Dan's Data letters #90
Publication date: 11 February 2004.Last modified 03-Dec-2011.
Leaderboard time
Where can I find performance tables for different CPU architectures?
Aaron
Answer:
Cross-platform performance testing is tricky; individual application benchmarks
can tell you more about how well optimised the compiler was than about the
actual architecture performance, and general
SPEC-type benchmarks are irrelevant to
most desktop computer tasks.
With that said, though, the old CPU Info Center is now quite badly out of date, but still a valuable source of info on older chips; the General Microprocessor Info PDF file there, while incomplete and not at all current, is about as close as you'll get to an overall ready-reference guide to the performance of different desktop and lighter workstation processors.
For more up to date information, check out ChipGeek. For x86-only data, try Sandpile.
x86 cranio-transplant
I am upgrading a friend's old PC, and instead of going through the trouble of upgrading the processor from a 260MHz AMD K6/3D, I have bought a second hand motherboard with a AMD K6-2 at 500MHz, with better video and sound cards included.
Now my question is: Can I just unplug the old motherboard and plug the other one in, or do I have to be prepared to spend two days on my knees praying to God for help when WinXP won't work?
Christophe
Answer:
Oh, XP'll freak out, all right. Windows never likes a brain transplant.
However, you ought to be able to get running without a full reinstall by "repairing" the existing XP installation, after you swap the boards. Just boot from the XP CD (which you should be able to do with a mobo of this vintage, though you'll probably need to fiddle with the boot order in the CMOS setup program) and follow the prompts.
RCR-V3 debate continues
Regarding the rechargeable CR-V3 battery mystery in your latest letters, here's what (I think) is going on.
The impedance of a new Li-Ion cell is much less than that of a new NiMH cell, and the difference gets worse as the cells age. So under a moderate-to-heavy load the voltage on NiMHs would drop far enough that camera would quit, and most of capacity would go unused, whereas Li-Ion would still keep going.
7 [Possibly my first Borg correspondent.]
Answer:
Lithium ion internal resistance can be lower than that of NiMH, but
you're still only likely to be talking around 0.35 ohms at worst for 7.2
volts worth of NiMH cells (around 0.25 ohms, for the same number of NiCds,
at worst; lower resistance NiMH cells can score 0.25 ohms, too). 7.2 volts
of lithium ion cells (two cells, since they're 3.6V each versus 1.2V for
NiMH and NiCd) will give you a range from 0.15 to 0.25 ohms, and lithium
polymer cells (also 3.6V per cell) will be around 0.2 to 0.3.
So low-resistance NiMH cells are about as good in ohms per terminal volt as lousier lithium ion cells, but they're all so low that only when you're talking 1C and above currents (current in amps equal to capacity in amp-hours) does the difference really matter at all. NiCds are most happy with high current loads (which is why they're still widely used in radio control and other super-high-draw applications), NiMHs are next happiest, and lithium ion are less happy again. Their internal resistance may be a tad lower than NiMH, but they're much more fragile.
The new rechargeable CR-V3s must have some jiggery-pokery going on in there to give them a 3V terminal voltage; presumably they've got an internal regulator of some sort, which will slightly improve their current delivery capability if it's a whiz-bang DC-to-DC converter (maybe that explains their thrilling price per unit), but which won't if it's a plain old linear regulator. It could just be a simple voltage-dropping diode, for all I know.
But yes, the lower terminal voltage of NiMH cells might indeed cause some cameras to quit early. Never mind internal resistance; an RCR-V3 has three nominal volts to the 2.4 nominal volts of the pair of NiMH cells it replaces, which could tip the balance. In badly designed cameras, that is.
With normal photographic loads, NiMH and NiCd rechargeables will stay above 1.1 volts per cell for most of their discharge curve, and will be basically empty by the time they drop below 1.0 volts per cell. It's not hard to design a camera power supply that can cope with everything from one volt per cell (near-flat NiMH) to 1.7 volts per cell (brand new non-rechargeable lithium AA). Some cameras definitely don't deal well with the lower end of that range, though, as we'll see...
The little camera that almost could
I have some comments on your observations about rechargeable CR-V3's, and my own very preliminary results, using my Olympus D-560.
I've been sitting here firing off shots in pretty rapid succession at high quality resolution, LCD on all the time, all shots requiring flash. No time for the batteries to rest up.
My baseline with the 560 was that my (rather old) using Olympus NiMH cells, fully charged, would cause the camera to display its low power light after only ten photos, but would be perfectly good for approximately another 40 shots with the LCD on, flash used occasionally, and one download to my computer. With the LCD off, except for reviewing after shots and changing some settings, I got well over 220 shots and the camera was still doing just fine, so I don't know how far it would go. The power low light was on most of the time, though it caused me virtually no trouble.
Now for the rechargeable CR-V3s
Shooting in the HQ mode most of the time at 2048x1536 (maybe ten shots were done at SQ1 1024x768) I fired off 176 shots, including one USB download to my computer, and at no time did the power low indicator come on. This in contrast to about 40 with the Olys. The Olys are far from new (they've had enormous use over the years and I've used a Rayovac fast charger when not traveling), so I may do some testing with newer ones with higher capacity.
Still, the only sign that the L3s were running down after 176 shots was that it was taking a long time for the USB connect to my computer for a second download, which is a general sign of weariness. There's still simply no comparison with the NiMHs I'd been using, though.
To me, the RCR-V3s are a pretty spectacular improvement over NiMH, and it eliminated the D-560's "power low" problem, that many users have reported on. If I were to switch over to my LCD-off mode I would expect to be able to take the 450 shots suggested by Steve's Digicams.
Ron
Answer:
The results you report do suggest that the NiMH cells you're using are pretty
clapped out; just because they look OK on a battery tester doesn't mean
they've got more than half of their original capacity left. Near-dead rechargeables
can still deliver full rated voltage and a respectable current when freshly
charged - they just won't keep on keeping on for very long.
Even taking that into account, though, you're getting strangely awful performance from your D-560, by modern digicam standards. Digital cameras used to eat batteries like popcorn, but they don't any more; my old piece here on building big batteries is now of little more than archaeological interest.
As you say, the "power low" problem with the D-560 is frequently reported; it seems to be normal for that camera. A quick Web search reveals several other people complaining about crappy battery life with D-560s using NiMH cells, including one interesting comment here, which says that the camera decides its batteries are flat too early, leaving (according to the poster) about 70% of the cell capacity unused.
That sounds like an overstatement, since he then says he could only take another 30 (presumably flash) pictures with another un-named camera, unless he gave up very early on the 560. But it does line up with what you say about the 560 giving a too-early low battery warning.
(And then there's another poster claiming "over 140 mostly flash photos from a single pair" of NiMH cells. So who knows. But I'm on a roll now, don't spoil things for me.)
The rechargeable CR-V3s have higher voltage than NiMH and NiCd AAs, so they'll last longer in a camera like yours, that wigs out when the supply voltage drops.
Only 40 shots from a vaguely recent couple of NiMH cells, though, is really lousy. Run my old Olympus C-2500L (a more than four year old camera, now; modern designs should be a lot more efficient) from some bog-standard current 1800mAh NiMH cells, and you might be able to take as many as five hundred flash pictures, if you keep it fed with memory cards and never stopped shooting.
In real-world use, you could expect the thick end of 200 flash shots, easily.
OK, the 2500L runs from _four_ AAs and the D-560 has only two. The 2500L is also an SLR camera, with no LCD viewfinder function at all [in later correspondence, Ron pointed out that he could take a lot more pictures if he minimised LCD use...].
But even if you allow for half as much battery energy, 40 shots from a 2003-model digicam totally bites the wax tadpole, dude.
Off-brand 1700mAh-plus NiMH cells are cheap these days, and much better value per amp-hour than fancy brand name 2Ah-plus cells; I recommend you pick up at least a couple of sets to supplement your zooty new lithium ion pack. You've got the chargers already, so a few bucks on a couple or four AAs won't be a big deal.
EPROM power!
I have a question about your page on chip upgrades to improve car performance.
Mainly, my question is why what you say, when the Powerchip site pretty much says the exact opposite on all counts.
Would Powerchip lie outright, and provide a three year warranty with possibility for an extension for the drivetrain?
In searching through the Web I only come across your opinion of a chip swap being a bad choice to upgrade. If you can refer me to your references I can make a better judgment on whether or not it really is not good to upgrade my ECU.
Tom
Answer:
First up: I didn't say that drop-in Electronic Control Unit (ECU) upgrades
for otherwise stock vehicles were outright fraud, though some companies
in that market have certainly been snake oil merchants. I just said that
a drop in chip isn't likely to be good value compared with various
actual mechanical upgrades. Powerchip, like various other chip vendors,
will charge you several hundred Australian bucks for a new chip.
Now that I've said that, dig this.
A while after I put my piece on ECU chips up on the Web, one Wayne Besanko of Powerchip contacted me.
He did not offer any independent evidence to support Powerchip's claims. Nor did he point out anything I'd said that was wrong.
Instead, he offered me money, plane tickets and accommodation if I'd travel to Powerchip's HQ and write a "white paper" on Powerchip's products.
He didn't say "here's a bucket of cash, if you write what we say", but our correspondence led me to the firm belief that, um, only one viewpoint on their products would be acceptable, were I to take up the offer.
So there's that.
And, again, as I write this, I remain unaware of any proper independent testing that indicates that these pricey drop-in ECU chips are good value, compared with a variety of actual mechanical modifications.
Sure, you can get a bit more juice from a stock engine by goosing up the ECU programming; drop-in chips from reputable companies like Powerchip don't generally do nothing. I wouldn't be surprised if there were quite a few cars, particularly turbo diesels, that have sub-optimal stock ECU programming, leaning further towards the "green" end of the scale and away from the "performance" end than their owners would choose, given the option.
The particular oddities of individual engines (in high performance cars, at least) may also benefit significantly from custom-tuned ECU maps, even if you aren't going for new cams, an after-market turbo, blah blah blah.
But drop-in chips aren't tuned for individual engines. They're one-size-fits-all. If you want a chip that fits your car's engine in particular, you have to go to a speed shop that'll test your engine and blow an EPROM to suit.
In the vast majority of cars, I think it's quite sensible to say that if you aren't making significant mechanical modifications to your engine, then the money you'd spend on a "hot chip" would be better put towards those modifications (or, you know, spent on the rent or something, but we're not talking about sensible life choices here). I think that even something as simple as a less restrictive air filter is likely to give you more horsepower per dollar than a hot chip.
Even Powerchip themselves admit (or, at least, did admit at the time I corresponded with Wayne; I haven't groveled through their specs lately) that a 15% power and torque gain from a plain chip swap is unusually high. Figures closer to, or below, 10% are common. Some people would question even that - but even if you get a whole 15%, that ain't a hell of a lot. Worth having for free, sure, but Powerchip's products aren't free.
If someone's got $AU500 or so (are there any Powerchip upgrades cheaper than that?) to spend on performance modifications for their car, and they've already got good shock absorbers and tyres, then they want a better air filter, maybe a better exhaust system, other mechanical upgrades as appropriate to the weaknesses of their particular vehicle, and maybe to save up for a turbo.
If they've already got a turbocharged car, then they can get a boost control. And if they go so far as to rework their engine with new cams and valves and so on, then they'd probably do well to get a chip blown to match, and maybe also a fancy-pants new ECU if they're a real racer, or just dig techno-toys.
Shortly after I mentioned to Wayne that that I'd read this edifying article and was not impressed by the politician's answers he gave there, he clammed up.
Makers of "hot chips" for basically-stock cars are, by and large, an industry with no proper objective testing of their products, lots of gullible consumers, and plenty of fraudulent claims. I've no reason to think that Powerchip themselves make fraudulent claims, but some chip vendors certainly do.
Since Powerchip appear to have, more than once, advertised drop-in chips for otherwise stock vehicles that give superior engine performance to high-spec factory-custom versions of the same car, I can't help but think that something must be fishy. As has been said numerous times - why would factories bother with a whole new ECU, exhaust, porting, cams, et cetera, if they could get functionally identical results by just blowing a new EPROM?
Rainbow sneezes
Dan, I just sneezed. Why do you care? Well sometimes its hard to get full coverage of the projection zone with my hands before all hell breaks loose and the odd bit of well, lets call it liquid, breaks through and ends up on my CRT monitor.
That little dot of liquid then becomes multi-colored! Why? Why must something that is considered gross end up mimicking a rainbow? Its like caterpillars turning into butterflies!
Now, if I was to splash a cup of water onto my screen, would it look any good before I got zapped? I may not see your reply, as I'm pretty eager to try it out.
Brendan
Answer:
The clear... liquid... is probably just saliva; it's acting as a tiny magnifying
lens, and giving you a larger view of the red, green and blue phosphor dots
behind it. You'll be able to see this more clearly if you look at it through
a magnifying glass - a cheap jeweler's loupe will give you a great view
of the phosphor.
And no, this phenomenon will not scale well. Cups, buckets or barrels of water thrown on your monitor will not produce proportionally more beautiful results.
