When I was a child I received severely inappropriate treatment at several occasions: codeine drops against allergy induced cough, phenylmercure borate for fever blisters, a tooth inlay implanted in my gums. Do you still wonder why I don't trust medics anymore and prefer to conduct a web search for every diagnosis and medication?
It seems you need to find better doctors. Sometimes what you pay is what you get. You can get top-notch health care in the US, if you can afford it. If you want to do web searches for everything, by all means do. It is your health after all. Just don't be too eager to assume that you can do better than a properly trained doctor (we doctors have discovered the web, too).
It's usually very hard to put a diagnosis without proper data. In this case, even though I'm currently preparing for USMLE-like examinations (you are presented with some data and you try to make a diagnosis), I find it very hard to trust Pat's "clues" because I don't know what is real and what is *his* idea of a diagnosis.
As a doctor I really like to hear my patients tell me actual facts and not their interpretations. E.g. "I have fever and sore throat" and NOT "I have the flu". Infectious diseases that may present with fever and sore throat are many (ranging from primary HIV infection to infectious mononucleosis to common cold) and it's highly unlikely that the patient has considered all of them. By focusing on a possible "diagnosis" the patient may ignore other signs that would be useful to the doctor.
I could list quite a few diagnoses that would fit Pat's description, but guessing is quite useless, especially in important health matters. Maybe some doctors did not follow proper standards of care but the fact that an assumed serious condition did not alarm so many of them is quite suspicious.
As a simple advice (I hope Pat is reading this!): IF you have fever (defined typically as over 38.3 deg. Celsius) plus a NEW audible cardiac wheeze (not mitral prolapse, which is quite different) you should be admitted to the hospital on the basis of an assumed diagnosis of bacterial endocarditis (unless *proven* otherwise). Bacterial endocarditis usually develops on PRE-existing pathological conditions (e.g. old rheumatic fever, IV drug use). Typically, cardiac ultrasound (why don't you go have one, if you are so worried?) will give very useful clues. Examination of the retina and blood cultures (at least three) are also necessary. If no signs of bacterial infection are found, several viral pathogens can cause pericardial inflammation but I can only remember Coxsackie and echoviridae off the top of my head. Viruses usually cause milder disease.
Finally, please do not trust the web, google, medline, nih. These are excellent data sources, but you are unable to properly interpret what you read without proper training. You can't just open "harrison's internal medicine" and hope to acquire the skills to make a diagnosis in a few hours/days/weeks. Find a good doctor and trust him. Sure, some people say that they correctly diagnosed their condition, even though the doctors where wrong. It happens, doctor's make mistakes. But on 99.99% of cases, your doctor knows better than you.
You mean by NOT buying it after you didn't buy it in the first place? I'm sure they'll notice the difference.
NOT!
No, I mean by not buying even though I intended to. I thought that was quite clear. Many people have propably waited a few days to read the reviews and hear other opinions before buying. I'm one of them.
Annoying policies like this one may increase short-term profit but people will, I hope, gradually react. I must say that from all the best-selling FPS epic has showed the best company policies. They released a linux version concurrently (on the same CD!), ut2004 does not require the CD to be in the drive and they keep releasing free files in the ut2004 community (extra maps etc).
Sure, everyone wants to play HL2. I know I do. But I think I'd rather send them a message by NOT buying their game. Would you go to a restaurant where the waiter insults you, even if it is the best one in the town?
Right now, US economy is pretty much sustained by the fact that, if you want oil, you must purchase it in dollars - thus you are buying a small part of the left-pondist's debt each time.
European countries actually use the euro to buy oil. I think most OPEC countries do accept euros. After all, many of them are not very friendly towards the US.
The real reason why USA prefers oil is, I think, that the US has a much stronger control of oil supply than the EU. They can afford to lose money on expensive oil, because the EU loses more. My brother studies this kind of stuff (MSc international politics), so this is my vague impression.
Actually solid-state does render vacuum tubes obsolete, to the rational mind. Once you've admitted that the sound you really like just involves lots of second order distortion it's no big deal to make a processor using opamps or discrete transistors to add that distortion to a reliable, efficient, cheap amplifier. As many manufacturers have done! Boss, Line 6, and Roland to name just 3.
Many people like to think that we know everything there is about music reproduction so we can model almost everything using enough bits and MHz. Sure, you can do a good job in most cases and with a fraction of the cost but guess what:
a) We do not always KNOW why machine 1 sounds better than machine 2, and no THD and response curves are not enough to describe the complex phenomena of musical reproduction. So we can't model this.
b) If a hypothetical machine using silver cables or tubes or pixie dust sounds better then maybe instead of trying to disprove several people that enjoy its sound you could try to find out WHY it sounds better. People are not always victims of the placebo effect.
And let's not consider technology to be the perfect substitute for everything: sure, you can try to imitate a piano using N processors and god knows what else, but I'd prefer a plain (plain in the sense that it does not contain transistors;-)) steinway piano (~100k $) any day of the week. It's a matter of taste. If a musician likes the sound of tubes and he can afford them then let him happily enjoy them.
For the record, the absolute best audio reproduction I have encountered came from Audio Research (www.audioresearch.com, have a look) tube amplifiers.
Seriously, sugar is hardly the problem with obesity in America. The problem is primarily one of poor eating habits, coupled with lack of exercise. (Not that anecdotes prove anything, but just to pull out one random example; I used to know a gal who was a strict vegetarian, and I *never* once saw her eat a piece of candy or "junk food" - yet she was overweight.)
It is certainly true that surplus caloric intake
causes obesity, regardless of macronutrient source. However, all macronutrients are not
processed in the same way. Sugar is a special case for the human body because it quickly increases
blood glucose level (thereby triggering homeostatic
mechanisms--insulin) and equally quickly disappears.
The net result is that, even though your empiric
evidence may point to the contrary, research is
still necessary in order to fully determine whether sugar consumption is linked to obesity.
I could point you to some interesting articles,
but the general idea in the relevant bibliography
is that eating sugary foods does seem to predispose
to obesity. And, to put it in a different light,
sugar does not seem to offer anything
useful (except in cases of hypoglycemic coma!)
besides plain old calories, so you are better off
eating more complex carbohydrates.
The idea that only "intelligent", "educated" etc people should vote is actually wrong. The logic behind voting does not have to do with getting the best possible decisions (best decisions come from think tanks and specialists). The REAL reason for voting is MORAL, and has to do with sharing responsibility. The idea behing democracy is that people should bear some of the burden of power.
After all, if you HAVE to vote you will, even out of curiosity, try to learn something in order to make an intelligent choice. This would really force the political parties to address voters in a completely different manner.
Once more, democracy is a system for distributing *blame* (moral burden of power), not a system for making optimal decisions.
As an owner of a 9700 and a hobbyist developer, I'm very familiar with the limitations. The shader length is highly restricted, conditional branching can't be done, so loops have to be unrolled. For this reason, even the latest ATI cards can't fully support the OpenGL Shading Language. What can be done on an FX or a Geforce 6 in one pass could take 10 or more passes on an X800. Many important features for shadow mapping are hopelessly missing, such as rendering to a depth texture, and hardware linear filtering.
Copying from ATI's web site, the specifications
for RADEON 9800 PRO say:
# SMARTSHADER(TM) 2.1
* Full support for Microsoft® DirectX® 9.0 programmable vertex and pixel shaders in hardware
* 2.0 Vertex Shaders support vertex programs up to 65,280 instructions with flow control
* 2.0 Pixel Shaders support up to 16 textures per rendering pass
* New F-buffer technology supports pixel shader programs with unlimited instructions
* 128-bit, 64-bit & 32-bit per pixel floating point color formats
* Multiple Render Target (MRT) support
* Shadow volume rendering acceleration
* Complete feature set also supported in OpenGL® via extensions
So, it seems that ATI cards do support flow
control (i.e. loops) and infinite pixel shader
length. Shadow volume acceleration might be
somehow limited as you say, but I haven't
really looked into it. Maybe some ATI white
paper will clarify these details.
NOTE that I'm referring to a 9800 product and
not to the high-end X800.
Benchmarking is not about *proving*, it is about *measuring*. Note that Goedel's incompleteness theorem does not necessarily imply that ALL propositions are unprovable! It merely states that there is at least one proposition that is true but cannot be proven (Goedel constructed this proposition in an ingenius manner[1]). It suffices to build ONE such proposition to derive incompleteness. MANY other propositions can be proven, including, possibly the fact that card A is better than card B (if you can call this thing "proof").
Therefore, proof that card A is better than card B may be feasible. However, the idea of proving something like this is ridiculous. What you need is an ordering of the cards according to some way of measurement. This means that for some arbitrary benchmarking procedure P() for card X you get a number P(X) and order all cards according to that. IF P() is relevant to you (e.g. quake 3) the ordering is sound (assuming many other variables, like system, software etc are the same). Nothing esoteric about that.
(by the way, the profound theoretical significance of Goedel's incompleteness theorem is completely opposite to its practical impact, which is almost none in everyday mathematics--most things have nice proofs)
P.
[1] The importance of the theorem lies in the fact that even if you convert an unprovable TRUE proposition into an axiom you will get a new system that will AGAIN be incomplete. You can't get away by building bloated systems.
To quote from memory : "the program will detect discrepancies in a user's thinking and alert the user".
This sound an awful lot like "Clippy" raised to the Nth power. I am certain that MS products will be the first to feature this Smart-Ass computer technology, whereby the computer will constantly correct you and interrupt your thinking with irrelevant bullshit ("are you sure you want to do this? maybe you want to do that").
On the other hand, just like spell-checking helps pepole (sic) write clearly, maybe this will allow the less-than-privileged computer users to follow a logical path of thinking (quite possibly, an eye opening experience for some).
Sometimes I feel that AI is awfully close.
P.
Slowly but surely the UNIX crept upon the Nintendo user.
Recently some of us have observed that CD-R quality has gone down the drain. Too often I buy disks that fail really soon.
The real question for me is: which of the two formats is most reliable in the long term? Will I be able to read my disks 5 years from now? Which is most reliable, DVD-/+R or CD-R?
Capacity and speed is nice but I don't tolerate data loss very well...
Certainly onboard audio is acceptable for many uses (office desktop etc) but a decent sound card can be bought for a ridiculous amount of money so it may be worth it.
I recently bought an Audigy Player (OEM) for roughly 50 Euros and it is great. You get an effects processor, real 5.1, a load of utilities that are quite nice and MUCH better signal quality than most onboard cards for the price of 3 CDs. Many CPU coolers cost more than this and they are just blocks of metal!
You should definitely stick with onboard if you own cheap speakers and don't play a lot of music or many games but in my opinion a decent card is worth the small extra spending and it will last many years. As a matter of fact I also suggest getting decent speakers (like Cambridge Audio 2400). For less than 100-120 Euros/$ you can get very respectable (computer) sound quality.
I had a very noisy computer and I spend some money to improve it. I used the following components:
- Akasa pax.mate sound insulation 30 E - Thermaltake Volcano 9 fan/cooler 35 E - Zalman 400W PSU (EXCELLENT quality!)
at 118 Euro - High quality case with 2x12cm fans (not
the standard 8 cm) at 50 Euro without
PSU
I also stripped the fan from my Northbridge (VIA KT400 chipset) and it's only 3-4 degrees C hotter (~42 deg. C with ~30 deg. C ambient).
I am now considering a passive VGA cooler (Zalman, 27 Euro) for my video card (Geforce 3 Ti 200).
> Did you know you can fertilize your lawn with used motor oil?
OK, this is totally off topic, but you should NOT fertilize your lawn with motor oil. It contains several mutagenic/carcinogenic substances and you don't want them going underground to pollute water/plants etc.
Used motor oil should be properly disposed of in places that later sell it/use it where appropriate (ships can burn it, it has other CLEAN uses).
Regarding the hard problem: it looks a lot like creating a Huffman coding tree.
You need a binary tree where the weight (information = -log propability) influences the depth. Rare things go deeper and need more bits to find. Common things go higher. Therefore if, say, you want to describe something common you know it is at 01 (2 bits instead of 8) while something rare is at 1001111101 (11 bits instead of 8). Bits describe the path on the tree.
This is not particularly hard and you can construct an optimal solution VERY quickly.
Anway, the problem is a bit different (Huffman trees store information only at the leaf nodes and they are not ordered) but I'm sure that the same principles apply.
The general idea is that you progressively built subtrees starting from the less propable (less weight) items, combining them two at a time while preserving ordering[1]. You then treat each subtree as a single item (classical dynamic programming) and see if you can build any further using it.
Many problems in such competitions are solved with dynamic programming.
P.
[1] This is possible because you can always choose between left and right at the same depth. Depth is based on weights and left-right ordering on key values.
P.S. Could be much harder, though. I just read the problem description and I might be saying a load of crap. Sometimes a small change can transform a O(N) problem to an NP-hard problem.
>special interests have corrupted >mainstream benchmarks to make them an > unusuable guide.
This is true, to a certain extent!
People used to benchmark with, say,
Office scripts or other "office" jobs.
Most modern benchmarks are almost
100% multimedia (Internet content
creation, Divx, MP3 photoshop etc
come to mind). This is very convenient
for Intel because P4 is a multimedia
design (long pipes, high MHz, small
cache, fast FSB, SSE2) designed for
serial operations (small loops, no
branches) with huge data sets (fast
memory is good here).
The Athlon is a completely different
design and it cannot compete. I am
really wondering how it performs e.g.
with kernel compiles or other
system scripts.
Anyway, this reminds me of the situation
with K5 and Quake. Until the era of
Quake (and Pentium Classic) all
competing processors (AMD K5, K6,
Cyrix) had very very slow FPUs.
Quake was the first widespread game
to use FPUs (it relied on the FPU to
compute 1/z with fdiv) and even though
the K5/K6 were VERY fast with integer
code the Pentium Classic had vastly
superior performance under Quake
(popular benchmark at the time!).
Anyway, this whole situation is not
necessarily the fault of the benchmark
designers. Frankly, even though the
AMD may be faster at e.g. Excel I don't
intend to use huge multi-megabyte
spreadsheets (most of us don't) but
I do intend to use MP3/MPEG2 etc
algorithms for which (like Internet
Content Creation) the P4 may be
a better choice for those that can
afford it.
Actually I do own an Athlon 2200+
and I am very pleased with its
performance mostly (and most
importantly) because it is CHEAP.
(AMD always has better MIPS/FLOPS per $, that's why I buy AMD)
Some people here have pointed out that this '3d' effects desktop will not cause a performance hit because it will use the GPU that is currently under- utilised.
This is clearly wrong. If anyone of you is thinking that you are NOT using VGA acceleration then please try running the XFB server on the framebuffer device.
All current desktop environments do use the 2d acceleration including things like BitBlockTransfer (Blitting) and resizing and drawing polygons/lines etc.
By definition a 3d environment will require more resources, especially memory and CPU overhead to keep track of 3d properties. Even if texture mapping is somehow faster than blitting still the 3d algorithms are generally way heavier than 2d primitives.
Anyway, I strongly favour the idea of better GUI but frankly everything has a certain cost and 3d GUI is definitely not cheap in CPU/GPU/memory terms.
My lifting consists of 1-2 hours typically (depending on how focused I am), and I have a 6 day split (2 days on, 1 day off, all 6 to cover my whole body). I aim for at least 30 minutes of cardio per day, sometimes I do more sometimes I do less...
Your advice is interesting but I'd like to point out that
only highly dedicated people are likely to follow such
an extreme fitness program.
It's like Linux vs. Windows. Some people like to tweak
stuff, some don't. Some people like to work out and
do special diets etc and some don't.
General advice towards the population should be
a) easy to follow, b) not very restrictive c) beneficial
It would be good enough (from a medical perspective)
if every overweight person started eating 200-300
calories less per day (one snack less) and doing
20-30 minutes of light aerobic workout every 2
days (walking the dog, shopping, walking in the
park).
Telling the people to follow your schedule is likely to
make most of them quit! (especially the 10% body
fat is unbelievably low, I'm even wondering whether
it healthy... I mean, fat is a tissue not waste, right?)
Actually there is a difference in the way CPU and GPU see memory.
A CPU cares a lot about latency because typical code will have "random" accesses scattered with calculations in between. The same data and code areas are often accessed many times and data are small (e.g. a Word document is small) while code maybe quite large.
That's why CPU's don't have enormous 256-bit buses (which have the same latency as a 64-bit bus)
A GPU performs "multimedia" calculations which typically involve serial access to memory where caching can be of very little help. You cannot "cache" a whole texture set and code is of really trivial size (until now, maybe PixelShader 2.0+++ will change all that). Therefore a GPU needs serial access to huge areas of memory, involving items of similar size and in regular intervals. That's why a GPU needs BANDWIDTH (not necessarily latency, because when the calculation starts latency is hidden inside the calculation loop).
Considering the above, P4 is a "multimedia" design (much more like a GPU) that's why it was made to work with very high FSB and RAMBUS (high bandwidth) originally. Contrary to this, AMD Athlon is a "generic" design which does not depend on huge bandwidth but on very low latency (hence the HUGE L1 cache). That's why P4 needs HyperThreading : its long pipelines do not care a lot about latency but can cause a big bottleneck if they stall. Intel feeds them continuously by drawing instructions from 2 processes at once (so that the pipeline does not remain empty if one process is stalled from the front side bus or something...).
Anyway, I expect GPUs to drift slowly towards the generic CPU design because pixelshader language has become quite complicated with long loops etc. Gradually this means that GPUs (esp. with DirectX9) will start being compute-limited and not texture-fill-rate limited (anything over 2 GTexel/s is really absurd for typical screen sizes). This will propably become apparent with DOOM III.
> Noatun plays MP3s with only modest smoothness. mpg123 > suffers similar problems. Skips are common when switching or > redrawing windows. Real users stick to command lines, I guess.:)
This is quite hard to believe, especially since I was using an Athlon 700 with great success for multi-track recording and other media work. Especially MP3 playback is usually at the 3-4 % range cpu load so it is quite trivial as long as the process gets some time now and then.
You could use the low latency patch (I'm not using it right now) but you also need to have a close look at your interrupts. Definitely use hdparm and make sure you sound card is properly configured. I'm using sb live! with drivers by creative (opensource.creative.com, or try sourceforge).
And of course all my drives are hdparm -u1 (unmask interrupt).
This kind of stuff is VERY important. With a tbird 1000 you could do up to 3-4 44.1/16bit channels + some FX easily (try ecasound).
Also try increasing buffers and see if it helps. Note that increasing buffers increases latency but decreases skipping.
And finally (not the best idea, but you should do it for serious work) you could give root privilege to your recording application so that it can use real time scheduling. Ecasound does that.
P.
P.S. I'm not sure about usb keyboard/mice. I never used
them but they could be quite interrupt-hungry (serial
line is VERY hungry, e.g. serial modem). You could
take a look at that, too.
Quite possibly SSE2 support is a bigger short-term gain than 64-bits because many applications are "tuned" for SSE2 use. By supporting SSE2 (and possibly, later Hyper-Threading) AMD will immediately increase performance. Creating a new "set" like "3dnow Extra" is much less likely to get developer support so fast.
After all, SSE2 is not such a bad idea!
P.
P.S. Don't get me wrong, x86-64 is cool, but it isn't something you will immediately enjoy with games and commercial (!open source) apps. Maybe in a few months....
> 2) the partioning scheme. Only 4 partitions!!!! this is an artifact from the days of the original PC.
You can definitely have more than 4 partitions. I think 2 primary plus 8 logical is the maximum (total of 10). It would be nice to have an arbitrary number of partitions but this is not very important in my opinion.
Regarding the BIOS, I have to say that I like it the way it is. Complex things fail and BIOS should be 99.999% error free. I don't see the need for a themeable GUI-Bios with transparent windows. If it works, don't break it.
The fact that 64-bit XP exists does NOT mean that it will work on ALL 64-bit processors! Clearly, IA-64 (Itanium) is NOT the same as x86-64 (Opteron) and XP for Opteron needs quite a lot of different low level code. It is a different processor!
However, AMD has released working silicon (and complete specifications, AND an emulator) to partners a long time ago. Please check http://www.x86-64.org. The reason Opteron has taken quite a long time to release to the public is that it has to be competitive with an already fast processor (Athlon 3000+, P4 3.06 etc) so it has to reach a very high clock rating AND it has to be widely available.
I believe the first "unofficial" benchmarks had been available a few months ago. Also note that according to AMD, test systems are available (www.amd.com).I'm sure you can read about working systems presented during the last year.
Anyway, to sum this up, I'm sure that if this was an issue, MS would have had BETA (or ALPHA!) silicon a VERY long time ago for developement. Hell, even UT2003 has been recompiled for x86-64 and linux/arch/x86-64/ is already 35000 lines of C and assembler code!
P.
P.S. I just found out that Tom's hardware had seen x86-64 silicon from 27 February 2002. Go check http://www.tomshardware.com/cpu/20020227/ I'm sure they could spare a few extra systems for MS;-)
Slashdot Mirror
It seems you need to find better doctors. Sometimes what you pay is what you get. You can get top-notch health care in the US, if you can afford it. If you want to do web searches for everything, by all means do. It is your health after all. Just don't be too eager to assume that you can do better than a properly trained doctor (we doctors have discovered the web, too).
P.
It's usually very hard to put a diagnosis without proper data. In this case, even though I'm currently preparing for USMLE-like examinations (you are presented with some data and you try to make a diagnosis), I find it very hard to trust Pat's "clues" because I don't know what is real and what is *his* idea of a diagnosis.
As a doctor I really like to hear my patients tell me actual facts and not their interpretations. E.g. "I have fever and sore throat" and NOT "I have the flu". Infectious diseases that may present with fever and sore throat are many (ranging from primary HIV infection to infectious mononucleosis to common cold) and it's highly unlikely that the patient has considered all of them. By focusing on a possible "diagnosis" the patient may ignore other signs that would be useful to the doctor.
I could list quite a few diagnoses that would fit Pat's description, but guessing is quite useless, especially in important health matters. Maybe some doctors did not follow proper standards of care but the fact that an assumed serious condition did not alarm so many of them is quite suspicious.
As a simple advice (I hope Pat is reading this!): IF you have fever (defined typically as over 38.3 deg. Celsius) plus a NEW audible cardiac wheeze (not mitral prolapse, which is quite different) you should be admitted to the hospital on the basis of an assumed diagnosis of bacterial endocarditis (unless *proven* otherwise). Bacterial endocarditis usually develops on PRE-existing pathological conditions (e.g. old rheumatic fever, IV drug use). Typically, cardiac ultrasound (why don't you go have one, if you are so worried?) will give very useful clues. Examination of the retina and blood cultures (at least three) are also necessary. If no signs of bacterial infection are found, several viral pathogens can cause pericardial inflammation but I can only remember Coxsackie and echoviridae off the top of my head. Viruses usually cause milder disease.
Finally, please do not trust the web, google, medline, nih. These are excellent data sources, but you are unable to properly interpret what you read without proper training. You can't just open "harrison's internal medicine" and hope to acquire the skills to make a diagnosis in a few hours/days/weeks. Find a good doctor and trust him. Sure, some people say that they correctly diagnosed their condition, even though the doctors where wrong. It happens, doctor's make mistakes. But on 99.99% of cases, your doctor knows better than you.
P.
You mean by NOT buying it after you didn't buy it in the first place? I'm sure they'll notice the difference.
NOT!
No, I mean by not buying even though I intended to. I thought that was quite clear. Many people have propably waited a few days to read the reviews and hear other opinions before buying. I'm one of them.
P
Annoying policies like this one may increase short-term profit but people will, I hope, gradually react. I must say that from all the best-selling FPS epic has showed the best company policies. They released a linux version concurrently (on the same CD!), ut2004 does not require the CD to be in the drive and they keep releasing free files in the ut2004 community (extra maps etc).
Sure, everyone wants to play HL2. I know I do. But I think I'd rather send them a message by NOT buying their game. Would you go to a restaurant where the waiter insults you, even if it is the best one in the town?
P.
>> European countries actually use the euro to buy oil.
> No, they don't.Well, it seems they do. At least my country does use euros to buy oil. Don't know about other european countries.
P.Many people like to think that we know everything there is about music reproduction so we can model almost everything using enough bits and MHz. Sure, you can do a good job in most cases and with a fraction of the cost but guess what:
a) We do not always KNOW why machine 1 sounds better than machine 2, and no THD and response curves are not enough to describe the complex phenomena of musical reproduction. So we can't model this.
b) If a hypothetical machine using silver cables or tubes or pixie dust sounds better then maybe instead of trying to disprove several people that enjoy its sound you could try to find out WHY it sounds better. People are not always victims of the placebo effect.
And let's not consider technology to be the perfect substitute for everything: sure, you can try to imitate a piano using N processors and god knows what else, but I'd prefer a plain (plain in the sense that it does not contain transistors ;-)) steinway piano (~100k $) any day of the week. It's a matter of taste. If a musician likes the sound of tubes and he can afford them then let him happily enjoy them.
For the record, the absolute best audio reproduction I have encountered came from Audio Research (www.audioresearch.com, have a look) tube amplifiers.
P.Seriously, sugar is hardly the problem with obesity in America. The problem is primarily one of poor eating habits, coupled with lack of exercise. (Not that anecdotes prove anything, but just to pull out one random example; I used to know a gal who was a strict vegetarian, and I *never* once saw her eat a piece of candy or "junk food" - yet she was overweight.)
It is certainly true that surplus caloric intake causes obesity, regardless of macronutrient source. However, all macronutrients are not processed in the same way. Sugar is a special case for the human body because it quickly increases blood glucose level (thereby triggering homeostatic mechanisms--insulin) and equally quickly disappears.
The net result is that, even though your empiric evidence may point to the contrary, research is still necessary in order to fully determine whether sugar consumption is linked to obesity. I could point you to some interesting articles, but the general idea in the relevant bibliography is that eating sugary foods does seem to predispose to obesity. And, to put it in a different light, sugar does not seem to offer anything useful (except in cases of hypoglycemic coma!) besides plain old calories, so you are better off eating more complex carbohydrates.
P.
The idea that only "intelligent", "educated" etc
people should vote is actually wrong. The logic
behind voting does not have to do with getting the
best possible decisions (best decisions come from
think tanks and specialists). The REAL reason
for voting is MORAL, and has to do with sharing
responsibility. The idea behing democracy is that
people should bear some of the burden of power.
After all, if you HAVE to vote you will, even out
of curiosity, try to learn something in order
to make an intelligent choice. This would really
force the political parties to address voters in
a completely different manner.
Once more, democracy is a system for distributing
*blame* (moral burden of power), not a system for
making optimal decisions.
P.
As an owner of a 9700 and a hobbyist developer, I'm very familiar with the limitations. The shader length is highly restricted, conditional branching can't be done, so loops have to be unrolled. For this reason, even the latest ATI cards can't fully support the OpenGL Shading Language. What can be done on an FX or a Geforce 6 in one pass could take 10 or more passes on an X800. Many important features for shadow mapping are hopelessly missing, such as rendering to a depth texture, and hardware linear filtering.
Copying from ATI's web site, the specifications
for RADEON 9800 PRO say:
# SMARTSHADER(TM) 2.1
* Full support for Microsoft® DirectX® 9.0 programmable vertex and pixel shaders in hardware
* 2.0 Vertex Shaders support vertex programs up to 65,280 instructions with flow control
* 2.0 Pixel Shaders support up to 16 textures per rendering pass
* New F-buffer technology supports pixel shader programs with unlimited instructions
* 128-bit, 64-bit & 32-bit per pixel floating point color formats
* Multiple Render Target (MRT) support
* Shadow volume rendering acceleration
* Complete feature set also supported in OpenGL® via extensions
So, it seems that ATI cards do support flow
control (i.e. loops) and infinite pixel shader
length. Shadow volume acceleration might be
somehow limited as you say, but I haven't
really looked into it. Maybe some ATI white
paper will clarify these details.
NOTE that I'm referring to a 9800 product and
not to the high-end X800.
P.
Benchmarking is not about *proving*, it is about
*measuring*. Note that Goedel's incompleteness
theorem does not necessarily imply that ALL
propositions are unprovable! It merely states that
there is at least one proposition that is true but
cannot be proven (Goedel constructed this
proposition in an ingenius manner[1]). It suffices
to build ONE such proposition to derive incompleteness.
MANY other propositions can be proven, including,
possibly the fact that card A is better than card B
(if you can call this thing "proof").
Therefore, proof that card A is better than card B
may be feasible. However, the idea of proving
something like this is ridiculous. What you need
is an ordering of the cards according to some
way of measurement. This means that for some
arbitrary benchmarking procedure P() for card
X you get a number P(X) and order all cards
according to that. IF P() is relevant to you (e.g.
quake 3) the ordering is sound (assuming many
other variables, like system, software etc are the
same). Nothing esoteric about that.
(by the way, the profound theoretical significance of
Goedel's incompleteness theorem is completely opposite
to its practical impact, which is almost none in everyday
mathematics--most things have nice proofs)
P.
[1] The importance of the theorem lies in the fact
that even if you convert an unprovable TRUE proposition
into an axiom you will get a new system that will
AGAIN be incomplete. You can't get away by building
bloated systems.
To quote from memory : "the program will detect
discrepancies in a user's thinking and alert the user".
This sound an awful lot like "Clippy" raised to the Nth
power. I am certain that MS products will be the first
to feature this Smart-Ass computer technology, whereby
the computer will constantly correct you and interrupt
your thinking with irrelevant bullshit ("are you sure
you want to do this? maybe you want to do that").
On the other hand, just like spell-checking helps
pepole (sic) write clearly, maybe this will allow the
less-than-privileged computer users to follow a
logical path of thinking (quite possibly, an eye opening
experience for some).
Sometimes I feel that AI is awfully close.
P.
Slowly but surely the UNIX crept upon the Nintendo user.
Recently some of us have observed that
CD-R quality has gone down the drain. Too
often I buy disks that fail really soon.
The real question for me is: which of
the two formats is most reliable in the
long term? Will I be able to read my disks
5 years from now? Which is most reliable,
DVD-/+R or CD-R?
Capacity and speed is nice but I don't
tolerate data loss very well...
P.
Certainly onboard audio is acceptable for
many uses (office desktop etc) but a decent
sound card can be bought for a ridiculous
amount of money so it may be worth it.
I recently bought an Audigy Player (OEM)
for roughly 50 Euros and it is great. You
get an effects processor, real 5.1, a load
of utilities that are quite nice and MUCH
better signal quality than most onboard
cards for the price of 3 CDs. Many CPU
coolers cost more than this and they
are just blocks of metal!
You should definitely stick with onboard
if you own cheap speakers and don't
play a lot of music or many games but
in my opinion a decent card is worth the
small extra spending and it will last
many years. As a matter of fact I also
suggest getting decent speakers (like
Cambridge Audio 2400). For less than
100-120 Euros/$ you can get very
respectable (computer) sound quality.
P.
I had a very noisy computer and I spend
some money to improve it. I used the
following components:
- Akasa pax.mate sound insulation 30 E
- Thermaltake Volcano 9 fan/cooler 35 E
- Zalman 400W PSU (EXCELLENT quality!)
at 118 Euro
- High quality case with 2x12cm fans (not
the standard 8 cm) at 50 Euro without
PSU
I also stripped the fan from my
Northbridge (VIA KT400 chipset) and it's
only 3-4 degrees C hotter (~42 deg. C
with ~30 deg. C ambient).
I am now considering a passive VGA
cooler (Zalman, 27 Euro) for my video
card (Geforce 3 Ti 200).
P.
> Did you know you can fertilize your lawn with used motor oil?
OK, this is totally off topic, but you should
NOT fertilize your lawn with motor oil. It
contains several mutagenic/carcinogenic
substances and you don't want them going
underground to pollute water/plants etc.
Used motor oil should be properly disposed
of in places that later sell it/use it where
appropriate (ships can burn it, it has other CLEAN uses).
P.
Regarding the hard problem: it looks a lot
like creating a Huffman coding tree.
You need a binary tree where the
weight (information = -log propability)
influences the depth. Rare things go
deeper and need more bits to find.
Common things go higher. Therefore
if, say, you want to describe something
common you know it is at 01 (2 bits
instead of 8) while something rare
is at 1001111101 (11 bits instead of 8).
Bits describe the path on the tree.
This is not particularly hard and you
can construct an optimal solution
VERY quickly.
Anway, the problem is a bit different
(Huffman trees store information only
at the leaf nodes and they are not
ordered)
but I'm sure that the same principles
apply.
The general idea is that you progressively
built subtrees starting from the less
propable (less weight) items, combining
them two at a time while preserving
ordering[1]. You then treat each subtree
as a single item (classical dynamic
programming) and see if you can build
any further using it.
Many problems in such competitions
are solved with dynamic programming.
P.
[1] This is possible because you can
always choose between left and right
at the same depth. Depth is based
on weights and left-right ordering
on key values.
P.S. Could be much harder, though. I
just read the problem description and I
might be saying a load of crap.
Sometimes a small change can transform
a O(N) problem to an NP-hard problem.
>special interests have corrupted
>mainstream benchmarks to make them an
> unusuable guide.
This is true, to a certain extent!
People used to benchmark with, say,
Office scripts or other "office" jobs.
Most modern benchmarks are almost
100% multimedia (Internet content
creation, Divx, MP3 photoshop etc
come to mind). This is very convenient
for Intel because P4 is a multimedia
design (long pipes, high MHz, small
cache, fast FSB, SSE2) designed for
serial operations (small loops, no
branches) with huge data sets (fast
memory is good here).
The Athlon is a completely different
design and it cannot compete. I am
really wondering how it performs e.g.
with kernel compiles or other
system scripts.
Anyway, this reminds me of the situation
with K5 and Quake. Until the era of
Quake (and Pentium Classic) all
competing processors (AMD K5, K6,
Cyrix) had very very slow FPUs.
Quake was the first widespread game
to use FPUs (it relied on the FPU to
compute 1/z with fdiv) and even though
the K5/K6 were VERY fast with integer
code the Pentium Classic had vastly
superior performance under Quake
(popular benchmark at the time!).
Anyway, this whole situation is not
necessarily the fault of the benchmark
designers. Frankly, even though the
AMD may be faster at e.g. Excel I don't
intend to use huge multi-megabyte
spreadsheets (most of us don't) but
I do intend to use MP3/MPEG2 etc
algorithms for which (like Internet
Content Creation) the P4 may be
a better choice for those that can
afford it.
Actually I do own an Athlon 2200+
and I am very pleased with its
performance mostly (and most
importantly) because it is CHEAP.
(AMD always has better MIPS/FLOPS
per $, that's why I buy AMD)
Some people here have pointed out
that this '3d' effects desktop will not
cause a performance hit because it
will use the GPU that is currently under-
utilised.
This is clearly wrong. If anyone of you
is thinking that you are NOT using VGA
acceleration then please try running the
XFB server on the framebuffer device.
All current desktop environments do use
the 2d acceleration including things
like BitBlockTransfer (Blitting) and
resizing and drawing polygons/lines
etc.
By definition a 3d environment will
require more resources, especially
memory and CPU overhead to keep
track of 3d properties. Even if
texture mapping is somehow faster
than blitting still the 3d algorithms
are generally way heavier than 2d
primitives.
Anyway, I strongly favour the idea of
better GUI but frankly everything has
a certain cost and 3d GUI is definitely
not cheap in CPU/GPU/memory terms.
P.
My lifting consists of 1-2 hours typically (depending on
how focused I am), and I have a 6 day split (2 days on, 1
day off, all 6 to cover my whole body). I aim for at least 30 minutes of cardio per day, sometimes I do more
sometimes I do less...
Your advice is interesting but I'd like to point out that
only highly dedicated people are likely to follow such
an extreme fitness program.
It's like Linux vs. Windows. Some people like to tweak
stuff, some don't. Some people like to work out and
do special diets etc and some don't.
General advice towards the population should be
a) easy to follow, b) not very restrictive c) beneficial
It would be good enough (from a medical perspective)
if every overweight person started eating 200-300
calories less per day (one snack less) and doing
20-30 minutes of light aerobic workout every 2
days (walking the dog, shopping, walking in the
park).
Telling the people to follow your schedule is likely to
make most of them quit! (especially the 10% body
fat is unbelievably low, I'm even wondering whether
it healthy... I mean, fat is a tissue not waste, right?)
P.
Actually there is a difference in the way CPU and GPU see
memory.
A CPU cares a lot about latency because typical code will
have "random" accesses scattered with calculations in
between. The same data and code areas are often
accessed many times and data are small
(e.g. a Word document is small) while code
maybe quite large.
That's why CPU's don't have enormous
256-bit buses (which have the same latency as a 64-bit
bus)
A GPU performs "multimedia" calculations which typically
involve serial access to memory where caching can be of
very little help. You cannot "cache" a whole texture set
and code is of really trivial size (until now, maybe
PixelShader 2.0+++ will change all that). Therefore
a GPU needs serial access to huge areas of memory,
involving items of similar size and in regular intervals.
That's why a GPU needs BANDWIDTH (not necessarily
latency, because when the calculation starts latency
is hidden inside the calculation loop).
Considering the above, P4 is a "multimedia" design (much
more like a GPU) that's why it was made to work with
very high FSB and RAMBUS (high bandwidth) originally.
Contrary to this, AMD Athlon is a "generic" design which
does not depend on huge bandwidth but on very low
latency (hence the HUGE L1 cache). That's why P4 needs
HyperThreading : its long pipelines do not care a lot about
latency but can cause a big bottleneck if they stall. Intel
feeds them continuously by drawing instructions from 2
processes at once (so that the pipeline does not remain
empty if one process is stalled from the front side bus or
something...).
Anyway, I expect GPUs to drift slowly towards the generic
CPU design because pixelshader language has become
quite complicated with long loops etc. Gradually this
means that GPUs (esp. with DirectX9) will start being
compute-limited and not texture-fill-rate limited
(anything over 2 GTexel/s is really absurd for
typical screen sizes). This will propably become apparent
with DOOM III.
P.
> Noatun plays MP3s with only modest smoothness. mpg123 :)
> suffers similar problems. Skips are common when switching or
> redrawing windows. Real users stick to command lines, I guess.
This is quite hard to believe, especially since I was using an
Athlon 700 with great success for multi-track recording
and other media work. Especially MP3 playback is
usually at the 3-4 % range cpu load so it is quite trivial
as long as the process gets some time now and then.
You could use the low latency patch (I'm not using
it right now) but you also need to have a close look at your
interrupts. Definitely use hdparm and make sure you
sound card is properly configured. I'm using sb live! with
drivers by creative (opensource.creative.com, or try
sourceforge).
My interrupt table is (truncated):
14: 91022 IO-APIC-edge ide0
15: 14 IO-APIC-edge ide1
16: 402363 IO-APIC-level nvidia
18: 29 IO-APIC-level eth0
19: 17168 IO-APIC-level EMU10K1
21: 0 IO-APIC-level usb-uhci, usb-uhci, usb-uhci
NMI: 0
LOC: 461516
ERR: 0
MIS: 0
And of course all my drives are hdparm -u1 (unmask
interrupt).
This kind of stuff is VERY important. With a tbird 1000
you could do up to 3-4 44.1/16bit channels + some
FX easily (try ecasound).
Also try increasing buffers and see if it helps. Note that
increasing buffers increases latency but decreases skipping.
And finally (not the best idea, but you should do it for
serious work) you could give root privilege to your
recording application so that it can use real time
scheduling. Ecasound does that.
P.
P.S. I'm not sure about usb keyboard/mice. I never used
them but they could be quite interrupt-hungry (serial
line is VERY hungry, e.g. serial modem). You could
take a look at that, too.
Quite possibly SSE2 support is a bigger short-term
gain than 64-bits because many applications are
"tuned" for SSE2 use. By supporting SSE2 (and possibly,
later Hyper-Threading) AMD will immediately increase
performance. Creating a new "set" like "3dnow Extra"
is much less likely to get developer support so fast.
After all, SSE2 is not such a bad idea!
P.
P.S. Don't get me wrong, x86-64 is cool, but it isn't
something you will immediately enjoy with games
and commercial (!open source) apps. Maybe in a
few months....
> 2) the partioning scheme. Only 4 partitions!!!! this is an artifact from the days of the original PC.
You can definitely have more than 4 partitions. I think
2 primary plus 8 logical is the maximum (total of 10).
It would be nice to have an arbitrary number of partitions
but this is not very important in my opinion.
Regarding the BIOS, I have to say that I like it the way it
is. Complex things fail and BIOS should be 99.999% error
free. I don't see the need for a themeable GUI-Bios with
transparent windows. If it works, don't break it.
P.
The fact that 64-bit XP exists does NOT mean
I'm sure they could spare a few extra systems ;-)
that it will work on ALL 64-bit processors!
Clearly, IA-64 (Itanium) is NOT the same as
x86-64 (Opteron) and XP for Opteron needs
quite a lot of different low level code.
It is a different processor!
However, AMD has released working silicon
(and complete specifications, AND an emulator)
to partners a long time ago. Please check http://www.x86-64.org.
The reason Opteron has taken quite a long time
to release to the public is that it has
to be competitive with an already fast processor
(Athlon 3000+, P4 3.06 etc) so it has to reach
a very high clock rating AND it has to be
widely available.
I believe the first "unofficial" benchmarks had
been available a few months ago. Also note that
according to AMD, test systems are available
(www.amd.com).I'm sure you can read about
working systems presented during the last
year.
Anyway, to sum this up, I'm sure that if this
was an issue, MS would have had BETA (or
ALPHA!) silicon a VERY long time ago for
developement. Hell, even UT2003 has been
recompiled for x86-64 and linux/arch/x86-64/
is already 35000 lines of C and assembler
code!
P.
P.S. I just found out that Tom's hardware
had seen x86-64 silicon from 27 February
2002. Go check
http://www.tomshardware.com/cpu/20020227/
for MS