System comparisons (CPU, etc). How?

OK, so I feel like an old fogey now.

Back in MY day, things were pretty straightforward.

If you were doing advanced work on your PC, you pretty much ALWAYS craved more power.

CPU was fairly easy to compare - mainly it was about megahertz, though type/generation of CPU mattered, too.

More RAM was better - fancy types of memory seemed interesting, but had little impact on speed.

And the same was more or less true for hard drives - bigger was better, type didn’t matter much for speed.

Right now, my primary home PC is a somewhat old build (about 3-4 years old - but was originally fairly top of the line). My wife’s is also old (~4 years old - but was originally fairly basic).

I’m trying to decipher PC specs.

IIRC, there used to be a site where people posted the results of some kind of CPU benchmark, and so you could see scores for many generations of CPUs. I thought it was on Tom’s Hardware, but I don’t see anything quite like I remembered. I’m not interested in the nuances of this benchmark versus that - just want something that’s broadly accurate (within 10-20% or so), and preferably has not only the latest CPUs, but older stuff too.

Re: memory - there used to be a ~2-3 GB limit on what the OS could address. Did that go away? With which OS? Other than “more is better”, is there much to know about memory?

Re: hard drive - any weird quirks to know about here, or simply “more is better”?

FWIW, if I get a new system for my wife, it’d likely be something off the shelf - HP or something. For my own needs, if any, I might go off the shelf or build it. Both of those would likely be desktops. Also, I might consider a cheap new notebook for my kids, who are currently using something ~8 years old.

Aside from normal web, e-mail, and the like:

  • I use my PC for all kinds of CPU-intensive stuff.
  • My wife does photos and video

High end 3D graphics don’t matter.

So, of the various specs on new machines, aside from obvious stuff that I can determine on my own (size of RAM and hard drive, connectors supported and the like), which of the more subtle specs matter, and how can I find out more, especially about CPU speeds?

Just build yourself a midrange system (midrange by today’s standards):

Asus P7P55D Pro
4GB or 8GB DDR3-1600
Intel Core i7 860 (2.8GHz)
AMD Radeon HD 4770 or Nvidia GTX 260 Core 216
Any good 600W PSU
Windows 7 64-bit

If you’re using a lot of memory intensive tasks, up it to:

Asus P6T
6 or 12GB RAM (one or two triple-channel DDR3 kits)
Intel Core i7 950

The new Nehalem CPUs from Intel are much faster on photo and video editing and transcoding.

Other gear can be the same. But this setup will cost about $500 more than the first one.

As for storage: if you want really fast boot and load times, get an SSD, but they’re quite expensive. Otherwise, any 1TB or larger, 7,200RPM SATA hard drive will be fine (Seagate, Western Digital or Hitachi.)

You might have been thinking of the charts section of Tom’s Hardware, which has pretty extensive benchmark comparisons over wide ranges of hardware.

(Hey, this is thread #55555)

Case - is there any good CPU benchmark site that will allow comparison of a broad range of CPUs (new and at least a few old). Among other things, I’m curious how much of a CPU win I’d get by upgrading the various systems.

Fugitive - yeah, I’d found my way there, but when I click on the “2009 Desktop CPU Charts” link, I get all kinds of sub-benchmarks. I just want something simple. Perhaps that is the “Performance Index” which is apparently a bit of a blended benchmark. But that one doesn’t show all that many CPUs - maybe it’s just what’s currently on the market (or was around the time the benchmarks were conducted).

I’m not sure which, if any of these CPUs are notebook-oriented CPUs.

Interestingly, the performance range is relatively tight - just over 2 to 1 from best to worst. Back in tha’ day, IIRC, there was typically a much larger performance spread from the top to the bottom of whatever was on the market at a given time. But maybe this list is only a small portion of what’s out there.

There’s a chart there specifically for mobile CPUs, charts from previous years (you just can’t merge them all into one big chart), and if you just want a single number you could look at just the 3DMark06 CPU test.

But that’s kind of where things break down since it’s not necessarily all that meaningful to compare a 2009 CPU’s 3DMark06 score against a 2005 CPU when 3DMark06 hadn’t even been released yet. Or the list of games used were completely different. Or the other products were different versions. I think Tom’s charts are about as broad as you’re going to find while still being meaningful.


CPU naming has gotten stupid with the 4 digit numbering sequence along with GPU naming sequence.

Not so much. But if you are looking at photo and video editing, Nehalem based systems are well over 2x what the older Core 2’s were.

Anecdote from real life:

Last year, I was working on a project in Adobe Premiere. Burning the DVD took about 50 minutes on a Core 2 Quad Q9770, which runs at 3.2GHz.

I moved the project over to a Core i7 965, also running at 3.2GHz. The DVD encoding and burn time went down to 15 minutes. Similarly, re-encoding previews took a lot less time.

So yeah, get some flavor of Core i7, with Hyper-Threading (800 or 900 series), if that’s the type of work you’re doing.

Yikes, so to describe a CPU now, you need to use 4 different identifiers? (Core i7 965 3.2Ghz)

And a slight difference in a couple of those identifiers (swap 2 Quad Q9770 for i7 965) and you’re talking a 3X+ swing in performance for one key type of app?

How do regular consumers sort this stuff out?

Yeah I haven’t even figured out the Core shit yet. I figured out the naming with Intel for the last generation:

E2200 E6600 E6800 Q9400 Q9650 etc.

E = dual core, Q = quad core
higher the number the better, 2nd number indicating …oh fuck it. Same sort of shit for Radeon/Geforce.

If you’re talking everyday, run-of-the-mill Joe Consumers, they typically buy on features and price – rarely on performance. There’s some understanding that higher priced systems (usually) perform well, but that’s about it.

Another anecdote:

Years ago, when I was writing graphics card roundups for Computer Gaming World, I was at Fry’s Electronics. I saw a guy walk in, carrying an issue of CGW, with one of my roundups. He peered at the magazine, looked at all the boxes on the shelf.

“Cool.” I thought. “I helped someone out.”

After a few minutes, he shook his head, tucked the magazine under his arm and bought the cheapest card on the rack.

They ask someone for advice, same as ever. Regular consumers have never understood technology – 486 vs. Pentium might seem obvious to you, but not to randoms.

But yes, it’s difficult to figure CPUs out these days – they’re now like the GPU market, where there’s no single marketing number you can look at to see what’s better.

Well OF COURSE E means dual.

Everyone knows that.

It’s because, umm, E is the second vowel in the alphabet.

Once upon a time you could tell which Intel CPUs were the good ones because of the Blue Man Group commercials, or something like that.

I worked at Intel doing processor marketing during the 80s until 2000, and communicating to consumers which processor was “good”, “better”, “best”, was always challenging. At this point I think is safe to say that is pretty much incomprehensible even to former chip heads like myself. In discussion, with other Intel alumni we’ve shaken our heads over the overly confusing naming scheme.

The good news is that it appears that at least some of my former Intel colleagues have figured this out.

In the meantime, I think we are back to old days where you talked your “local” PC guru to figure out what CPU to buy. The good news is Mr Case clearly qualifies. :)

While I agree it’s more confusing than it was (because MHZ no longer dominates all) the above is a little disingenious.

“Core i7 965 3.2Ghz” isn’t 4 unique identifiers. “Core i7” is a single term that denotes the processor series (e.g., pentium or pentium D). The 965 and the 3.2GHZ are redundant. All 965’s are 3.2 Ghz.

So, this isn’t ridiculously different than saying “pentium D 2.66Ghz”. In fact, if you say “i7 965” that tells you everything.

Probably the biggest difference is Intel releasing new processor families at a faster clip than before (i3, i5 and i7 all in a short span) which includes variations like cache sizes, etc., that you didn’t see in the early chips.

No no, see, it’s because the capital letter “E” looks like a picture of two little cores sitting next to each other… see?

Except not really, because how does a Core i7 860 compare to a Core i7 920? Well, the 860 is a little better, but that’s nonsensical by the numbering. Likewise, go on and explain how the Core 2 Duo 6600, 8200, and 9300 relate to each other in a way that makes sense of their model numbers.

It’s actually a sideways pitchfork, like the devil carries, and devil starts with D which stands for dual.

I was mainly responding to the characterization of the naming issue (4 identifiers are not needed to figure out the processor). My post opened with recognition that comparisons are more difficult in the post-MHZ only era.

On actually making comparisons, like the video card guys, intel has gone with the first digit (“8” or “9”) describing a subfamily, and the later digits describing the part’s “rank” in that family (“x60” or “x20”). Hence, a 860 being better than a 920.

Obviously its a lot more complicated than before, but I’m frankly not sure if there’s a more straightforward naming structure possible, once you have a larger array of parts than what chipmakers use to put out.