AMD/Intel CPU "Bang per Buck" study.

[Rastalovich]

May not be newsy, but it`s definately current affairs

Concerning both AMD & Intel, seeing as I`m gonna be in the market pretty soon, out of interest I wanted to see just how these processors work for their money.

Assumptions and data retrieval

• CPUs taken, I considered as both the comon and the popular, along with market leaders and power leaders.
  
• None of the results are based on overclocking. Stock levels are used in this case as overclock results are too inconsistant, especially when u can`t get 100% the same result from theoretically identical hardware.
  
• To grade the processor, it`s 3DMark06 CPU score was used. About 99% of the results were taken from Tom`s Hardware. 1 or 2 had to be taken from the FutureMark website. Using a single site, with it`s single conditioning, gives a fair-ish chance to each cpu.
  
• For each 3DMark06 CPU score, a P35 Mobo was used for the Intel cpu and an NF590 for the AMD. Consistantly.
  
• As with the above, none will benefit from faster memory (other than the known differences between NF590 & P35), graphics cards or anything known about at the time of writing.
  
• Prices were obtained from NewEgg, again merely to keep the singularity of data. Prices may vary from store to store - but on a whole, the level of the price isn`t `that` important, more the price in relation to each other. Again, 1 or 2 prices were not present at the shop in question, so an alternative was used for them, adjusting as a factor of their comparison to other cpus. In relation to US vS UK prices, again although they differ by something like 1.6x in some cases, the relation cpu to cpu within the store in pretty consistant.

Column explaination

• Name is obvious.
  
• 3DMark06 CPU result as explained above.
  
• Percentile takes the highest score from the 3DMark06 results and uses the highest value as a maximum benchmark. Other cpus therefor are a %age of it.
  
• Costing as explained above.
  
• $ per %, is a reflection of what the benchmark grading vS money spent to achieve it.
  
• BANG per buck, is merely the $ per % displayed as a factor of the worst performer in that category. Similarly to a maximum benchmark, this is taken from the minimum.

The data-spread is sorted by the BANG field, as this is the object of the study.

Conclusions are there to be drawn by the individual. It`s easy to see if u want the best processor `at any cost` or if you want a high BANG in relation to a big value in 3DMark06 CPU points.

To me, the Q6600 stands out. Whilst u can still make a comparison between the two camps (AMD/Intel) as to what they`re trying to do within the market.

Potential for overclocking has so many factors involved; individual`s skills, mobo, memory, `batch`; that to try and meld this into the results could be fatal to the whole study. But I think taking the results above, flavoring what other people have done - then judge urself, the hardware u will integrate with the cpu - those results would be subjective.

 

[!TIMMY!]

nice post, very helpful to see how the cpus compare! might have to grab myself a athlon x2 for cheap

 

[mrapoc]

Nice post rast! Reps

 

[Toxcity]

Nice one Rasty! Reps from me too.

Doing my job as well!

 

[prosser13]

Nice, but it'll take more than one benchmark to convince me of that order

 

[Rastalovich]

Thanks guyz.

Fair comment tbh.

Way I look at it atm, the one benchmark out of about 50 million that were available is meant as a "gaming" grader. It`s a loose enough a stat to use in itself, but what it`s meant to represent is how well a processor does against `modern gaming` - even with that said, the 3DMark is 06, we`re almost in 08. And we all know that gaming performance differs wildly when u team up CPU with mobo, gfxcard, phys card, sound card, drive acces method, OS.... frankly the list can go on.

A good arguement for a different stat would have been to include something like a UT3 bench. Something current and demanding. Might think about that some this week.

But, as with all reviewers, u would assume, particularly as the same mobo was used in each family case, that the best available ram was used and so forth. Reviewers of note tend to use the best they have at hand to push stuff, or allow the stuff to be pushed.

To be infinately more critical, would be to introduce all the benchmarkers and percentile each one... form an average, then gauge that against pricing. Greenies would like to factor in power consumed too. But I assume that gamers firstly don`t care if their gaming pc does well at rendering 3D Studio stuff or DivXs -OR- handle rar`ing with the greatest efficiency - as long as Crysis runs fastest with the CPU they can purchase in the money bracket they have. In the same breathe, if the CPU that runs Crysis @100% but uses more power than one that runs it @90% - imo the gamer couldn`t give a damn if we`re honest. Their parents might, or their company might - but for our arguements sake, the trees can be cut down, I`m running @ a constant 120 fps, so to speak.

It`s not really a table to criticize the CPUs in terms of what`s better than the other either. It`s merely there as a guide where u can look at it in a number of ways. If money has no bounds, and u may not be interested in oc`ing, u just want to look at the best Mark score. If u have a budget, u can line up the money u have put aside for a CPU and choose that way, whether it`s worth picking one that costs $40 dollars more (that could stop u getting a GTX Ultra), u can judge if it`s worth it using the BANG column against it. So in these senses, the table is just a tool.

The other thing to be clear about too is that as time, particularly months and release dates, go by, the table makes less and less sense. In that case it`s a snapshot for the beginning of November. Pretty soon a handful of these will become unavailable, some are rare already. But then ones are going to be appearing soon enough. Both these instances blow the whole thing out of proportion.

EDIT: Way I`ve used this guide so far is to look down the table (the 3DMark06) and notice when there is a big jump in performance whilst going down, as it`s sorted by BANG. Q6600 jumps right out and the Q6700 is something between going that extra in cash and jumping into crazy money.