Performance/Price Ratio (Standard Needed?)

[ElCapitan]

I'm new to this whole overclocking thing, but I used to design signal processing boards for cellular base stations. When you're doing this, you're constantly worrying not just about performance, but also about cost. It seems to me that many here might also be concerned with the amount of money they're spending, so it seems we ought to think that way at least some of the time.

I mean, sure it's interesting to read about people who climb Mount Everest, but for weekend mountaineers, it's really more important to maximize your climbing in the time you have available.

Similarly, I'm sure that there are plenty of people out there who just don't have $5k to spend on a machine and would love to see a top-ten "best bang for your buck" list. I don't know, maybe somebody has already done this?

As a starting point, you could just divide your 3DMark06 score by the amount you spent on the machine. To maintain a standard, the cost should include one DVD-R/W, and one hard-drive of at least 150GB. Cost would include Case, PS, all Fans, no monitor, etc.

As a starting point, my numbers (config shown below) would be:

3DMark06 / $ = PricePerf

13220 / $1300 = 10.2

Thoughts?

-------------------------
--> CoolerMaster C690
--> Intel Core 2 Quad Q6600 G0 @ 3.150GHz (350x9) - 61deg under load
--> EVGA 122-CK-NF68-T1 680i SLI
--> 4GB (4x1GB) Kingston DDR2-1066 @ 5-5-5-15-2T @ 1.89V
--> XFX 8800GT 512MB @ Auto-Overclock
--> Hitachi 750GB 7200RPM
--> Sony EIDE DVD-RW Dual Layer
--> Antec 750W
--> Zalman CNPS9500
--> 3DMark06 - 13,220

 

[Idontcare]

http://www23.tomshardware.com/...6&model2=882&chart=444

It's a noble endeavor, but fraught with complexity and logistics management overhead. One natural issue is choosing a benchmark to standardize on. Not everyone will agree with your choice, so you will be forced to include multiple benchmarks.

Its a downward spiral to "unmanagable land" from there. Or so it has been for everyone who has tried such to my recollection.

 

[Cogman]

sorry, but it is impossible. Because processors are better in different areas there is no way for a person to make an unbiased judgment on Price per Performance.

Not to mention, if some system like this was created, you can bet the driver tricks will be created by the different companies to give their processors the leading edge. It was done in the past, you can bet it would be done in the future.

 

[ElCapitan]

Ok, I agree that a perfect indicator is, well, not necessarily impossible, but pretty damned hard. Still I can't help but notice that everyone seems to have picked their standards: 3DMark06, and Framerate in a few different games like Crysis, etc.

The point here, is not to be perfect, but to give people a general idea of where they can get for a certain dollar value, and what combos produce (generally) the best ratios.

Thanks for the CPU chaft< Idontcare, but it's CPU only, which doesn't take into account the complete package.

Cogman, to your point, would you say most OC'ers are probably interested in gaming? If so, than we don't need to worry about DVI->MPEG4 frame conversions or anything. Hell, if you wanted to, you could create a price-perf ratio for that too.

 

[ElCapitan]

One other point. I'm not talking about an industry benchmark. I'm just talking about something OC'ers could all use just to compare.

 

[DSF]

For the heck of it, 11162/$940 = 11.9.

--> Antec Solo
--> Intel Core2Duo E4500 @ 2.93GHz (266x11)
--> Gigabyte GA-P35-DS3L
--> 4GB (2x2GB) Supertalent DDR2-800
--> XFX 8800GT 512MB
--> Samsung 500GB 7200RPM
--> Corsair 520HX
--> ZeroTherm BTF90

 

[Foxery]

Originally posted by: ElCapitan
Similarly, I'm sure that there are plenty of people out there who just don't have $5k to spend on a machine and would love to see a top-ten "best bang for your buck" list. I don't know, maybe somebody has already done this?

The best chips always become common knowledge... you can tell what people buy just by reading the thread titles in this forum, without even reading the messages. If you need more detail, there are a dozen good review sites who will give you pretty graphs and a list of price tags.

Hell, ArsTechnica does regular updates of complete machines within certain price ranges. No independent thought required. If it were any easier, hardware companies would lose a lot of profit margins that currently come from suckers.

 

[Idontcare]

Originally posted by: ElCapitan
Thanks for the CPU chaft< Idontcare, but it's CPU only, which doesn't take into account the complete package.

Was merely trying to point out what others have already done in their endeavors to do as you are attempting. Well-funded organizations have attempted to crack this nut, tomshardwareguide got the closest I've ever seen.

This biggest challenge, bar none, is figuring out how you are going to deal with the fact that prices are variable, not constant. If I calculate my performance/price this week it will be different than last week (because prices changed) and will change again next week.

 

[LOUISSSSS]

thats dumb because of many things
-3dmark sucks
-people might want to spend money in departments where there is no performance increase, such as storage space, high quality case/psu, sound card, quiet pc parts, etc etc etc x 20.
-theres no sole "performance-test" for everyone's PC cuz everyone uses their pc differently and not only got one game.

 

[ElCapitan]

"Dumb" is probably a bit strong.

To the person who commented on changing prices, I think the whole point of this would be to track the moving target.

Regarding things that don't matter to performance, we could leave off HD and Optical drives. Anything else?

I guess the whole point here is just that some might want to go cheap. I can certainly afford $5k for a machine. My new $1300 box is sitting right next to one I paid $5k for about 5 years ago. I LOVED that machine. This time, though, I wanted to try to squeeze more juice out of my dollar. (I'd rather put the other $3700 in the kids' college fund...)

May some of you naysayers have a better idea?

 

[LOUISSSSS]

we can eliminate the motherboard too for features such as wifi, more sata, raid, pci-e slots, firewire, cooling solution... etc.

 

[nyker96]

the P/P ratio will differ depend on your specific app and setup. Say if you game on hi-res and the vid card is limiting, then the difference between a Q9xxx vs. X2 6000+ will be negligible and since Q9xxx is much more expensive, you can say it will have very low P/P. On the other hand if you doing H264 compression, that Q9xxx will be incrediblly fast, so fast in fact, the P/P ratio might be infavor of it. So like I said P/P will be specific to your app, your setup.

 

[Cogman]

Originally posted by: ElCapitan
...
Cogman, to your point, would you say most OC'ers are probably interested in gaming? If so, than we don't need to worry about DVI->MPEG4 frame conversions or anything. Hell, if you wanted to, you could create a price-perf ratio for that too.

Nope, I wouldn't say that. Most OCing in general is in the minority. Some that are interested in it are interested because of gaming, but most don't have a clue about. Me, I enjoy overclocking and doing video encoding, gaming is one of my lower priorities. You might as we say that everyone that gets a computer is interested in gaming, and you be just about as close to the truth.

So, if we are just going to create separate benchmarks for each area of computing, why not just leave the PPP out of there and let the computer guys handle it? After all, anyone that is interested in actually reading a review with 10 more graphs dedicated to PPP would be able decided from earlier graphs where his cpu should fit in.

And on the my real point, How do you determine which application has what effect on the PPP ratio. Because even across different games you see that different CPU's preform different ways (Heck, even in video encoding we see this from codec to codec). So do you judge all the games equally? or do you say "More people play this game so lets give it a stronger influence on the results". Even if you keep everything even, there is still the problem that you haven't tested every game. Because of that, you will need a pretty heavy testing pool to get accurate results.

And if you testing pool is too small, like I mentioned before, Driver optimizations will creep in to give one product a leading edge over the other.

Sure, it would be nice, but I honestly believe that anyone interested enough to read reviews with that sort of ton will be capable enough to do a price comparison. By all means though, if you want to try and set it up, go for it. But just be warned that its not as simple as a task as you may be envisioning.

 

[heyheybooboo]

As noted - I think there are too many variables ...

and that we spend much too much time looking for a 'one size fits all' standard when, as noble as it may seem, it is impossible.

If you used your rig 100% of the time for one specific task, it's performance can be measured and ranked against systems performing that similar task 100% of the time.

Beyond that, your measurement would have to be 'weighted' in a dizzying array and the design standards for ""P/P"" would become more and more ambiguous the more ambitious they became.

And when you introduce the human element (we're all different), your measuring stick becomes more an issue of conflict and arguement, instead of satisfaction, which I think is the ultimate measurement.

As opposed to brute force, I tend toward the minimum. I underclock and undervolt but off-load intensive tasks to '2p' systems when necessary. I'm also a 'late-adopter'. I look forward to my new e8400 - in the summer of 2009

I've got 6 apps open, I'm typing this post, listening to 'Meet The Pest' on the HD tuner, motoring 1005MHz at 1.125v and 16-19% cpu utilization. I've tried dropping the voltage further but the %^@#!! locks up - lol

All that said, under your Futuremark Standard, on the latest rig I built I 'qualified' a 12.2 (9152 3dmarks - $750). I could have easily scored over ""13"" by selecting a different mobo, PSU, case, etc.

MSI 790fx - x2 5400+ at stock 2.8GHz - 2900pro OC'ed - 2Gb ram - OEM XP - single HDD & optical - Silencer 610 - Enlight case - 3 fans & a 'recycled' Opty heat-pipe cooler

And I don't even like to game ...

 

[Idontcare]

Originally posted by: ElCapitan
To the person who commented on changing prices, I think the whole point of this would be to track the moving target.

And who do you propose does this tracking? What currency do you intend to normalize your database too? You aren't assuming everyone here lives in the USA are you?

This is merely getting back to the point that the gating issue is management...someone has to manage the collection, maintenance, and constant adjusting of all the data to "track the moving target".

Ready for a full-time job? Or are you hoping somehow the collective would just do it?