High End Processors and ROI over life cycles?

W

WBurchnall

2[H]4U
Joined
Oct 10, 2009
Messages
2,622
TL;DR Is there better Return on Investment for spending more these days with multi-core processors than there was 20 years ago? As now you get a direct upgrade in core/thread count rather than in the 90s/2000s where you might just get a better binding and a 5-15% increase in clock speed and slightly more cache?

I just wanted to get people's opinions on high end processors and their return on investment. I'm going to be doing a new build soon and traditionally, I tend to get processors in the $350-$400 dollar range. Mostly because in the past I noticed the 'Extreme' versions for $900 dollars more seemed to be identical in core count and just gave a small improvement in the form of say "0.4 Ghz faster" and the performance gains for the extra $900 dollars were in the range of 20%. So I always went for something mid-range for the gaming crowd as it seemed the best bang for the buck. Like my last processor likely 6 years ago was an i7-930; before that was a Q6700 and Core 2 Duo E6600 before that etc... A reasonable gap between processors.

That being said, I notice that now for $1000 you can start to get processors that have double the core and thread count when comparing it against the $350-$400 dollar processors and for $1500 almost triple the core/threat counts. So it seems like there might be a significant performance increase; especially 5 years down the road if we are moving towards programming for higher and higher thread counts.

IE 6 core(400) vs 10 core (1000) vs 14 core (1500).

I'm just wondering, what people's opinion is on what the return on investment over the lifecycle of these newer high core-count processor will be? Is it much better to spend the extra $900 now than it was a decade or two decades ago? IE those of you who bought say the Intel Extreme Edition processors back when the i7-930 was out, was it worth it? Did it help you continue to use the processor and keep it comparable to new ones long after it was long in the teeth?

I'm mostly motivated to upgrade my i7-930 as when I look at CPU-benchmarks on CPU-Boss or random websites, the raw performance per core seems to be 50% higher and 30% higher clock speeds are resulting in double the estimated overall performance at some tasks I commonly do (video conversion, gaming, SolidWorks, AutoCAD, Revit, etc).

I'm just wondering now that there seems to be almost linear-scalable-upgrades, is there any disadvantage to getting a $1500 dollar 14-core processor when intention of using it for 10 years? Will new processors have new instruction sets or something that holds it back significantly? I do know my i7-930 has limited support for x265 due to it's age relative to x265. I think it cannot natively decode 10-bit 4k and requires the software to do it during playback giving high cpu usage where one with x265 instructions barely sweats. IE a $60 dollar setup box from China out does my i7-930 at 4k 10 bit. Sad.
 
Last edited:
Do you make money off the work you do on your computer?
 
All depends on system load factor and how much time is worth to you/your process.

Also, most folks that are seriously considering this metric, aren't looking at 10 year life cycles for hardware. It's not uncommon to find ROIs measured in weeks or months in high density/high load factor deployments.

In the tech industry, apart from some niches, such as embedded or large enterprise use, 10 years is an eternity and considered legacy support at that time. You'll pay a heck of a lot of money just for the privilege of having repair and support options in years 6+.
 
Skillz said:
Do you make money off the work you do on your computer?
Click to expand...
Yes, which in Canada, just means I don't have to pay income tax on it but still have to pay GST/PST/HST/etc. So I get about 40-50% back and my income for the year counts as less according to the cost I spent which can result in more government benefits or pay less in MSRPs and other government non-tax, taxes. I doubt $2,000 will drop me a tax bracket though unfortunately.
 
Last edited:
you buy shit and you sell it to buy you more shit. basically the IT World.

I sold my 7820x to buy me a 9700k: i sold shit to buy more shit.
 
A very complex question to answer, and a lot of the answer depends upon the volatility of the market you are in. Is there significant value of time? If so, spending up the chain on what would normally be diminishing returns still makes sense.

I work in an industry which as both factors. Sometimes, being first AT ALL COSTS has a completely nonlinear payback. In other mature areas, you want to have a more temperate response to spending. Given this, when working in each area, your budget reflects it.
 
You could only make a ROI argument for CPU upgrades if you are sitting around waiting for it to finish something instead of making money. If you can just leave the thing running at night or something then it would be harder to justify.
 
bigdogchris said:
You could only make a ROI argument for CPU upgrades if you are sitting around waiting for it to finish something instead of making money. If you can just leave the thing running at night or something then it would be harder to justify.
Click to expand...
Exactly. Unless you are running one task 24/7 for income, then you aren't calculating ROI where one CPU would outperform another.
OP I am guessing you are over thinking this. Pick your budget and buy the most cores you can. Maybe look at a platform you can drop in a replacement with twice the cores three years from now (AMD). Also even a modest GPU will handle your x256. If you game you should be running something decent?
 
Also note that productivity software often favors more cores or more frequency, but not always both. Check the reviews to see if you are better off with a faster single threaded chip or with more cores.
 
bigdogchris said:
You could only make a ROI argument for CPU upgrades if you are sitting around waiting for it to finish something instead of making money. If you can just leave the thing running at night or something then it would be harder to justify.
Click to expand...

Well there is also the personnel time in that equation as well. Maybe it's automated and there is no people time, maybe it involves some setup before you can just let the CPU run. Maybe it's something liek running filters on images, where you just have to sit and wait for it to finish before you can proceed.

The people time isn't always insignificant, and it isn't always that you can just set a process (or batch) and let them run all night long. For myself, I value my time highly. If I have to sit and wait on a computer a few less minutes a day, even if that CPU is running idle all night long - that still has an ROI associated with it.
 
Well. Something to consider that I haven't seen mentioned here -

When you move into HEDT space, other components, specifically motherboards, cost significantly more. 10 year lifecycle is asking a hell of a lot, though.
 
If you have to deal with say software licensing per socket costs that are high, it could make sense to replace dual socket workstations with new higher core single-socket machines.
 
You must log in or register to reply here.
Back
Top