New 4700MQ vs old 3630QM

I was wondering if anyone knew how much better the new 4700MQ performs than the old 3630QM, both have the same clocks 2.4 Ghz to 3.4 Ghz and both have a 6MB cache.

I am curious because I do a lot of programming, computational power means a lot.

Thanks in advance,
Jake

 

Roughly the same. I'll guess somewhere between 5% to 10% increase in CPU power, though more likely nearer 5% or so. For non-Ultrabooks, Haswell seems pretty ho-hum.

 

Nothing too ground breaking on the processing side, gpu performance is impressive however and power consumption is down, so good for mobile users.

 

True, but it's a moot point if you have a dedicated GPU as far as graphics performance is concerned.

 

It'll mean more for ultrabooks users than for power users, and seeing as most i7 come with a discrete card, it is not a huge factor.

 

Anyway, in summary I don't think that Haswell would be too special for power users. Probably would be a better idea to get a discounted Ivy Bridge system and pocket the cost difference for a computer upgrade or a decent lunch .

 

My take on it differ, the power users will be likely to upgrade because every CPU cycle counts, so even a 5-20% (according to Anandtech) clock for clock performance increase can translate to some time saved if you spend your day crunching numbers and rendering stuff. So purely in terms of hardware, Haswell is still interesting, however if you factor in price/performance increase, it can get less interesting so power users with the cash to upgrade will likely upgrade.

Gamers on the other hand will likely not see much of a difference.

 

For casual users, gaming, occasional video/music editing, the difference is negligible. For profit, yes, every second saved is worth money.

 

Yeah, I'm personally looking at this from a price/performance ratio. IMO, if someone's not performing scientific computing or anything else where CPU cycles are absolutely critical, then it won't make much of a difference performance-wise.

 

Agreed.

/10

 

It's kinda funny how the public usually portrait scientists.

Most people with scientific research titles don't give a damn about CPU circles. They are very happily and proudly running their time-proven designed-in-the-1980s-by-a-layman single-threaded non-cache-aware just-loop-forever holy scientific software on their half-a-decade-old was-super-expensive-and-therefore-must-be-fancy-even-today Xeon®-powered PROFESSIONAL workstations. This is how science is done. Who will waste precious Scientific™ time on this Haswell thing? Prefetcher tweaking, EP rearrangement and AVX extensions are dirty jobs for the uneducated low-rank workers (aka IT-capable kids).


Now back on topic:

The theoretical boost in Haswell's pipeline is somewhere around 10%-20% depending on how you look at it, and there are some edge cases in which higher levels of improvement are expected. You may notice it from time to time, but in general there isn't much difference in terms of (expected) real-life CPU performance. If you've never done any assembly-level performance optimization, my suggestion is "just forget abort it". Compilers will catch up later and handle most of the work for you anyway.

 

Some of it goes that way, not all though, the main thing here is once again budget, so they have to make due without having to refresh their hardware. That said, most of the clusters my uni has access to is as of now mostly 2010 and up gear, the same goes for some of the workstations, not all of them though. People in CFD do give a damn, it's just that money to buy new equipment isn't exactly easy to come by.

Heck, my research supervisor upgrades his hardware every 2 years or less just to get number crunching done faster, so some scientists care, some don't and some care but just don't have the money to buy new equipment as often as they'd like.

 

A good comparison of 4700MQ and 3630QM is at notebookcheck:

Review One K56-3N2 (Clevo P157SM) Notebook - NotebookCheck.net Reviews

 

But wouldn't such job be better reserved for desktops that can do more with a smaller budget?

 

If you don't need the portability, yeah, but laptops have their uses as well.

 

Depends, how flash are you at optimizing your code for AVX/AVX2 or hyperthreading? if you can then you will notice at least 25-50% speedup at the same clockspeed otherwise your performance increase is only about 10% on unoptimized code.

 

Because you have a budget of $8.1 million for a staff of scientists. But those new analysis equipments are going to cost you a total of $8.09 million, leaving $10K to upgrade several dozen computers.

And you can't upgrade if the equipment manufacturer refuses to release a Windows 7 driver to replace the Windows NT 3.0 driver. You thought printer driver compatibility was bad enough? Try working in the scientific field where the equipment manufacturers have no interest in lowering their prices or maintaining driver support.

 

If the equipment needs a Windows NT 3.0 driver to run (I've been playing with some.), buying an ancient computer or a few costs nothing. Then they can do the heavy lifting on the faster ones.

 

Not all programs have the ability to off-load computing to a different computer. Coding that functionality in might not be any better than putting pressure on the application developers to make an updated version of the software (and then charge $$$ for it).

 

i7 4700mq is 47 watts and better suited for 15-17" laptops for the added 200mhz clock and is slightly better than the i7-3630qm

Meanwhile there is the i7-4702mq which is also a quadcore at 2.2ghz-3.2ghz and since it uses 37watts, it is perfect even on lighter and smaller laptops.

 

@Jarhead
Being forced to run the same app for both online data collection and offline processing? In that case it would be very tricky.

@Benchmade 42
I find the naming scheme confusing...

 

Unfortunately, it's the way it goes sometimes, some spectrometers and HPLCs I had to use, use the same software for running the tests as well as doing some post processing. Of course, there are other software you can use to process the results sometimes. If you're lucky you can easily export your data in .txt or .csv and then port it to other software or the software you use for post processing of the results can take the native file format. However, there are other times when the software you use for online data collection stores the results in some sort of database which can make transferring the data to another computer somewhat tricky. Then there's always the cost involved with buying a second piece of software to do the post processing when you already have something capable of doing it and running the tests on top of that.

Intel's CPU naming schemes have been confusing for quite some time now.