Slashdot Mirror
10 Years of Intel Processors Compared
jjslash writes to Techspot's interesting look back at the evolution of Intel CPUs since the original Core 2 Duo E6600 and Core 2 Quad processors were introduced. The test pits the eight-year-old CPUs against their successors in the Nehalem, Sandy Bridge and Haswell families, including today's Celeron and Pentium parts which fare comparably well. A great reference just days before Intel's new Skylake processor debuts.
I wish they compared early PIII Katmai or Coppermines to the Duo 6600. Because what the pessimist in me is seeing, isn't a cherrypicked 11 x increase in one bench but overall core performance stagnation.
I was hoping to see an article discussing the changes in architecture and how the improvements have been made, not just regurgiting lists of bench marks
Not only that, but most people talk about "speed" when they use their computer mainly as a terminal for the WWW. They would probably receive more benefit and apparent "speed" from upgrading their internet speed package instead of buying a faster machine.
Seven puppies were harmed during the making of this post.
It wouldn't be a meaningful test then, would it? Hey let's race these cars, but since the old ones don't have turbos, let's disable the turbos in the new ones too.
"So we compared 8 years of Intel processors....."
sigh
That's not just about the pace of change, it's about how well the machines handled the typical workloads of the day. In 1995, mainstream PCs struggled with the typical workloads of the day; my first laptop was a 100MHz Pentium with 8MB of RAM bought around then, and it was slow just booting up, getting online, word processing etc, even after I upgraded it to a massive 24MB of RAM. So any real-terms performance increase made a huge impact in terms of your day-to-day experience using the machines.
Whereas in recent years, even a low-end machine isn't taxed by basic workloads, so the performance improvments don't really have much impact on your day-to-day experience. I can go from my 16GB i7 work laptop to my 4GB i3 ultrabook, and I'll only really notice a performance difference if I wanted to do something like encoding video or whatever.
Because a lot of programs don't use those features because they are compiled to run on a wide variety of hardware.
Ancient? I've got a Pentium D 805 running as a general family use machine and a Pentium III running as a file server. I just got rid of a bunch of Pentium (as in the original) boards year. I have an 8080 I pulled from an old board, framed, hanging in my workshop. Now get off my lawn.
Well crappy Javascript code (as is prevalent on the web today) is pretty power hungry, but it's terrible everywhere anyway.
Benchmarks are hard for comparing computing systems already. Design trade-offs are made all the time. As the nature of the software these systems run change over the time, so does the processor design changes to meet these changes. With more software taking advantage of the GPU there may be less effort in making your CPU handle floating points faster, so you can focus more on making integer math faster, or better threading...
2005 compared to 2015...
2005 - Desktop computing was King! Every user needed a Desktop/Laptop computer for basic computing needs.
2015 - Desktop is for business. Mobile system Smart Phones/Tablets are used for basic computing needs, the Desktop is reserved for more serious work.
2005 - Beginning of buzzword "Web 2.0" or the acceptance of JavaScript in browsers. Shortly before that most pages had nearly no JavaScript in they pages, if they were it was more for being a toy, at best data validation in a form. CSS features were also used in a very basic form. Browsers were still having problems with following the standards.
2015 - "Web 2.0" is so ingrained that we don't call it that anymore. Browsers have more or less settled down and started following the open standards, And JavaScript powers a good portion of the pages Display. the the N Tier type of environment it has became a top level User Interface Tier. Even with all the Slashdot upgrade hate. Most of us barely remember. clicking the Reply link, having to load a new page to enter in your text. And then the page would reload when you are done.
2005 - 32 bit was still the norm. Most software was in 32 bit, and you still needed compatibility for 16 bit apps.
2015 - 64 bit is finally here. There is legacy support for 32 bit, but finally 16bit is out.
These changes in how computing is used over the time, means processor design has to reweigh its tradeoffs it choose in previous systems, and move things around. Overall things are getting faster, but any one feature may not see an improvement or it may even regress.
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
You've never used the "new and improved" Google Maps interface have you?
It's just like the old one, but 10x slower.
Right with you on a C2Q-6600 that's been running at 3ghz since the day it was first booted. You and I also share the same reason for upgrading - virtualization; although mine isn't as much for performance as it is for memory. I can only buy 8GB of ram for my current machine (DDR2) and would like more.
Karnal
What strikes me the most is that today's processors are barely any faster than the 2011 processors.
4 years and only a small speed increase in real performance - 4% for games!!!!! FOUR PERCENT over 4 years. Time to ditch silicon and to start using materials that support higher clock speeds.
Waterfox - a Firefox fork with legacy extension support, security updates and better privacy by default.
If you buy a highend Haswell CPU, they're faster than the first generation by enough to make it matter. Anything newer and you're actually slowing down. Intel's new strategy is to make slower CPUs with faster graphics and lower power consumption.
One thing I found when speccing a new computer was single thread performance. Properly designed intense workloads will multithread (even video encoding), but for day to day use, it's usually one thread that's bogging down the system. AMD in particular pushes for multithread performance at the expense of single thread performance. For Intel, i7 Haswell chips do great at multithreading, but only slightly better at single threaded performance than i5 Haswells, for substantially more cost. Since I don't game, I went with a higher end i5.
And how many of those programs are computing intensive? And how many of the computing intensive ones don't use libraries for some or all of the intensive parts? A lot of of software that is cpu bound got the message years ago that there are newer technologies and it is worth having more than one code path to take advantage of newer cpus or gpus or using a library that does that for you.
I have an 8008 in a piece of equipment I could go in the other room and turn on. And 3 or 4 full tubes of Harris (Intersil clone) 6100s if I want to run the PDP-8 instruction set. My Kaypro has an 8080 in it, and isn't just a gutted part either.
You're welcome to mow my lawn if you're just going to stand there.
I was surprised by how 'flat' the performance increases are as well. If you interpret Moore's Law as expecting a doubling of performance ever 18 months (yes I know this wasn't what Moore's Law originally said, but for long time the transistor count on a die and performance ran hand in hand), 8 years gives time for 5.333 doublings. 2^5.333 is about 40.3. That is a big difference from the 'up to 11 times faster' results these benchmarks produced. If you're more lenient and allow for 2 years between expected performance doublings, the current processors should still be 16 times faster than those from 8 years earlier.
RETURN without GOSUB in line 1050
Not on the AMD side as far as "bang for the buck" is concerned. In 2007 I paid $140 for a quad core CPU and last month I paid $132 for an octocore. Blows through transcodes like a boss and at $362 for the processor AND a gamer board AND 16GB of RAM? Its still got the bang for the buck like they did in 07.
ACs don't waste your time replying, your posts are never seen by me.
A much better comparison would have been if they'd compared the same CPUs at the same frequency so that IPC gains could be immediately spotted. Also I've never understood the point of all-in-one benchmarks like PCMark which measure everything and nothing because various PCs with wildly different CPUs/GPUs/RAM configurations have very similar results.
Because what the pessimist in me is seeing, isn't a cherrypicked 11 x increase in one bench but overall core performance stagnation.
Well, you can't say you weren't warned; there have been about a zillion articles along the lines of "everybody better learn how to multithread, because we've hit the wall on single-core performance and the only way to make use of extra transistors now is to add more cores".
I don't care if it's 90,000 hectares. That lake was not my doing.
Ancient?! I just went from a ca. 2006 2Ghz Core2Duo to a 1.7Ghz i3 a few months back. I don't game or do anything particularly CPU-intensive, so I wasn't expecting big changes, but DAMN! I had no idea. I think the drastically improved memory bandwidth really shows, particularly in Chrome. These are both 4GB machines with an SSD too. Anyway, the i3 was really excellent bang/buck. No regrets.