Please register or login. There are 0 registered and 1264 anonymous users currently online. Current bandwidth usage: 326.30 kbit/s December 16 - 09:08pm EST 
Hardware Analysis
      
Forums Product Prices
  Contents 
 
 

  Latest Topics 
 

More >>
 

    
 
 

  You Are Here: 
 
/ Forums / Intel's Core i7, is it ever enough?
 

  Everyone likes more speed 
 
 Author 
 Date Written 
 Tools 
Continue Reading on Page: 1, 2, 3, 4, 5, Next >>
elliott trevino Jan 10, 2009, 01:03am EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
I agree with you in that point with the economy only the truly able will upgrade to an i7 the prices are terribly high. I have been pushing the AMD/ATI 780g Chipset with a midrange X2 this combo really helps the overall performance of the whole system. I am a system builder and most of those people I get are on a serious budget many are just now upgrading from AthlonXP socket 462 & Pentium 4 Northwoods, yes most of thier pc's are dinosaurs but they were happy, however now with more and more online entertainment from Hulu.com to WarRock they are wanting more options and better speed. You can call them late to the table however this works to my advantage as the 780 & 790 chipsets from AMD can both be upgraded to Phenom AM2+ or AM3 boards in this way AMD has created an upgrade path that at the least envolves a processor upgrade and at best inlcudes a ram update to 1066mhz and soon the DDR3 will be available from what I have read on MSI, ASUS, BIOSTAR & ECS they plan to offer bios updates to enable better upgrade paths on existing motherboards granted some won't run the newer ram due to socket limitations but a good number of the processors will work. This path for me makes alot of sense and to my customers. While these boards are not Awesome they are very very good and provide more than adequate performance, if everyone really wanted a faster processor the Celeron & Sempron would never have come to any motherboard. I also realize that doesn't characterize 99% of the posters to this site, but the average person is just learning how to burn a dvd once they get to doing that is when I see them wanting an upgrade. Not to open a program faster but to rip, encode & burn video, edit, make a home movie, watch hd video content online, play many of the new free mmorpg's hitting the web and pc gaming is seriously making a grass roots comeback. I pick up games at thrift stores, garage sales, second hand stores and have been making a very nice extra profit while converting the consoled masses to pc gaming with the Athlon X2 processors.
My high end customers have been very happy with the Phenom though I admit it is not anywhere near where I had hoped it would be when it debuted my buyers placed second and third orders especially this past christmas as gifts to thier family. Most cite the affordable quad-core experience and again on the mentioned chipsets they perform quite well with the right memory and when needed the right ATI video card in hybrid Crossfire mode.

Want to enjoy less advertisements and more features? Click here to become a Hardware Analysis registered user.
Radomir Jordanovic Jan 10, 2009, 09:17am EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
You would need to get a new motherboard to run DDR3.

Silverstone 750W, Asus P5KC, C2 Q6600, 2GB OCZ DDR3 1600
ATI HD 4850 512MB, 1TB Caviar. 1TB Seagate
Tim Magraw May 03, 2009, 02:27am EDT Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed

u64 Jan 08, 2010, 03:49am EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> The need for speed is dying
There comes a time for each tech detail to reach the final 'good enough' point.
Remember the awsome jump from in from 16bit to 32bit color. After that
no one complained why we didnt get 64bit colors. Because the human eyes and
brain cant even see all the 32bit colors. Back in the day, overclocking a
300MHz CPU to 400MHz made a huge noticable difference. Nowadays, going from
3GHz to 3,1GHz feels the same. Even going from 3GHz to 6GHz has very little
difference because throwing twice the Hz at something dont make it go twice
as fast. (Blind-test before you critizise :P)

I continues. DualCore isnt twice as fast SingleCore. 64bit dont make anything
go twice as fast. Modern SLI dont deliver twice the fps.

We dont need faster CPUs, we need cheaper CPUs.

AMD gives the best bang for the buck. That matters the most to me.

And we need lower idle Watts, to improve the electric bills.

I dont buy computers anymore, i just use parts that others toss. Saves
money that is better spent elsewhere.

P4 1700 256k SSE2 256M-800MHz Geforce3 200/460 128bit 64M SystemIdle:99W 0

100G 8M
D: FAT16 64k Pagefile Evt
C: NTFS 4k 32bit TweakedXP3Pro (boots in 16s)

80G 8M
E: FAT16 64k Pagefile Evt
F: NTFS 4k
dark41 Jan 08, 2010, 12:22pm EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
We've had a rush for new systems since xmas, all i5 750s. I can build them with 2x2GB DDR3 1333 and 320GB HDD for a whopping $20 more than an E8500 and 2x2GB DDR2 (both higher end motherboards with E-SATA, RAID, and Firewire). Me thinks there's more demand for fast systems than some people care to admit. And we're still not at the point where we have enough speed, as nothing is instant.

Dual core alone isn't twice as fast as single core. But dual cores with new architecture are twice as fast as single cores of the latest architecture (Pent 4 and AMD 4000+), and then some, at everything. EG: Each individual core of the new i7s and i5 750 is about twice as fast as a single core on the Pent D 960. Multiply X 4 the speed of a both cores of the Pent D 960 and you're about up to a i5 750.

So please don't give false information to people considering upgrading. If you're happy with your dinosaur, power to you. To most of us in this day and age, time is money. The less time we waste waiting for our computer to start, or programs to open, or DVDs to encode, etc., is the more time we have for other things.

It really comes down to your usage. But anyone who thinks decoding a DVD is remotely close to the speed of a new quad core CPU is kidding themselves. They are more than twice as fast as the fastest Pent D for anything that'll utilize the quad cores. As far as opening a web page or creating a Word .doc, of course the difference would be negligible.

For some of us there is a huge payback to having a faster system, and the components for one fast enough for me don't exist yet. :)

EX38-DS5
E8500@4.0GHz (445x9, 1.40v) TRUE Black
Corsair HX620W
2x2gb Kingston HyperX 9600
HIS IceQ4 HD4850
2X1TB F1s (RAID 0) XP Pro/Win7 Ult 64
Auzen X-Fi Prelude 7.1
Cambridge Soundworks 500w 5.1
G5, Antec 1200
Suspended User Jan 08, 2010, 12:30pm EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
u64 said:
Even going from 3GHz to 6GHz has very little
difference because throwing twice the Hz at something dont make it go twice
as fast. (Blind-test before you critizise :P)

I continues. DualCore isnt twice as fast SingleCore.


AMD gives the best bang for the buck.


I am certain going fomr 3GHZ to 6GHZ will make a huge difference, sorry that is wrong.

A Core 2 Duo is twice as fast as a single core Pentium 2. Please, be specific, that is a horrible statement to make.

AMD does not give best bang for buck.

Troy Holley Jan 08, 2010, 01:19pm EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
So what kind of budget (PRICE RANGE) are we talking about with those AMD systems?

And how about a graphics card along with those?

Do you use a board with integrated sound card/network card/etc?

u64 Jan 09, 2010, 09:12am EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> The Need for speed is dying
Btw, you didnt blind-test, did ya :D

I never compared C2D to P2. I'm not THAT silly :P
I say, a DualCore P2 wouldnt be twice as fast as a SingleCore P2.
And a DualCore C2D isnt twice as fast as a SingleCore C2D. At least i havent seen
any bechmarks suggesting otherwise. It dissapoint me because i would like them
to be twice as fast.

Going from 3GHz to 6GHz never cuts the job time in half. My point is that for each
Hz added we get lesser and lesser performance out of each Hz. Again, show me a
benchmark that say otherwise and i'm very happy to be proven wrong.

Yes, not ALL AMD give more bang for the buck. Just most of them.
Pick a random gaming benchmark and do the math, fps/buck.

In short, spending twice the money never give us twice the bang.


P4 1700 256k SSE2 256M-800MHz Geforce3 200/460 128bit 64M SystemIdle:99W 0

100G 8M
D: FAT16 64k Pagefile Evt
C: NTFS 4k 32bit TweakedXP3Pro (boots in 16s)

80G 8M
E: FAT16 64k Pagefile Evt
F: NTFS 4k
Radomir Jordanovic Jan 09, 2010, 05:52pm EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
It used to be that we would just go up in clock speed. Then when it was seen people would like multiple cores, the clock speeds went down since they weren't necessary. Intel didn't release the 4GHz P4 because AMD would continue trashing Intel in its 2.XGHz range anyway.

A clock speed is simply how many times per second a cycle happens. A 2GHz CPU has 2 billion clock cycles per second. But what does that even mean to you? You've seen you can't compare these like apples to apples. How about this: What's a clock cycle? How many processes per cycle? Then, how do cores access their memory? A CPU is nothing without access to memory, and how fast is the access, how much memory is there, etc?

The reason these questions didn't come up about 5 years ago is that they just weren't important. Almost all the post-Willamette P4 CPUs were built identically, and even when AMD made CPUs that were faster at lower Hz ratings, you could still look at purely the clock speed if you stayed in the same brand. However, these questions were still applicable even back then and answering them will help you figure out whether a 3GHz CPU will be slower than a 6GHz CPU of the same type. That, and just thinking about it in general.

Silverstone 750W, Asus P5KC, C2 Q6600, 2GB OCZ DDR3 1600
ATI HD 4850 512MB, 1TB Caviar. 1TB Seagate
dark41 Jan 09, 2010, 08:46pm EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
I still don't see AMD as being better bang for the buck, in any model. Maybe 2 months ago for a few cheaper models as Intel had nothing to compare to (Celeron has always sucked), but that was before the new i5 and i3.

The Core 2 Duo prices are now dropping and scheduled to drop more, becoming Intel's "low end" CPU (replacing Celeron), which will blow away anything low end AMD has. (I can currently get an E8500 or Q8500 for $210AUD, if that drops $50AUD it seriously buries anything I can get from AMD at all but much higher price points.)

Socket 1156 is where its at now for low-medium-medium/high, and nothing AMD currently has will touch i7.

So I'm not sure how anyone could think AMD is better bang for the buck now at any level. But admittedly, I'm a much bigger fan of Intel than AMD, always have been, and probably always will be. I'm sure there's bigger fans of AMD than Intel out there somewhere too. :-)

EX38-DS5
E8500@4.0GHz (445x9, 1.40v) TRUE Black
Corsair HX620W
2x2gb Kingston HyperX 9600
HIS IceQ4 HD4850
2X1TB F1s (RAID 0) XP Pro/Win7 Ult 64
Auzen X-Fi Prelude 7.1
Cambridge Soundworks 500w 5.1
G5, Antec 1200
u64 Jan 10, 2010, 03:59am EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Jan2010, AMD still wins BangForTheBuck
I admit, on the second-hand market i got no clue if Intel give more bang for the buck.

For new CPUs, i've compared AMD vs Intel bang for the buck continiously since i bought
my K6-2 350. AMD gives by far the better bang/buck, even more so when including costs
for the mobos, which costs more on the Intel side.

I'm a AMD fan, currently. Also because i want more competition. Currently AMD is
near-bankrupt or something while Intel has 90% market or such. That's an unfair place
for me, the consumer. So i favor AMD whenever i find it makes sense.
I hope ARM succeed someday.
I was very interested in Transmeta back in the day, Transmeta had technology to blow
Intel and AMD out of the water. But as i turned out, Intel had too deep pockets and
helped to kill Transmeta. And we never had the chance to enjoy dirt-cheap 1W 256bit CPUs.
Those kind of things make me truly dislike unfair market.
(I'm not a fan of markets at all but i'm keeping it short, and pragmatic)

http://www.tomshardware.com/charts/2009-desktop-cpu-charts/GTA...,1402.html
GTA4 + LeftForDead (fps) / $$$$ = BangForTheBuck

Athlon2 X2 250 (36.9+111) / 67.00 = 2.207BangForTheBuck
Phenom X4 9650 (48.8+102.1) / 104.99 = 1.437
Phenom2 X4 965 (62.1+156) / 184.99 = 1.179
C2D E8400 (45.8+129.8) / 190.00 = 0.924
i7-975E (62.4+198.8) / 999.99 = 0.261

I picked these at pseudo-random, already there's a 100%-pattern. Case closed.


P4 1700 256k SSE2 256M-800MHz Geforce3 200/460 128bit 64M SystemIdle:99W 0

100G 8M
D: FAT16 64k Pagefile Evt
C: NTFS 4k 32bit TweakedXP3Pro (boots in 16s)

80G 8M
E: FAT16 64k Pagefile Evt
F: NTFS 4k
dark41 Jan 10, 2010, 09:41am EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List

Edited: Jan 10, 2010, 09:49am EST

 
>> Re: Everyone likes more speed
u64,
The only thing closed here is your mind.

This article and the consequent posts are about speed. You've made it abundantly clear that speed isn't important to you, which is fine for you. I'd guess you represent a very small percentage of PC users in that regard.

Fine by me if you want to turn it into a "bang for the buck" debate, but at least get your facts straight.

You chose the site, so I'll stick with TG's CPU comparison charts. I don't believe for a second that you chose a random benchmark, as there are very few that show the AMD beating the i5 750 at anything. Even the example you chose (0.3FPS difference, 61.1 vs 60.8) could be considered a draw. Ironically they don't list a price for the i5 750, and they don't have the i5 and i3 CPUs either. That's not exactly realistic for either Intel or AMD fans today.

Then there's the simple fact that there is no way to measure bang for the buck for everyone. Bang for the buck has to be measured by individuals, according to their own usage. Any rating system places value on what that particular author sees as important, which is strictly opinion rather than fact. I value my time more than a few lousy FPS in a game, so for me this is the true picture of the bang for the buck with the i5 750 vs the Phenom II x4 965:

EG: The i5 750 costs the same amount as the Phenom II x4 965 on Newegg right now (both $199.99USD).

I'll use your own link since your "random" choice gives decent comparisons.

The 750 beats the 965 by 21 secs on the Adobe 9 Professional bench, 91s vs 112s. Thus, if someone uses Adobe 9 Professional all day long, that's a serious reason to go with the Intel rather than the AMD. In other words, for every 5 documents the AMD user does, the Intel user does 6.

Converting music is your thing? The i5 750 converts iTunes to ACC 35 secs faster than the x4 965 (98x vs 133s). For every 3 songs the AMD user converts, the Intel user converts 4.

The i5 750 beats the Phenom II 965 by 39s on lame (142s vs 181s). For every 3 conversions on the AMD system, the Intel system does 4.

The list goes on and on.

In fact, you'll struggle to find a benchmark that the AMD Phenom II 965 can even compete with the Intel i5 750. Other than the one you pionted out, I see that the 965 beats the 750 by a whopping 1 sec on an AVG scan. Again, call it a draw.

And you think you have a case for the AMD being better value/bang for the buck now? Think again. As I see it, for the same price, I get roughly 20% better all around performance with the i5.

These benchmarks make a serious case that any bang for the buck comparison would heavily favour the Intel in this price range now. For people who can be doing other things rather than waiting for a CPU to finish, its impossible to put a price tag on the time saved by the Intels in these benchmarks. For me, it would be like 25% of the score. That would give the i5 a 45% advantage for me over the Phenom II at this price point. Obviously the comparison you chose to show doesn't see it the same way as I do.

For those who care more about saving a few dollars than saving time, AMD is still a valid option. The low end market is appealing to those who want the cheapest solution possible, and the cheapest AMDs are still a better option than the current Celeron offerings. But once you get to mid range, AMD is no longer competitive, and AMD offers absolutely nothing to compete with Intel's high end solutions (i7).

I won't waste much space talking about motherboards, as your statement about the prices of them was just as outdated as the bang for the buck comparisons you used. But suffice to say that any motherboard that has the features I want (E-SATA, RAID, Firewire 800, 5.1 HD sound, USB 3.0, SATA3, etc.) will cost around $150-$200USD, whether its for AMD or Intel. :)

EX38-DS5
E8500@4.0GHz (445x9, 1.40v) TRUE Black
Corsair HX620W
2x2gb Kingston HyperX 9600
HIS IceQ4 HD4850
2X1TB F1s (RAID 0) XP Pro/Win7 Ult 64
Auzen X-Fi Prelude 7.1
Cambridge Soundworks 500w 5.1
G5, Antec 1200
Robert Eachus Jan 10, 2010, 12:08pm EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
The speed that matters to you is the speed that matters to you. Seems almost a waste of electrons to say that, but it is important to remember. There is what economists call a "virtuous cycle", in which different sets of customers buy the products that meet their needs, and the manufacturer of that product designs the next version for their customer base, and product differentiation results.

I'll stay away from the current allegations that Intel's compiler releases have intentionally tried to make the resulting code run slower on competitors products. However, there is a selection effect that is very powerful: Intel compiler developers do their development on Intel hardware. If a new compiler feature seems to run faster, the developers will include it. Later in the development cycle the new code will be tested on all sorts of hardware, and if it causes a serious decrease in performance, it goes back to the developers.*

Why explain all this? Different groups of users use different programming languages, and as a result use different compilers. I tend to use GCC, not because it is faster than the Intel C compiler or other C compilers. (It isn't) But the front-end language selection better meets my needs. The big difference for me is that GCC has an Ada front-end (GNAT), while the Intel compiler does not.

I also write programs that run for hours, or days. Some of these programs are doing things that were prohibitively expensive when I first thought about them ten or 20 years ago. Now I can run them on my desktop. Writing such code is all about data movement. I don't care which modern CPU you use, you can do more number crunching in a clock cycle than you can do data movement to and from L1 caches. Moving data to or from main memory is about 100 times slower, with L2 somewhere in the middle. If you want to multiply two matrices that don't fit in L1 cache, you end up tuning the code to a particular set of cache sizes.

Having said all that, right now if you need to do trigonometric, operations, think discrete cosine transform used a lot in signal processing, an Intel CPU will win the race. For straight linear algebra, the 64k L1 caches on an AMD CPU are the big win.

The problem though, and it is a problem, is that you need to optimize the code for a particular system. Not just cache sizes, memory timings, main memory sizes, etc., all affect processing speed. So you need a good BLAS (basic linear algebra subroutines) version tuned to your system. Once you have that, the effects of the compiler on throughput are marginal at best.

Now to go back to hardware after having said all that. For the average application, tuning the code can result in a factor of 100 difference in speed. So I usually look at (hardware) benchmark suites and laugh. Hardware sites are choosing test programs with expected run-times in minutes. (And this includes SPEC by the way.) These benchmarks are very useful--if you run that application. If you run hardware optimized code that does the same thing several times faster, the bottlenecks move. So what was a "good" CPU benchmark is suddenly a good memory benchmark, or even a disk I/O benchmark,

By the way, this level of extreme attention to performance is just not worth it for most applications. But simulating galaxy evolution? You go through the effort to insure that square roots are done inline, by the proper instruction, and that the evaluation of sqrt(a^2+b^2+c^2) involves no function call overhead, or even a check that the sum of squares is positive. Getting a factor of ten or even 68 ;-) in the time to evaluate one time step may result. Then you spend most of that gain improving the accuracy of the simulation. :-(

There are lots of applications like that where the code developed to make the evaluation possible on one generation of supercomputers turns up on desktops a few years later. The simulation code used to improve the speed of automobile crash testing on computers is now starting to show up in games as "physics effects", I don't expect games to do a good Navier-Stokes simulation of hydrodynamic effects when wading across a river. But it would be nice if they got to the point where wading downstream was easier than wading upstream. This is starting to happen...

But the final nail in the coffin of highly optimized code for CPUs is the current move to using GPUs for HPC (high-performance computing). This has been waiting on decent double-precision floating-point support on video cards. It also requires support for integration of GPU use into programs written to run on the CPU. There is a rule of thumb that 10% of the code takes 90% of the time. It is that 10% you want to move to the GPU. Remember what I said above about data bandwidth driving high performance code? GPUs have much higher bandwidth to graphics memory than CPUs do to main memory. Modern video cards now have what amounts to an explicitly programmer managed L1 cache, and a general purpose shared (with other shader processors) L2 cache. SInce you have hundreds of SPs available, total SP to L1 or L2 bandwidth is one or more orders of magnitude higher than CPU core to CPU L1 or L2 bandwidth.

Windows 7 has introduced DirectCompute, but I expect OpenCL not OpenGL) to become the preferred API for GPU computing before the end of the year. Of course, this will be a very fuzzy line, since nVidia's CUDA and ATi/AMD's Stream are evolving toward being proprietary development environments, but using the OpenCL API. (Not that difficult, since OpenCL can be thought of as a standard interface to CUDA and Stream technology.)

* This is called regression testing. Yes, the solution may be to turn off that feature in generic code, and code targeted at specific hardware. But much more usual is to look at a trace of the code on a processor where it runs slowly, and see a bug in the code. A bug in this context can be anything that makes the code run slower.

My favorite horrible example of this was early Dhrystone versions ran at two very different speeds on 68020 based Amigas. The bug was not in the compiler at all, but in the loader. The code it generated correctly checked the CPU type and did stack alignment on two or four byte boundaries. But it also put the executable name on the stack. It did this at load time. You could rename executables and the name pushed was the current name.. However, this name was only two byte aligned. Global variables got pushed onto the stack next. So the first program that came along that extensively used global variables, as opposed to variables in the stack frame of the main program, exposed the bug.. Dhrystone was the first benchmark written in Ada, and although the widely used version is a port to C, it still made extensive use of global variables (which were in packages in the original Ada).

In any case, regression testing is important in compiler development, and every compiler group has a set of regression tests they run constantly. It may be that many of the tests in the suite check for bugs long fixed to make sure that they don't creep back into the code. But a lot of the programs tend to focus on a particular instruction set feature, such a matrix operations on complex numbers.

Meats_Of_Evil Jan 10, 2010, 12:33pm EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
u64 said:
Going from 3GHz to 6GHz never cuts the job time in half. My point is that for each
Hz added we get lesser and lesser performance out of each Hz. Again, show me a
benchmark that say otherwise and i'm very happy to be proven wrong.



It's all about the hard drives baby! I'm baffled at how slow hard drive progress has been. We could get an octo core by next year and SSD's will still wouldn't be the norm.

-------------------------------------------------------------------------------------------------------------------------
Everything I write is Sarcasm.
Radomir Jordanovic Jan 10, 2010, 01:27pm EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
Robert, that explanation was absolutely amazing! Thanks a whole lot!

Silverstone 750W, Asus P5KC, C2 Q6600, 2GB OCZ DDR3 1600
ATI HD 4850 512MB, 1TB Caviar. 1TB Seagate
Sean Costello Jan 10, 2010, 07:48pm EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
Why on gods green earth would you type that much D:




______
Story of My Life- http://www.youtube.com/watch?v=c764JWVt5Fw
Radomir Jordanovic Jan 10, 2010, 08:32pm EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
It's for a good cause! I read it all, and I enjoyed it!

Silverstone 750W, Asus P5KC, C2 Q6600, 2GB OCZ DDR3 1600
ATI HD 4850 512MB, 1TB Caviar. 1TB Seagate
dark41 Jan 10, 2010, 08:59pm EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
I thought I covered all that by simply saying bang for the buck is different according to each individual's usage. :~

I reckon less than 1% of the people in the world would understand any of that explanation, but I do appreciate the effort. :)

EX38-DS5
E8500@4.0GHz (445x9, 1.40v) TRUE Black
Corsair HX620W
2x2gb Kingston HyperX 9600
HIS IceQ4 HD4850
2X1TB F1s (RAID 0) XP Pro/Win7 Ult 64
Auzen X-Fi Prelude 7.1
Cambridge Soundworks 500w 5.1
G5, Antec 1200
Robert Eachus Jan 10, 2010, 10:44pm EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
Why on gods green earth would you type that much?

Because the entire computer industry is built on an incredibly small base, and there is a problem. Running out of compiler experts would limit innovation in computers. It doesn't matter if the new computer has the same instruction set as an older model. Changes in computer architecture change which code should be generated for a particular operation. An easy example is that a compiler can use three XOR operations to swap two operands. Is that faster than generating loads and stores? Usually.

I grew up with computers. Yes, i had a home computer as a child, well two or three. But they were a Burroughs E100 and an analog computer* my father designed and built himself. While I was in high school, I took two summer programs, one at Temple University with an IBM 650, one at Univ. of Pennsylvania with a Litton LGP-30, and a Royal McBee RPG-4000. While I was at Temple, the first compiler came out for Fortran. The Penn program concentrated on compiler design, but I also did some OS work on the RPG-4000, which didn't have an OS when I started. (It really didn't until later that year, but I contributed to it. But I had to write in machine language, since there was no compiler.)

After I got my Master's at RPI I took a course in Compiler Construction and wangled an invitation to a seminar on software engineering held jointly by RPI and General Electric. The conference was dominated by many of the original compiler development experts, and the word had just come down from the DoD that Green was Ada. Green looked like a better language for my thesis than APL*, but there were a few things that needed to be fixed. About 12 pages worth. ;-)

Within a few months I was working at Honeywell on a compiler that had started out to be a compiler for Green, but became an Ada subset compiler. That had to wait for the ANSI standard in 1983. By that time I was a compiler expert and Ada language lawyer. There were some things in Green, Preliminary Ada, and even the first Mil. Std. Ada that were impossible to implement correctly. Some were obvious, unintentional, editing bugs, which I could help fix--I wrote over 50 pages of comments on one draft of the ANSI standard, most of which were trivial issues. But some were very deep. I showed a simple program that compilers were supposed to accept as correct if, and only if, Fermat's Last Theorem was true. Now that is not a design bug, but I could have changed it to the Reimann Hypothesis. (Oops! That may have been proved now. ;-) The obvious solution to the problem, was not requiring compilers to find a legal elaboration order if one existed. But allow them to raise an error at run-time if the order they chose caused access before elaboration problems. Some people picked on this as a "problem" with Ada. But they were not Ada programmers--or experts in Computability Theory. It is really something that Godel proved almost a century ago. It is possible to embed deep Number Theory problems in "simple" equations. Any program that doesn't read input before elaboration (setting up storage) is completed, won't have ABE problems. But sometimes it is useful to make the storage allocations depend on something read in at run-time, like the size to make an array. The "issue" in Ada was that allowing arbitrary (integer) expressions to change the order in which storage was allocated could be turned into a theorem prover. ;-) The expressions to be evaluated could be pretty bizarre though, and unlikely to occur in any real program. If they do? No biggie. Instead of getting an error message at compile time, you get it the first time you try to run the program. (Or the program runs forever trying to find a counter-example to FLT.)

Anyway, the number of expert compiler designers has never been large, and is shrinking. Getting even one person with the mental flexibility to join the group interested is worth a lot of effort. What is required? When I was teaching Compiler Construction, I would start the class by saying that if you can think in two languages at once, you will do well in this class. If you can think in three at once, you will enjoy it. ;-) We tended to look for people who were bi-lingual as a children, that helped. (I didn't, though.) Another test was a killer. At one point we were unable to identify a single programming language expert who hadn't spoken English, German, Dutch, or Russian as a child. Jean Ichbiah grew up speaking English and French. I once visited Alsys Headquarters in France. I don;t think I heard a single word of French spoken inside, except by secretaries answering the phone.

I found my limit applied to me in real life when I was on an SAS flight from Copenhagen to Frankfurt. With conversations going on in Swedish, Danish, German, and Norwegian in the group I was travelling with, a stewardess came by and asked: "Coffee, tea, or milk?" I had no idea what she had said. My wife would also complain that if I was working on the parser, or the parser generator, and she called to ask what I wanted for dinner, that I wasn't processing English then either. However, sometimes the pause was just to mentally or physically bookmark my place in a very large program listing. ;-)

Why do you need to process more than one language simultaneously? You are often describing a mapping from one language to another in a third language.. (Sometimes even in English, or whatever, to yourself. ;-) It helps if you can write the compiler in the language it will be compiling, since that sometimes cuts the simultaneous language requirements way down. But you still often need to use many formal languages (such as BNF, regular expressions, or finite state machines) before you are done.

Oh, and yes, I type around 50 wpm, so something like this doesn't take me long. ;-)

* I won't go into what analog computers were. Anyone who knows what an op amp is will know. Other than that I'll say it was put together to test the power supplies for the Univac I, and later used to test power supplies for the radars at Cape Canaveral, and elevator motors and controllers for skyscrapers. Today, we would do all that with (digital) computer simulation.

dark41 Jan 11, 2010, 10:13am EST Reply - Quote - Report Abuse
Private Message - Add to Buddy List  
>> Re: Everyone likes more speed
Ah.. technology has come a long way.

Not sure what qualifies as a first "computer". I started with a 2 x 5 1/4" disk drive system (read disk and write disk, and that's right I'm old). I remember being amazed at how fast and elegant my first PC (100MHz) was, and how p**sed off I was when I could no longer get a new belt for my printer.

Finally slowing down from the post xmas rush for new systems, so I'm building my own new system slowly, (probably going with the i7 860) as my overclocked E8400 seems slow anymore compared to the many i5 750s I've just built. I'm taking my time modding my Antec 1200 with fans and lights, and then will pick out the MB and CPU. But that's just where my interests lie, hardware.

We also have a cap on our uploads/downloads in Australia. Once upon a time downloading a movie was unrealistic, and HDD space was expensive. Now my 2x1TB RAID 0 data partition is about full of movies and music, and its time to upgrade again. Luckily I was able to change from our 40GB peak/40GB offpeak to a cheaper plan of 70GB peak/60GB offpeak. Converting these files to various formats is a big part of my weekend, and anything that does it faster is more time I can spend with my wife and riding my BMW. So for me, computer speed = priceless. :)

EX38-DS5
E8500@4.0GHz (445x9, 1.40v) TRUE Black
Corsair HX620W
2x2gb Kingston HyperX 9600
HIS IceQ4 HD4850
2X1TB F1s (RAID 0) XP Pro/Win7 Ult 64
Auzen X-Fi Prelude 7.1
Cambridge Soundworks 500w 5.1
G5, Antec 1200

Write a Reply >>

Continue Reading on Page: 1, 2, 3, 4, 5, Next >>

 

    
 
 

  Topic Tools 
 
RSS UpdatesRSS Updates
 

  Related Articles 
 
 

  Newsletter 
 
A weekly newsletter featuring an editorial and a roundup of the latest articles, news and other interesting topics.

Please enter your email address below and click Subscribe.