Price-performance ratios for CPUs are completely bogus, because you never run a CPU by itself. If your case/power supply/motherboard/RAM/disk collectively $400 and you have a choice between a $100 CPU with performance X and a $200 CPU with performance 1.5 X, the faster CPU has a performance/price ratio which is 25% worse -- but using the faster CPU gives you a system with a performance/price ratio which is 25% better.
That is assuming CPU is the defining factor in the speed of your computer. For vast majority of users, this is just plain false -- a 60$ Athlon II dual core with a 200$ SSD massively outperforms a 950$ i7 975 that is paired with a traditional HD in nearly all of the use cases a normal desktop user encounters -- And if you are into gaming, that same 60$ processor paired with a 300$ HD5870 high-end graphics card easily outperforms the same i7 paired with HD5850, the next-best model that costs only 50$ less.
Good point, but not always true. It's a myth that game performance is universally GPU-bound. Most notably, most MMORPG's, many RTS's and pretty much anything made with the Source engine are more dependent on CPU performance.
Sure. All performance tuning depends on figuring out where the bottlenecks are; but having found said bottlenecks, many people make the mistake of considering price/performance ratios for those individual components rather than for the system as a whole.
This isn't just relevant for CPU performance -- people do exactly the same thing with disks. If you want the most disk space for the least expense, it's worth buying 2 TB drives even if they're more than 2x as expensive as 1 TB drives.
Are you sure? A 2T drive puts all your eggs in one basket if it happens to fail. And you can't RAID it (though you can't RAID 2 x 1T disks either, unless you are crazy and want to strip).
I still find it a useful list. It lets you focus on the stuff you know is well priced.
My priorities for stuff to splurge on - PSU, case and cooling (all very important to protect your other gear, and reusable), monitor (since you'll probably be stuck with it for 10 years, and it pays off even when you are doing lightweight stuff like reading yc), hard drives, then finally RAM, CPU and mobo (in that order).
For gamers, 3D card and a faster memory path (recent CPU, good mobo, fast RAM) are probably more important.
But we always get sucked into caring about the CPU, for some primal reason.
Apparently I'm the only one who only saw a single CPU listed, with a hundred in the no price section?
Clicking update I get tons of:
Warning: preg_match() expects at least 2 parameters, 1 given in /var/www/paul.slowgeek.com/compare/cpu/update.php on line 19
{"name":"686 Gen","score":"267","rank":"718","search_name":"686 Gen"}
Warning: preg_match() expects at least 2 parameters, 1 given in /var/www/paul.slowgeek.com/compare/cpu/update.php on line 19
{"name":"AMD Athlon 1640B","score":"705","rank":"442","search_name":"1640B"}
Warning: preg_match() expects at least 2 parameters, 1 given in /var/www/paul.slowgeek.com/compare/cpu/update.php on line 19
{"name":"AMD Athlon 2500+","score":"423","rank":"567","search_name":"2500+"}
Using a synthetic benchmark is not exactly very accurate in order to determine price/performance, as not everyone is going to buy a processor to do Passmark's benchmark all day, not to mention the fact that certain processors might not contain the same features as others (ex: virtualization support).
Since I haven't bought a new desktop machine in nearly three years (until this machine, I upgraded religiously every 18 months), because my current CPU is still plenty fast for the work that I do (partly because I no longer work with C/C++, and partly because processors are just so damned fast these days), I would like to see power consumption / performance mapped out like this. So, energy efficiency is a much more important factor to me these days.
From looking at the list, it seems the benchmark assumes a task that largely scales well to as many cores as you have. That's not nearly always the case, many or most games are not well threaded and programs like Lightroom only use 2 cores max. It would be interesting to have the same table but ranked on single-thread or dual-thread performance, too.
Great site. Would love to see sliders for min/max price and performance. That way you could slide the minimum performance to 4000 and find the best cheapest CPUs that fit that criteria.
The whole line has been marred by the them continuing the ship buggy processors with the bios work-around. Would you buy another Phenom if you had the misfortune of getting one of these?
I was excited to see that the chart had an Intel 800MHz, but when I clicked the link, it was to a AMD Sparta 2.3GHz. There might be other mislabeled products on here.
Yeah, the poorly labeled CPUs don't search well on newegg. Please let me know if you find any others. I've hand checked the ones I cared about and they seemed good.
Interesting how AMD kills Intel on Performance/Price, but Intel's raw performance is much higher. Still, the highest-end Phenoms are pretty quick and still a reasonably good value.
You need to factor in the other costs too. If you buy a computer you need more than just a processor. If the rest costs $200 you have an extra cost of $400 + processor costs for two AMD's against $200 + processor costs with Intel (assuming two AMD's are equivalent to 1 Intel).
Most gamers, I know pick the best 100$ CPU they can on price performance because CPU's are just not that important. Granted if your building a 3,000$ system you might want to look at the CPU, but why pay more so you can have extra cores doing nothing all day.
The top listed performance/price AMD processor is known to be buggy, which is why it's so cheap.
People that are buying the high-end systems are usually people who play PC games competitively. Intel is also making compilers that are optimized for their processors. In addition they are creating tools for optimizing parallel processing. I don't see AMD doing that. If they don't they won't be able to compete in the high-end desktop market.
Really, are you talking about the B2 version, something wrong with the cache? Because I bought my 9750 a year ago August (it had the lowest power consumption at an acceptable speed) and it was a B3 even then. So, if that's what you're referring to, I doubt those are still in the channel.
Come to think of it I think they renamed the fixed ones (the ..50) so that there was no risk of confusion.
I'd say that's the obvious result, actually. Where there is competition (in the low to mid range), the smaller vendor generally has to pick the lower price point as a growth bet (because they're not the "default" choice). Where there is no competition (at the high end), Intel gets to set whatever price the market will bear without worrying about what AMD does.
I've never had problems with Intel (and I have with AMD) so I am almost certainly biased. That said, if [1] the best AMD can do in a performance test is half of Intel (at any price), I would say AMD has a big problem and needs to improve badly.
For normal users, unless you are doing video-processing or any equivalently CPU intensive work, pretty much anything will do these days and the low-cost, whatever-performance-will-do market is highly saturated. If you want to make money and stay in the game, you need to appeal to the big spenders and they want more power.
[1] I hereby acknowledge that the test might be flawed or not representative for actual use.
Yeah - looks like it might be a problem with IE. I opened it in Firefox and it looks totally different. In IE (8) its just a plain list - not a table with zebra stripes and sortable columns. It looks like the javascript for the datatable is throwing an error in IE.
