handle that without a instant crash. Also the memory is getting overclocked when you raise the
FSB also and since the i7 uses a lower FSB than the Core series did that means the ram uses a higher
multiplier. That means each small change to the FSB is going to effect the RAM more. So I hope we find
a way to disable turbo boost for the cpu to get the OC a bit higher, but then you still have the RAM that
may cause us to fall short and so far there is almost never any way around that.

So I am not saying the i7 is a bad cpu, infact I think its great for average user who never knew about
overclocking it will offer more performance and do great things, but for the rest of us it may not be so
great. Did Intel finally catch on that all of us were buying the cheap cpu’s instead of the expensive ones?
was this maybe some undercover way to end the trend of overclocking for performance gains and to
save money? I don’t know but no matter how you look at it this dilemma is definitely some food for
thought and could have a strong impact on your decision of what cpu to buy.

I have real life benchmarks soon to come of my old W90 with the Q9000 vs the new G3 with the 720QM. Forgive the formatting MS word does not paste into the forums very well.

 

If you can break your opinion to small paragraphs, it might easy to read.

BTW personally i believe 920XE is not a perfect enthusiast processor.

 

I will fix it later but i pasted it from word and it messed it up bad, it was worse than it is now before my quick fix.

 

I don't think you will get the BSOD when you overclock.

It's like with IDA, my CPU can't hande 2.8ghz with both cores active, but with IDA it is fine at 3ghz.
But maybe turbo is different, I don't have an i7 to test

I'm sure that there will be software soon that can disable turbo for you

But right now I agree that it's worth getting a Q9100 (and OC to 4ghz with volt mod) over the i7.

 

i'm an enthusiast, i want to know exactly how the cpu works in every detail, to know every feature and it's real gains in my usage, etc.

which is why i love the i7.

a modern enthusiast is about knowing how the computer does it's work best for the user. and the enthusiast knows the i7 handles all very well, so manual fiddling is unneeded.

but for a modern enthusiast, that's a big +. as now, he can fiddle on more important things than "getting it going". like optimizing his home network. or home automation, or anything fun new thing to explore.

i'm happy that i've outgrown the all-time-the-same tons of fiddling and tons of tweaking ages, and so did the hw i bought and will buy.

am i not an enthusiast now? i guess not, just come visit me to see how much work i put into these beasts, even without manual fiddling

 

Turbo Boost is essentially overclocking for the masses so yes, it doesn't really get along with enthusiasts (can't you disable it in the BIOS? I know you can in desktops). That said, a lot of the blame here needs to be laid not on Turbo Boost, but in the 720QM (and to some extent on all of the Clarksfield processors).

The fundamental problem is that Intel simply wasn't able to do quad-core properly in laptops using the 45nm process -- they're too hot, too expensive and/or too wimpy. The 720QM's 1.6GHz stock frequency means that it is on approximately the same level as dual-core CPUs like the 620M even in a fully multi-threaded scenario. It uses Turbo Boost to claw its way back to relevance in a scenario with one or two threads.

In a more reasonable scenario (e.g. Lynnfield), Turbo Boost is still unfriendly to overclockers, but it becomes more and more necessary as the number of cores grows larger. Sure, you can overclock all cores, but the power consumption and hence the heat are proportional to the frequency (and if you have to change the voltage, that goes factors in quadratically). It makes a lot more sense to factor in the number of cores necessary. AMD is doing it too: their 6 core processors due out this spring will feature essentially the same thing.

 

I think they can do quad core in laptops no problem. If I can run a Q9650 in my 15.4 laptop and even OC it to 3.35ghz stably in it then there is definitely not a problem with putting quad's in laptops. As for OCing I think enthusiasts find it fun I know I do and once you know what it can do it is hardly fiddly or time consuming. Takes seconds to set it to whatever OC you want. Or even set it to do so at startup. Another thing presently cost really is a big factor to. The 720QM is more expensive than the Q9650 I run and I am pretty sure it isn't better. The 920XM yes but then most of us haven't got that much cash to invest in it.

 

tl;dr

It seems to me many enthusiasts on this forum actually use the Core i7. Good luck convincing them that the i7s are no good enthusiast processors.

 

I was debating whether to put a qualifier on "laptops" in that context and decided that people would understand, but I guess not. Yes, I know it is possible to stuff desktop quad-cores into laptops and various manufacturers have been doing it for a while. I meant portable laptops (and since that is also liable to misinterpretation, I will clarify in advance: by "portable", I mean relatively light machines with a long battery life). The C90 is actually a bit of an exception -- most desktop/laptop hybrids are much larger and heavier -- but the issue of battery life remains.

 

As far as I can see the only problem is there's no readily available way to overclock the Core-i notebooks (outside of the M17x r2)....but a big part of being an Enthusiast is finding a way to make overclocking possible.

Desktops have proven the Core-i tech is an absolute beast for overclocking when you've got the tools to do so. Turbo Boost doesn't get in the way of this as it can be disabled, and usually is when overclocking on desktops.

 

The point of Turbo Boost is to use all the available TDP headroom to clock the CPU as high as possible depending on the task at hand. I think it is a fantastic new feature, like an extreme version of IDA for the Core 2 Duos.

 

Yup. I'm not sure how the Core i7's work, but I believe that they should be able to keep clocked up as long as the thermal head isn't reached. I'd like to see if it could keep the clocks higher on all cores as long as the cooling/power limitations were respected.

 

If you want to disable the turbo mode you can just download
Realtemp and start up the i7turbo program and disable the turbo mode in there

 

Yeah but it does not, any benchmark I use that puts all 4 cores under load (wprime, occt, video encoding) they go to the default 1.6ghz right away. Before it settled at 1.6ghz it spikes to about 2.8ghz on one or two of the cores as the process is loading as it hits just the first cores first before all 4 are fully inintated.

To answer the other question, its not about the thermal limits even @ 1.6ghz on all 4 cores its far from hitting a thermal limit but it still wont overclock the cores unless its only using a few cores on the cpu. Its like a hybrid dual core/quad core cpu but to gain that hybrid ability you lost the ability to make this a better quad core and have to fork out the $$$$ for a higher model, and honestly even the highest model is not that fast.

To answer the last question, no there is no way to turn the turbo boost off, not on the G73 anyways. Once we find a way to do this I think there will be some serious potential in this cpu by means of normal FSB overclocking but for now I think only a very small overclock will be possible before a crash due to the insanely high cpu multi turbo boost uses when you load a task.

Some of you may not have gotten the way I was trying to present this information, I am not trying to persuade anybody that the old core series is better than the new i7 cpu, I just wanted to bring up this interesting information I discovered for discussion so we can debate what is good/bad and make what otherwise would be a very scarcely known problem more well known. I did choose my words wisley, even the topic says "may not be" informing that its just a opinion and a discussion not a documentary

As it stands right now my old Q9000 is beating my new 720QM in performance and that is simply a fact, and one more thing that really bothers me that I did not have time to mention in the article is that the 720QM only gets close to the performance of my Q9000 when it has not 4 things to render but 8... It takes all 8 threads loaded using the hyperthreading before the performance is similar to that of the Q9000, with only a 4 thread load its very far behind.

 

Worse comes to worst there's always the 45nm AMD Danube's "Champlain" x4 due out this year.

 

Well if ur willing to pay and don't mind the weight , why not buy a notebook with core i7 desktop processors? It might weigh more but certainly has way more performance the core i7 mobile... and if u want an unlocked extreme cpu , the i7-975 would toast the 920XM.... Anyways , u can overclock the i7-975 in the Clevo D900F right?

 

Well I have just trying the orthos tool and it runs only on 2 cores and my core i7 jumps up to 2.66 ghz and when I start up thr real temt and disable the turbo mode my core i7 goes down to 1.6 ghz

 

Is there a possibility that we can get a software like setfsb to overclock the i7 one day? i7 doesn't have fsb so it will have to be a new program. Getting stuck at 1.6 on 4 cores is SAD.

 

One day perhaps, although without the bios being able to turn turbo mode off, visc is right, as soon as you have a solid overclock among the 4 cores, the instant it decides to go, right im done here, oh look hes on youtube, i'l OC this one core to make it nice and smooth, BAM, your going to hit a BSOD,

And i dont think its too nice for you ram either,

I can vouch for everything visc has really put forward here as i have actually managed to blow a bank of ram due to Pushing the FSB limit fairly hard and then IDA deciding the lock in an extra 1x on the multiplier (So i was running 2.9 and then BSOD.)

Had beeping confirming values were set too high (And i'd setfsb'd many times before) but it turned out something had messed up, and one of the banks actually said goodbye.

 

You must realize the Q9000 was a late Core2 model, it was the last generation before Core i3/5/7 was introduced so naturally they had a very effecient proccess in which a higher clocked quad core with low power consumption could included in a mobile package.

This is just the first generation of i7 and I believe there is much more to squeeze out them. In true multi-threaded apps, the i7 will win, and Intel proved clock speed is not everything.

I agree with your general idea though, Turbo Boost was a stock, and effective way, to utilize the whole processor for the general public. Overclocking takes a lot of patience and work, which most of the public would not be able to commit to.

Again, these are just the first generation of i7, and I believe near the end of their lifespan before a new architecture is released this same question will be asked.

 

Do you mean these?

http://www.fudzilla.com/content/view/16912/35/

If so, then they are going to be like Clarksfield, except most likely without Turbo Boost (unless AMD uses the same technology as for the Thubans). Looks like AMD can't make a decent mobile 45nm quad-core any better than Intel. I think we'll have to wait for 32nm quads before this makes sense.

 

Your right about the Q9000 being end of line and the new i7 being new but its still a much better cpu and can do a lot more, intel has set it up in a way that we just cant extract that potential. Temps do not go up very much from just regular overclocking and the cooling in the G73 is absolutely great just like the W90 was if it can do 2.8ghz on 2 cores it can do 2.8ghz on all 4 cores without a problem if you could just configure it that way.

 

What if the problem isn't the temps, but in the PSU to the chip? I'm not sure about the circuitry, but maybe the power distribution within the chip isn't up to the challenge, especially considering how different the desktop and the laptop sockets are (989 pins in the laptop vs 1156 in the desktop)? I mean, you're saying "if it can do 2 cores at 2.8GHz, it can do 4 cores!" and it seems like you're completely ignoring that is more than doubling the amount of power needed due to the extra interconnect circuitry and such.

The i7 is still a pretty power-hungry chip when it's going flat-out, 45W peak from what Intel claims. The lowest-power Lynnfields are 82W TDP, that's less than double the Clarksfield. That's a LOT of power in a CPU for a "mobile" machine.

 

Maybe, maybe not. I know up till now no other cpu had problems overclocking really high and I have yet to see a quad core really gain a lot of temp/power drain just from clock increasing its only when raising the voltage that it really starts to move the needle.

So my gut and experience says that its not a problem, but I have no technical data to back up that claim with proof.

 

I'm confused as to why some enthusiasts would dislike the i7's automated overclocking abilities. What is there to dislike about this feature, from the enthusiasts' view?

 

I think its the "do it yourself" aspect, something to be proud of? Obtaining the higest clock is somewhat of a sport I guess?

 

If I received a CPU that dynamically adjusted performance in an optimized manner in order to save power and produce less heat, I would be pretty pleased. If you think about it further, as a consumer we should expect a product to perform at its best, and any time that the enthusiast spends lovingly tweaking his system for hours on end is wasted time that should be grudged.

 

It's because 4 cores at 1.6ghz isn't gonna cut it for us. I want to push the i7 to 2.0 ghz at 4 cores then I'm happy. The turbo boost only kicks in when not all 4 cores are used which brings you to an average dual core processor speed.

 

Yeah those are them. The X920 BE would essentially be a QX9300 level chip with the on-die memory controller of the i7 quads. Of course no one knows if it will reach clocks as high as the QX9300 did but at least without TurboBoost we'll be able to find out. If the X920E can (and you can find it in a powerhouse notebook) then by the rationale of this thread it would be the best Enthusiast CPU.

The problem is that CPUs are not optimized....they're built under a specific threshold deemed as safe by the manufacturer that's always lower than the CPU's real potential. "The Enthusiast" seeks to max out that potential to it's limits.

 

The 720QM was never really aimed at being an enthusiast processor anyway, though. If you want that, wouldn't you be looking at a i7-920XM, which is clocked at 2GHz base? Or the 820QM at 1.73GHz, which can use a faster turbo during quad-core operation?

 

Thats the point enthusiast like me get the lowest one and save $$$ the Q9000 was not an enthusiast cpu nor was my desktop Q6600 that does 3.6ghz instead of 2.4ghz all day long.

enthusiast doesnt mean the guy willing to spend the most money to get the best stuff, it to me means the guy who knows how to get the best for less.

Because its not really overcloking the cpu read the entire first post I think I explained it really well.

the new i7 stuck at 1.6gh x 4 = 6.4ghz of power for a demanding task. My old Q9000 @ 2.7ghz = 10.8ghz of power... see how vast the difference is just because I can overclock the cpu?

Thats a HUGE difference in power.

The i7 only overclocks the core when part of the cpu is used not all 4 cores and its when all 4 cores are in use that you need the cpu power the most. Just overclocking one or two cores is almost completely pointless in todays computing world, there are no applications or games that need it. Plus the hyperthreading results show me that I need 8 threads to compete against the Q9000 with only 4 threads... So that means I need 2x more threads per core for the cpu to be at max efficentcy.

If turbo boost is on that means you do not have many threads its only using 1 thread per core and thus even slower than the old core series at that point.

 

The problem is that the 720 isn't going to do much, and you know that. There is no FSB to overclock, no way to modify it. Sure it's not "enthusiast" friendly, but they make one that's specifically for overclocking rather than dealing with so many kiddies OCing their machines past breaking point and then returning them. I've seen too many instances of someone running something out of spec, breaking it, and then exchanging it leaving Intel or a reseller stuck holding the bag. I can see why they wouldn't make a inexpensive modifiable CPU... it costs too much because of the people out there who don't take responsibility for their own risks and mistakes.

GHz is not a unit of computing power. Did nobody learn anything from the Pentium 4?

 

@ Pita 2.7 Q9000 still kicks the 720qm in the butt through

How is the desktop i7 being overclocked? Does it have fsb?

 

TLDR, but it seemed very interesting! ><

 

I have to agree with this. Rather than pushing a CPU to its limits for minimal performance differences, I've got a CPU that does the heavy lifting for me out of the box.
I'd already outgrown the whole overclocking thing, my E8400 ran at stock its entire lifetime. Now I can focus on more important things, and when a problem does crop up, I can get a better understanding of why it happened, and fix it too, rather than wondering if my system was unstable from modified voltages, clocks, and SPD tables. Overclocking is a fun hobby, but it's a far from important quality for an enthusiast, depending on what you're an enthusiast of.
I can still do lots of learning and understanding without spending hours tweaking BIOS settings and then running stress tests.

 

Even in single-core loads? There are a lot more of those than you'd think. And the 720QM is just barely priced above the Q9000. Plus, the i7 uses a lot less power when not actually ramped up. I just don't think that it's a big deal... it's like everyone expects the (lowest quad core) 720QM to be a massive increase over the previous generation in raw horsepower... it adds more features more than raw speed. If you want raw speed, go for the bigger 920XM and such.

 

I think what the enthusiasts expect is marginal improvement over the q9000 but keep the ability to overclock.

 

I know this, but thats a several generation old cpu the i7 and core are just one generation apart. I have stated more than once that I know the i7 is slightly more powerful clock for clock but not enough to make up 4ghz worth.

I have never seen somebody break a machine from a fsb overclock, only overvolting and thats not something just anybody can do. At worst your system freezes and you get a blue screen and have to reboot.

I would like to see an instance of these single threaded application and loads that you think a single core on a core2quad cant already handle with room to spare. Even if you did find one for every one you could present to me I can find 5 situations where all 4 cores are maxed out on the i7 and the old core2quad series could leave it in the dust. To trade off all 4 core overclocking for single core auto overclocking is like trading out your hand to get an extra finger, sure you have an extra finger now but nothing to use it with.


Really it comes down to encoding for me. I noticed even 2.0ghz on the Q9000 maxed out every single/dual threaded game fine. When overclocked there was almost no gains. Gaming is the most demanding single/dual threaded thing I ever run into. All my other programs that are cpu intensive are multi threaded especially the most intensive of them all when I do audio or video encoding. It always maxes out all 4 cores @ 100% for the entire duration of the process. What used to take me 10 minutes to encode will now take me maybe 15 minutes. It it took an hour now it may take an hour and a half. Thats a huge loss of performance and convenience.

I really do want you to present some of these many situations that are only dual/single threaded that the i7 will beat the old core2quad as I do not see it. Even if I was stuck in that situation in real life I always could just overclock to compensate. I also have a theory that this new hyperthreading will make performance of single/dual threaded stuff even worse than it was before on the old core2quads.

 

I don't think the desktop i7 uses Turbo Boost, so therefore it's pretty much safe to overclock like the C2Q's.

 

you think wrong. i5 and i7 on desktop have great turboboost. one of the main reasons i got an i5 there. one of my apps is 100% singlethreaded for one of it's main tasks. dances around the old quadcore, that i5

 

I think the desktops let you turn it off in the bios for high end overclocking.

Same I just said last post I cant think of any single threaded application that I do not have room to spare on my desktop quad. 3.6ghz on my Q6600 is still a 3.6ghz single core and it will offload any other tasks to other threads so that the single threaded application can have full use of that one core.

 

I was pretty sure the 920 and upwards didn't have it. Seems I was wrong.

 

As the very verbose initial post said, it messes with their overclocking. You either have to disable it (which apparently you can't do on some laptops) or overclocking becomes a lot more tricky. And even if you can disable it, you will be wasting a lot of power and not using a part of the hardware. It's aesthetically displeasing: whereas before overclocking cost you nothing except your time and extra electricity (assuming you did it right), you now lose the independent overclocking of individual cores.

It's great for everyone who doesn't plan on overclocking though.

Indeed. I recently built a machine with a Core i5 750 and it is fairly awesome -- not only does it effectively run at 2.8GHz (rather than the nominal 2.66GHz) with all 4 cores loaded, but for single or dual threaded processes, it runs at 3.2GHz.

That said, if you want to overclock the desktop i7s and i5s, it's fairly straightforward: you still have a base speed and a multiplier except that the former is now called BCLK (Baseclock) instead of FSB. You can increase it in the BIOS just as you did before and if Turbo Mode is getting in your way, you can turn that off in the BIOS too.

They do, but to a much lesser extent. Whereas Lynnfield's (870 and below) Turbo Mode is many bins, Bloomfield's (920 and above) is only one or two (or maybe a bit more, but definitely much less than Lynnfield's).

 

Hey Vicious, I've been working on some parallel programming stuff, so I have a simple program (matrix multiplication) that can take advantage of as many threads as available. I just ran it on my 2.4GHz C2D and managed to get an execution time of 178.1s (single threaded) to 93.4 (two threads). I'd like to see how your Q9000 fares against the 720qm in a situation where scaling is this good. I can also make it set the number of threads based on an input (or I can give you multiple versions with it hard-coded if that's easier).

 

I absolutely agree with the OP, and would like to put something on about energy efficiency. With any quad these days - the battery isn't great; but does mobile i7, or even i5 do much better than its predecessors?

I should think not. While the clock-per-clock performance did increase noticeably, and the turbo boost is suppose increase performance, while not compromising energy efficiency - by loading extra power the most in need, the extra power helps to do the given amount of workload. That's assuming faster a task is done - the better. How does a CPU with fixed number of cores do that? it uses Turbo boost. Where at the cost of increased instantaneous power draw, given numbers of core run into overdrive. On paper, that's good and all.

However in practice, the powerful army is only strong as the weakest link. Commonly any modern CPU shouldn't be the weakest, we have to look at the factors like the hard drive to see the problem. The limiting factors mean, unless there is costly SSD present, the CPU could only process as much as it's requested of. That means often the i7 ran at higher frequency than required, in assumption it would finish workload quicker - yet often the tasks couldn't be finished quicker - it's fixed timeframe - more on that later.

Not to mention the workload often are continous - where as the older desktop C2D's norminal TDP had been higher, the first generation desktop i7 had many power tweaks, like the power-gate, and managed to reduce electricity consumption by a bit - the same doesn't carry over to mobile side; also the thermal package is more flexible on DESKTOP platform, with far greater headroom. Because of the genuine archetitchural improvements, and the fact C2Qs were chewing so much power already, desktop i7 actually saw improvement in energy efficiency, per clock.

Needless to say, Mobile C2D were functionaling in far more efficient manner already - stright shrinking process for the latest and the greatest isn't necessarily enough to beat the last gen performance champ, at least by huge margin. Additionally with generations of 'tock' refinements, mobile C2D were tweaked considerably for mobility power; considered the state of art in power management for quite a while. When they had to reduce those hot 130W i7 chips for laptops, the output took too much the form of desktop chip - with little extra energy saving features. For e.g. 920 XM has even more single-core TB than its desktop Lynnfield siebling. On desktop environment, TB is great - grass isn't as green for mobility market, it's both a bless and a curse. By allowing cores to be dynamically loaded higher, the CPU risks higher levels of power draw in even light-medium usage, than C2D.

That's essentially the curse the i7 mobile had - they just couldn't do higher frequency on 45nm, while staying within a new high of 45-55W thermal package. 720-QM is good and all - but its TB implementation could be superior without threatening 820; where as respectively 920XM is hotter, and expontentially more expensive. With i7, while it's offering faster performance, and higher performance vs energy efficiency, the battery could be run dry MUCH faster too!

Thus often, if your doing anything ranging from light to medium load work, i7 would have to operate on potentially higher nominal TDP, with possibility of TB drawing disportionate power for certain longer tasks over last gen Q9000, that aren't actually greatly CPU demanding.

Additionally the efficiency, or rather the lack of efficient programming means these CPU go into high-bin very often - and stay that way for longer than really needed. Not really within intel's control, but could really cut i7's battery down a bit without user knowing. With older C2D, while you didn't get as much performance, the peak power draw was lower too, in a sense i7 removes the battery-friendly limit C2D introduced; you'd have to be careful with i7, you do get sharper knife, but careful not to slice your battery too much.

Even the newer i5 - they are still not at the battery level of last P series C2D. While the peak performance increases, the power could also be drained much quicker.

What I am saying, is i7 isn't ready yet, Consider, with all the tweaks, QX9300 with 45W TDP could be OC'ed as high as 3.2Ghz in M17X - faster than the 3 Ghz QX6850, at fraction of latter's 130W TDP - current mobile i7 has far to go in both effeciency and feature side. For e.g. 32nm, hardware encryption support, lower TDP, lower energy consumption.

 

And this is why I chose an arrandale i7-620, as I mostly game. Higher clocks FOR ME > more cores

 

I sold my W90 with the Q9000 I only have the G73 with a 720QM now.

 

True, especially with power usage. The Arrandale refresh will fix that:

 

I'm running an i5-540M, 2.53GHz nominal. I'm loading it down with Handbrake right now, encoding some high-def video, and CPUz is showing that I'm clocking at 2.8GHz. Something with the Turbo is going on, and that's with all 4 threads loaded. I'm ok with a free overclock from what I paid for.

 

Eh well that doesnt hold up too well, when you are only using 2 cores the turbo boost kicks in for higher clocks.

The real loss here is in multithreaded applications aka using all 4 cores, and I think we all know that there is more and more of these every single day including games. Most games are just gpu bound and the cpu doesnt have to work hard, the games that really kill the cpu 96/100 of them are using all 4 cores on a quad and the dual core is going to perform worse.

Yeah but the i5 is only a dual core, if you had a quad core of the same speed your encoding time would be cut in half. My old W90 with the Q9000 would literally finish in half the time.

Id have to say if your ok with some mild auto overclock your not included in that little group of enthusiast I am reaching out too