Xps 16xx Question/clarksfield Vs Arrandale

1640/1645/1647.....what are the differences? Just processor? Also, any recommendation between clarksfield vs arrandale processors? and Sorry if this is a repeat question. Thanks in advance.

 

I guess thats 1640 = core2duo, 1645 = clarksfield, 1647 = arrandale and yup that's it.

Go with clarksfield if you need 4 cores for heavy multitasking for example, and go with arrandale if you want more battery life, less heat and yet satisfying performances (avoid i7-620, that sucker heats/consumes almost as much as clarksfields).

 

I've heard nothing but good things about the 620.
It's got a higher clock speed than the rest of the Clarksfield and Arrandale i5's and it has the i7 name

Clarksfield would be best if you do a lot of video editing or if you rely on photoshop a lot. But at the moment a lot of software can't utilize the 4 cores and it may not be worth buying it for future advances in software because dual core looks like it will be here for a lot longer. Clarksfield for gaming though? I'm not a big gamer but that's what I've heard.

 

Thanks everyone. I assumed it was just the processor but wanted to be sure.

 

Another Question. Do all 3 have backlit keyboards?

 

Yes...............

 

thank you sir

 

Ya agreed with above, I'd recommend a 1647 for general usage/gaming and a 1645 for video editing or anything that taxes CPUs. Externally, all the 3 16xx are the same, but the 1640 is laid out differently internally. Components can be configured to be the same except the processor.

 

Most of the reccomendations are based off opinions and not actual real world use.

You don't need a quadcore to do video edit and/or picture editing.

I run premiere pro cs5 64-bit and edit full 1080p HD footage on 2 monitors and also have photoshop cs5 64-bit open editing images. My i7 620 handles it in stride. Doesnt even hit full load.

So again, you don't need a quadcore aka clarksfield to do video editing. Unless you get a 920XM I suggest the 620 over any other core laptop chip.

My cpu runs at 3.2-3.36ghz constantly with 4 threads. Rather have that than 8 threads at 1.7ghz.

 

@tenknics

Thanks for the response. but doesnt the clarksfield also support the turbo mode so it is actually more than 1.7Ghz?

 

Yes, it goes to 2.8.

 

k thats what i thought

 

Umm I think you're missed what I said or maybe you don't know how turbo works..Let me re-explain..

I have 2 cores, producing less heat. The turbo mode on arrandale's is more aggressive as they are based on the 32nm update. My turbo can kick in with both cores, in this case all of my cores, at full load. So I have 4 threads with HT at the max speed of 3.2ghz-3.36ghz


Yes the clarksfield turbo goes to 2.8 but not while running all 4 cores and hence not with 8 threads. The 2.8 kicks in on one or 2 cores, and it goes down from there..


Do you understand my original statement now? I'd rather have 4 threads at 3.2ghz than 8 at 1.7ghz (again you're turbo doesnt activate when 4 cores/8 threads are in use)

I cant be any more clear..

Let me make it even more easy to compare.


When 2 of clarksfield's cores are running, the max speed is 2.8ghz
When 2 of arrandales cores are running, the max speed it 3.36ghz.

You will use 2 cores more often then 4, and even then you still have 4 threads. Ask yourself if you need 8 threads, 8 slow threads. Or is 4 enough for you? 4 fast threads...

 

^^^^Nothing like being subtle about your opinions^^^^

Show a little respect plz

 

Teknics I think urv got ur threads and cores a bit mixed up.
Check this for speeds on different amount of cores.

 

Yep i7-720QM runs at 2.8 on 1 core, 2.4 on 2 cores I think. Dual Arrandale reach higher speeds, true, but real tests show that in some cases high-end dual are faster than a low-end clarksfield and in other cases they're clearly slower (when you have the full use of 4 cores).

 

I suspect your comment on the i7-620 is based on the notebookcheck review. However, thier original review was wrong. They corrected it on the german site, but last I checked had not fixed the US site. Anyway, attached is the corrected graph. The 620 may not be worth the extra cash, but the heat/battery profile is well worth the extra performance.

 

Not quite that high in dual-core mode.

i7 720QM: 1600MHz base, 1733MHz quad-core turbo, 1733MHz tri-core turbo (essentially quad-core turbo but 1 core is unused), 2400MHz dual-core turbo, 2800MHz single-core turbo

i7 620M: 2667MHz base, 3066MHz dual-core turbo, 3333MHz single-core turbo (I may be slightly off due to rounding)

Edit: I agree with tenknics that the 620M is the better and faster processor in nearly every case. However, configurations with the 720QM from Dell are often cheaper than with the 620M (and according to Intel the 720QM is worth more), and the 720QM allows memory speeds of 1333MHz (as opposed to 1066MHz with the 620M, I believe). So there's a little more involved than just raw processor speed.

 

Talk about it!

Go get some or rub one out, and then come back tec.

 

No I don't. Clarksfield 4c/8t. Arrandale 2c/4t. Thanks tho!



I can't help that you guys take my blunt nature as disrespect. Don't take things so personally. Lotta misinformation running around on this board, if you don't like the facts then don't read them.

Not quite that high? I was off by 27mhz, I said 3360 and its 3333, woa soooooo sorry. And you're basing this off of intel's spec sheet. Ya it says single core turbo, but I constantly hit that turbo with both cores active. Like I said the turbo mode in arrandale/westmere 32nm update is way more aggressive than the turbo in nehalem/clarksfield

 

I'm not referring to a less than 133MHz difference. I trust Intel's specs.

 

So you own a 620m? Because I do. Currently running at 25x which is 3.33ghz on both cores, not 1 like intel's spec sheet said, but both cores. I trust real world performance and first hand experience. I'm typing this up while I encode a 1080p file in Premiere Pro CS5 in the background.

Intel's spec for turbo is a guideline. It's based on power consumption and core load. There are a lot of variables that play into turbo, its not as cut and dry as Intel's spec sheet makes it out to be, especially the 2nd gen turbo.

 

what i dont understand is how dell only lets you get what you want in certain configurations...like on the canadian site right now if you want a 1080p screen and an hd5730 you have to get the 720QM or 820QM. Any other config only allows for a 1600x900 screen res and has the 4670 and less Ram. Im hoping by just switching the 4670 out for the 5730 i wont have overheating and throttling problems but if i could get a cpu that runs cooler and uses less power with the same config thatd be great. I dont see the 720QM as being very practical or necessary.

 

No, I don't have a 620M. But I don't see why you are listing the performance as higher than expected for ALL 620M's in dual-core turbo mode when you only have one test case.

You also said dual-core turbo mode for the 720QM is 2.8GHz, and Intel says it's 2.4GHz. Are you sticking with that number too?

 

No it was based off the number someone said in this thread. I don't own a 720QM so I don't know real world performance. One thing I have noticed that on most all Core's with turbo, they highest turbo that is reserved for 1 core on intel's spec sheet usually works on 2. My friend has a 620 in his Hp and reports the same results as I do.

And Im not listing performance higher than expected. I'm listing real world results based on personal use and based on my comparisons with other owners, its not uncommon for the 620to hit its 5x turbo on both cores.

Like I said, Intel's spec sheet is a guideline. Turbo is based on power consumption and load, 2 variables that aren't constant and completly dependent on your workload. It is possible to hit a higher turbo on more than 1 cores. I don't care if you don't believe me. I see it everyday. Thanks for your input tho man!

 

Intel® Core? i7-720QM Processor (6M Cache, 1.60 GHz) with SPEC Code(s) SLBLY ....... it is 2.8 guys.

 

Nice work. still cant decide between the two though...tough decision

 

Yea thats the max turbo for one core. We were talking about 2 cores, so in that case the max turbo is 2.4. I wouldnt be surprised if both cores could hit the 2.8 turbo though, like with the 620.

Here it is completely explained:

620m has 3/5. Meaning 3x133 for 2 cores = 2.66ghz + 399mhz = 3.06ghz and 5x133 for 1 core = 2.66 + 665mhz = 3.33ghz


And like I've said before, turbo is based on load and power consumption. Right now both my cores are running at 3.2ghz, the speed between the listed official Intel turbos. The 3.06ghz turbo is a 23x multi and the 3.33ghz is a 25x multi...My cores running at 3.2ghz is using a 24x multi. How is this possible gpig? Its not on your beloved Intel spec sheet. It's a half-step turbo, something that happens all the time with the new turbo modes on Intel chips. Its not always perfectly one turbo or the other, there are steps between.. Something that isn't listed on intel's spec sheet..But I guess because it's not, it's not real right gpig?

 

Just a question, what program are you using to measure your clock/multiplier?

 

CPU-Z, latest version, Everest Ultimate Edition benchmark and 3Dmark06.

 

So when you say its running at 25x on two cores, how do you know this? What program shows you what speed each core is running at?

 

Task manager will be up, showing load on all "4" cores, as well as Everest showing the real % load on each core and each core's frequency. cpu-z with show the multi hovering between 24 and 25x. I get a solid 24x while gaming at 1080p in SC2 and a 24-25x while working in Premiere and Photoshop. I have to have multiple sequences open and multiple projects open for the multi to dip. I rarely see it go below 3ghz.

 

So going by HWiNFO32 my 820qm is running all 4 cores at 21x plus...?

 

This is proof that the correct multiplier is not always reported. Sometimes you just have to use your best judgment.

For 'some people' if the program said 50x they would claim that their processor is super fast.

 

CPU-Z and all those other programs you mentioned do not follow the Intel recommended method to determine the multiplier. You are coming to conclusions about the speed of your CPU and how many bins of turbo boost it is using that are not accurate based on monitoring data that is not accurate.

When both cores of a Core i7-620 are in the active state, the maximum multiplier for both cores is 23.0. If your CPU is not fully loaded and one core goes to the C3/C6 sleep state, then the core that is awake can use a maximum multiplier of 25.0.

When lightly loaded, the maximum multiplier at any instant in time can be constantly swapping back and forth hundreds of times a second between 23 and 25. CPU-Z doesn't have a hope of showing this accurately.

ThrottleStop follows the method outlined in the November 2008 Intel Turbo White Paper and so does RealTemp.

-----------------------------------------------------------------------------
2.1 Problem with Today’s Software Reporting Frequency

Due to this, software that uses the Pstate structures, directly or indirectly,
will not properly display Intel® Turbo Boost technology frequencies.
------------------------------------------------------------------------------

If you really want an accurate look at what the average multiplier on each thread is up to then check out i7 Turbo.

http://www.fileden.com/files/2008/3/3/1794507/i7TurboGT.zip

Too many people have been brain washed by CPU-Z and Everest and 3DMark, etc. They are not accurate at reporting CPUs that support turbo boost when the CPU is lightly loaded. Picking a multiplier number out of a hat once per second is not the same as using the high performance timers that Intel specifically designed for the purpose of accurate multiplier monitoring and reporting.

I agree with you tenknics that the Core i7-620 is a better CPU than the Core i7-720QM for most users in most applications because of its increased speed. Do to increased power consumption, the Core i7-720 starts throttling its turbo boost back way sooner than the Dual Core. Even in well threaded applications, these two CPUs are very competitive and in applications that are not well threaded, the extra speed of the i7-620 will have it running faster in many applications.

That statement in your sig is wrong. Regardless of what CPU-Z might be telling you, there is no way to lock the multiplier on an i7-620 at 24X.

 

Lock might be the wrong terminalogy, but the point remains the same. Its not wrong because my CPU barely dips below 3.2ghz which is 24x. Regardless of the reported multi, the correct frequency is displayed. CPU-z is the standard for this, and although I understand what you're saying about it and the turbo multi, one thing is for certain is that the reported frequency is not wrong. 3 different programs are telling me the frequency. When I took a hardware survey through battle.net to start playing SC2, it reported I had an unknown Intel CPU running at 3.2ghz. Same with steam...Based on this and the 3 aforementioned programs I used, I made my sig. Again, lock is the wrong word to use, its more of an average, one that never changes...

I can't be anymore clear, my cpu never dips below 3.2ghz regardless of what I'm doing. CPU-Z may be wrong with the multi's but its not wrong with the frequency. CPU-Z is the standard when measuring clocks and overclocks, its used to determine world records... It's a certified respected program in the community, and if its reporting my clocks wrong then it's reporting everyone's Core clocks wrong who use it daily and rely on it daily.

 

I don't see how you could lock a multiplier in on any notebook unless the BIOS was open enough to allow for overlocking. When I had my desktop PC with an E8400 (3GHz stock), CPU-z would always report a 6x multiplier for a speed of 2.8Ghz even with speedstep disabled...sometimes it would switch it up to 7x and show 3.0Ghz but it was random and I wasn't doing anything else at that time which would put load on the cpu. IMO cpu-z doesnt consistently report accurate information.

 

How is that possible? If it is not accurate reporting the multiplier then how can the total frequency be correct?

Does that mean it is always correct? How about when it reports a 74 multiplier for a Core i7-920 which is a physical impossibility?

XtremeSystems Forums - View Single Post - i7 CPU-Z bug or WR ?

I used to try and make my software just like CPU-Z too until I found out that it wasn't always right. Now I prefer following the methods that the manufacturer Intel recommends.

You can find 101 programs that tell you your multiplier is 24 but that doesn't make any of them correct.

When I read stuff like that all I can say is, "You're wrong." You are being misled by the software you are using. The moment that a second core becomes active, the multiplier for both cores is limited to a maximum of 23. There is a few microseconds of lag time as the maximum multiplier transitions from 25 to 23 but that's it.

On a Windows computer you will never see the full 25X multiplier for an entire 1 second sample period. There is always background activity that needs to be processed that will wake up the second core and immediately drop the maximum multiplier to 23. That's how these CPUs work and any software that is leading you to believe something else, simply isn't accurate.

When you don't follow Intel's methods, that is the result. It's not always wrong but it's definitely not always right, especially with the Core i mobile chips that use multiple bins of turbo boost and rapidly transition between different amounts of turbo boost.

 

Of course CPU-Z is not wrong. Is the standard. Is the best. Is amazing. Is perfect.





But sometime is useless. The king is dead (CPU-Z) long live the king (RealTemp)!
Take my advice. For Intel CPU's it doesn't exist a better (and accurate) monitoring tool like RealTemp.
And as a side note I don't know another programmer as dedicated as Kevin to understand/find/solve Intel mysteries and FUD.

 

Hey, uh, I think your bird just died.

 

I have to disagree with your statements, at least regarding HWiNFO32.
I can confirm, that on Nehalem systems, HWiNFO32 might report Turbo Clock/Ratio on system which is not so heavily loaded. But the reason is not "Picking a multiplier number out of a hat". The true reason is that the CPU is extremely sensitive to load and puts the core into Turbo mode even if it's loaded for a very short time. This causes that the CPU frequency measurement method itself in HWiNFO32 can put the current measured core into Turbo Mode. I haven't seen such extreme sensitivity on previous CPU generations and there my results are in sync with Intel reference tools.
Intel has already admitted that the CPU shouldn't put itself into Turbo Mode so fast, so there will be an update to Turbo Mode behaviour (I can't disclose any further details).
There are of course other methods how to measure CPU frequency, ratio, etc, like the differential method used by T-Monitor. However after careful evaluation and testing I have decided to stay with the current method I use (and wait until Intel updates the behaviour )
I can also admit that Intel caused us many headaches with the way how Nehalem works, but one thing that I don't understand is why the CPU reports HFM+1 clock regardless in which Turbo Mode (how many bins) it currently is. I don't think it would be difficult to implement it reporting the correct ratio, so I'd kick them.... for this But this is not the biggest problem and can be worked out..
Each tool uses its own techniques - better for some situations, not so good in other... The fact is that Nehalem is a very dynamic CPU and changes it's operating point extremely fast, so it can cause P-State switch during the measurement method. Imagine that you measure the current frequency, read the ratio, measure BCLK as fast as you can. And between this the CPU switches it's state...
I'm still not convinced the Intel proposed method is the right one for our needs...

 

That's why I have a deep respect for Kevin. Without having access to inner Intel docs he managed to debunk most of Intel hidden secrets.
Now tell me who's right in that screenshot?

 

I agree.
I suppose you have Duty Cycles enabled. This is an issue I'm already aware of and plan a fix for this.
However, I'm trying to be in sync with Intel reference tools, and the current way of reporting (reduced frequency by duty cycles which result in reduced FSB/BCLK) is in sync...

 

Thanks for mentioning T-Monitor. It is the biggest joke going. The method it uses is completely wrong and results in nonsense being reported like this.



This CPU has the multiplier locked at 11.5 on both cores but because the CPU is idle, T-Monitor assumes that it must be using the 6X multiplier. That's wrong.

I like HWiNFO32. Are you using the method outlined in the Intel Turbo White Paper to determine the multiplier? If no, why not? It's as close to 100% accurate as you're going to get on a Core i CPU. At idle it might be only 99.9% accurate but as soon as you put any sort of load on the CPU it is 100% accurate. Intel wouldn't publish this method in a white paper if the results using this method weren't accurate.

The same method can also be applied to most Core 2 based CPUs other than the Atom CPUs that are missing the necessary high performance timers. Any method besides this method is not as accurate.

 

Still hating I see,


Everyone's entitled to their opinion. There is a reason why CPU-Z is used to officially validate speeds in overclocking contests and shows around the world, and that reason isn't to show false information. Go over to xtremesystems and look at any of the WR and OC threads. Every single one of them validate their scores with CPU-Z...


I guess all those hardware nuts who live to OC and use CPU-z are all just kidding themselves with false numbers....riiiiight

 

You must try to think yourself and not to be one of the herd.

 

When you find that reason, let us know. But who cares? CPU-Z isn't relevant to the Clarksfield vs. Arrandale debate since it has been proven that it does not show accurate information for these processor's speeds where it counts the most (i.e. maximum speed).

 

At full load when the multiplier is not rapidly changing, CPU-Z is usually accurate. When turbo boost is rapidly cycling on and off and changing the multiplier, it's not nearly as accurate.

Maybe when you're over on XtremeSystems you can check what program Asus engineers recently used to prove to users that the turbo throttling issues in their P6T line of motherboards had finally been fixed.

My ASUS P6T7 WS Supercomputer is here... - Page 20 - XtremeSystems Forums

I've been working with Asus engineers for months. It was i7 Turbo that showed them and showed users that their P6T motherboards had a serious problem when being pushed to the limit by enthusiasts. CPU-Z was totally ignoring this problem and not reporting to users what was actually going on inside their CPUs. Why? Because it's not accurate enough.



Mumak: In this example, using the Intel recommended method, ThrottleStop shows me that turbo boost (Intel Dynamic Acceleration) is being used 90% of the time on core 0 and the other 10% of the time it is using the default 10.50 multiplier.

( 90% x 11.50 ) + ( 10% x 10.50 ) = 11.40

The method HWiNFO32 uses and the method CPU-Z uses completely ignores this and tells me that this CPU is not using any turbo boost at all. Give the Intel recommended method a try sometime and I guarantee you that you will get excellent results.

 

Thankyou unclewebb. You told tenknics alot better than I could.

 

Because you knew none of what he said, of course you couldn't lol

How else can you learn? Its not about following the herd, it's about learning from people who do it for a living and seeing what tools and software are industry standard. Unclewebb is one of the rare few who work on this thing so I bet very few people knew about this CPU-Z issue with turbo, including yourself. I bet you think of yourself pretty highly when you make comments like that, haha, but you kinda just look silly. If I want to learn about the ins and outs of processor's and other computer hardware I'm going to go read about it and users who have it and experience with it, that's called being smart, not following the herd. By your definition everyone is just in a one giant herd on this messageboard because we all have the same laptop and ask questions about users who have this and that...

Go watch Good Will Hunting some more..lol