Processor upgrade for my old desktop

rendeverance

Hi all,

I sort of run at the value end of the tech curve when it comes to building my systems, and usually keep systems going for some time before going through the rigmarole of setting up a whole new system. I tend to build a new system a generation behind the bleeding edge, and thanks to intel and a lack of generational power leaps I have managed to run systems far longer than I previously have.

As it happens I have a few systems but my main desktop workstation has been flogging an old i5 2400 for nearly 7 years. Despite being locked it is overclocked and running just over 3.9GHz on all cores, and JUST ABOUT copes with my wildly eclectic workload (general productivity, some gaming, some video and photo editing, modelling 3d for printing and some mathematical simulations), but it is certainly showing its age. As of late it has become clear I need to renew the thermal paste, as temps have risen from sub 70C maximums, when I first overclocked, to peaking at around 90, and heating my bedroom a little too well. So, I thought, if I am going to rip the CPU out and refit, I may as well upgrade, the system is on 24/7 so efficiency gain of any sort would be appreciated.

With Ryzen 3 around the corner and wanting to save to build something substantial as my next build, I am hoping to get another year or so, and also to make the system a little more energy efficient. So I have just purchased a 3570K off ebay for £33. I would have considered an i7, but I wish to save money for the upcoming build.

I am hoping I will notice the difference in power consumption and speed, assuming I can run a mild OC of 4.2 and stock voltages? I am not running stock voltages on 2400.
A performance gain of 20% would make a huge difference to me. As would a 20W power saving. Does this seem reasonable?
Has anyone had much luck with undervolting ivy bridge with a mild OC?
Any thoughts on how successful you think my upgrade will be, or what I should look to for my next system would be welcome, I definately want more cores in the next build

My system is as follows:

i5 2400 @3.9 > i5 3570K @4.2
Gigabyte x68 mb
16gb 1333 ram
Radeon 390 gfx
Assorted SSDs and HDDs to the sum of around 16TB
I usually game on 1440P/2k on a single monitor, but have three screens in total (the other two are 1080P).
decent 600w psu
(cant recall exact components as not at home and its one of several PCs)

EndlessWaves

Anandtech did a big benchmark on the release of the 6000 series models and found that Sandy Bridge to Ivy Bridge was a 6% improvement at the same clockspeed. Combined with the clock speed change I'd expect around 14% improvement on average:
The Intel 6th Gen Skylake Review: Core i7-6700K and i5-6600K Tested

Generally overclocking ability has also reduced over the years as Intel have optimised the design so getting a better overclocker is also optimistic. .

You're making a common mistake regarding heat and temperature. You're thinking of it like an oven, where a higher temperature is due to more heat. It's more like a thermos flask, where a higher temperature is due to better insulation. The heat flowing out of the system is determined is the same whatever the CPU temperature. Or almost the same, temperature does have a slight effect on efficiency.

If the room is hotter it's more likely to be because your use has become more demanding and the CPU is spending more of it's time working.

rendeverance

Hi, thanks for your reply.

Useful link, thanks. My reasoning for expecting performance increase was on the assumption of about 10% due to the architecture and another 10% hoped for by terms of clock. With that information 14% seems a fair guess. As for how good an overclocker it will likely be - I assumed 4.2 would be a very mild clock that most chips should reach?


As a scientific person I should have been clearer in my OP with respect to my reasoning for expecting to save power and thus resultant heat - it is not because I have pictured it as you describe (though I would imagine most often do so hopefully your very easy to understand explanation useful to some that read this):
At the moment I have a 2400 which is OC'd and overvolted to ensure stability, it also has a higher TDP of 95 vs the 77 of the ivybridge. I am hoping to keep the powersaving features and stock voltages on the ivybridge, such that in normal use (i.e. not benchmarking) the chip itself consumes less power = less heat into the room. Since the TDPs of the chips are 18w apart, and the system is under medium-low utilisation the majority of the time, I hope to save power on average.
I am well aware that my cpu temp increase does not change the power consumption of the chip or heat emitted by the PC, and that the core temp issues now most likely due to breakdown of conductivity of the thermal paste, since cleaning the HS made little difference, and no settings etc have been changed for years.

So hopefully shooting for a 14% increase in raw power, not 20% but not nothing either, so still happy...

My main aim here was to save some power and heat, as that is what has drove me to take the system apart. Does my assumption that an overvolted 32nm chip will consume more than a 22nm chip on stock voltage seem reasonable here?

Sure, 20W was a guess, and its possibly hard to quantify, since few people seem to go for a balance of economy and power when they overclock, and as such I was struggling to make a more educated guess from reading benchmarks. Since this may see retirement as a media centre PC, I posted on AVforums as I thought there may be a few people that have done exactly what I intend, built a HTPC system with a 3570K, and thus didnt want extra heat or power use for an always on system.
If I am not wrong in my assumptions then the bonus is the energy saving over a year of 24/7 running would possibly also cover the modest cost of my chip

Absolutely...