Storage networking outfit QLogic has sued archrival Emulex for posting a web video that shows an egg frying on a QLogic converged network adapter. As noticed by The Street, QLogic filed suit on Monday in a California Superior Court, claiming the Emulex video is intentionally deceptive. "Emulex's video purporting to show an egg …
Physics by smoke and mirrors...
Point 1 is that egg is clearly not being fried - simply you cannot fry an egg at those temperatures. Typically you need oil to be up around 180 degrees C or so, not a miserable 190 Fahrenheit. So that's not even a boiled egg - more one that has congealed through a combination of being gently warmed and throroughtly bored by a long wait (hence the time lapse).
Secondly, the claim that sticking heatsinks on the chips reduces heat soak into the server. Well there is a slight point in that a cooler semiconductor will be slightly more efficient so there may be a tiny effect at the margin. A more likely reason is to improve reliability of operation as the chip is unstable at higher temperatures (and, at extremes, it will improve longevity). Hyperclockers know this sort of stuff. However, festooning a board with heatsinks is, as often as not, a sign that it's using a lot of power. Think of all those high performance graphics cards stuffed with fans and heatsinks due to the silly power consumption numbers.
If you are worried about heat soak into your server and datacentre, simply read the board specs on the power consumption (and you can guarantee that server builders pay attention). Choose the one that uses the least power that delivers the performance levels and has a decent reliability reputation (OK - there's not actually that much choice in this case) and don't worry about frying eggs (which they can't). In general, boards with fewer, and smaller, heatsinks (and fans) use less power.
As it is, that video is a joke and treating customers like idiots and is more in the line of the sort of pseudo-science that is used by celebrities to flog over-priced "wonder" cosmetics to the fairer sex ("and now for the science bit"). Just a play on psychology. In this case they haven't realised that they are trying to sell to a bunch of IT nerds, many of whom have a training in science.
one misleading item
is that they didn't specify how much the time was compressed in the video.
The second is that they didnt specify what load the boards were under to generate the operating temperature, however, if they were under the exact same load then this point is meaningless.
I could use a couple of those QLogic chips, there's ice and snow on the ground here in Atlanta.
Great demo, bad physics...
OK, the egg frying was a fantastic storyboard...but the physics is deceptive at best, and possibly badly enough to warrant litigation.
Certainly, having a massive board-covering heatsink a la Emulex is a good design practice, and it _may_ help increase chip life. But that is a POINT heatsource, and while it is true that the Qlogic board has the highest temperature point heatsource, there is NO WAY you can generalise that into saying their board - OVERALL - runs hotter and has more thermal impact to your server and data centre. The only way to draw that conclusion is by comparing the watts of power consumed by each board - and I notice that Emulex doesn't say anything about that...which leads me to believe (too lazy to look up the specs on each board) that they are roughly comparable - which you would expect, as they both are roughly the same performance, and probably both built on the same transistor sizes on die.
So it would appear that Qlogic may very well have a point to litigate...although I can't blame Emulex for trying - it was a great ad.
Fantastic as it is, it is not original
The first Athlons had no thermal cut-out so the silly egg frying trick was quite popular. In fact either tom's or Anand or one of the many other hardware sites had a video of it and it was smoking:). So in fact, it is just a bit of history repeating. What used to be used in scoring cheap points between CPU vendors is now used in the storage arena.
And similarly with the CPU wars of old the demo is unscientific and not representative of overall consumption.
Bad physics, perhaps
As you rightly say, the heat dissipates into the whole server pretty much the same whether it comes from a point source or a raft of heat sinks. The temperature of the whole server is a function of the power of all the boards and internal heat sinks only protect individual chips.
About the only real issue is in the area very close to a high temperature part. High temperatures can cause mechanical stresses in the solder joints, pcb vias etc. If there is difference in failure patterns then that's the most likely cause.
Slightly over 42 degrees Celsius is enough to make a "Fried Egg".
remember if you have a fever over 42 that is lethal due to proteins start breaking up.
sales ppl ......
190 F ~ 88 C.
Grumpy = Numpty
noroimusha is correct, Protien starts to break up at 42 celcius. you CAN cook an egg at 42C however it would take a very long time.. There is a chef near me who does a slow boiled egg. its cooked at 60C for 5hrs. (Sat Bains if your wondering)
Now back to the article. Yes 88c is hot enough to cook an egg (no not fry it) and yes as you decided to post 190F is approximatly equal to 88C but that's not the point that was being made.
...have nicked the bit from Carry on Screaming where Kenneth Williams meets his end shouting "Frying Tonight!"
Ought to be fairly simple to (dis)prove
Anybody have a QLogic and a tiny frying pan?
That recipe is tasteless
Without some drops of olive oil on the pan prior to the egg drops...
Some bits of dried minced garlic would help too...
Not mentioning that seeing the olive oil boiling would have made the video more spectacular.
Very cute pan by the way, mi 5 year old daughter now wants one.
Why would you
Fry an egg in olive oil? For that you want a relatively flavourless oil like sunflower or vegetable.
Seriously these TV chefs have a lot to answer for. Everyone seems to think olive oil needs to be used for everything these days.
QLogic should take an Emulex board and use the heat-sink to make little bits of toast to go with the fried egg....
..though I did get a lovely taste from my fried eggs using hemp oil. Just a shame the local Tesco doesn't sell it any more!
Already been done...
Ahhh yes you are correct
Except for one tiny little problem, and that is:
I dont believe the USB Toaster was using QLogic chips as its heating element.
They'd need to show that anyone but QLogic actually watched it to the egg-frying bit.
I saw a few seconds presented by a bore with a thick foreign accent, then turned off.
"..You can actually fry an egg on it."
No, you can't. The demo merely shows that part of the PCB gets hot enough to denature the protein bonds of a small amount of egg white. It's not fried (not even cooked) and it's not "an egg". Frying uses oil because the temperatures required exceed that of boiling water.
If you tired to cook a whole egg using this heat source, the volume of egg would probably dissipate the heat and cool the chip like a heatsink. Egg and chips?
I thought the presentation was rather crappy, with its wooden presenter, cliche rock music intro, superbike analogy and bad science.
I wonder if googling Emulex & turbo (or supercharger) would find some of their products.
Sunny side up please!
Where's the bacon?
Chips are not less stable at higher temperatures.
It all depends on the design and the level of testing that's been conducted.
A chip will be designed to run at a certain level of performance, a certain clock frequency.
The higher the clock frequency the more heat that is generated. As part of the original design specification, the chip will be required to work at a certain frequency, and this will be simulated during the design stage, and then tested upon manufacture of the chips.
If the die works at that frequency then it's accepted and packaged.
The issue with over clocking, is that you're deliberately running the design at a higher clock frequency than the chip was originally designed. It may work, it may not, but often does run faster.
When you exceed the manufacturers design requirements, when you clock the chip faster than than it was desiged to run at, then yes, it will become unstable and the manufacturer holds no liability for that, and this can happen at any temperature.
It doesn't matter if the chip runs hot, the question is, are you running the chip outside of its design parameters?
The issue with hot chips is that there is a correlation between the operating temperature of the chip and it's life. The hotter the chip runs, the shorter it's life. And if you're running from a battery, then the shorter the battery life as it uses more power.
If you read my post properly you will see that I said it is sometimes necessary to bring the temperature down to make it run reliably (typically by adding a heatsink). Precisely the opposite of what you are claiming I said. Also read my post and you will see I directly refer to the incfreased longevity that you might get with lower temperature running (although in real life, a server is likely to go obsolete before heat-induced failure of an FC card occurs).
As for all the other stuff in your post, what on earth are you on about battery life on a fibre channel card? Clearly using more power shortens battery life. That's nothing to do with the question here.
None of this hasn't anything to do with a deeply misleading video. And yes, for those that care, you can cook an egg (very slowly) at 190F, but you most certainly can't fry it which, why definition, requires oil heated to near its boiling point.
Granted, I've got a mid-2007 MacBook Pro with the terrific iStat Menus app that continuously displays temperature (and other data) for various components in real time on the menu bar so it's always visible.
When near-idle, the MBP's NVIDIA 8600M-GT (with heat sink) is usually closer to about 138 degrees (Celsius) at the GPU temperature diode and about 118 degrees at the GPU heatsink. I don't recall numbers when the GPU's really active but it's higher. 190 degrees F is only about 86 degress C... so does seem plausible for a GPU without a heat sink.
Just see the funny side....
Dodgy physics be damned its still pretty funny . Yes it was a lawsuit waiting to happen but as geek jokes go its pretty cool .
... I am insulted that the people issuing the law suit think I am dumb enough to believe that the chip is larger than a frying pan,,,
Do all of the Electrical Engineers at Emulex have webbed hands? (Watch the video again for the awesome green-screen fail)
Webbed hands, (fake) frying eggs in miniature chip-pans and goofy science. Do they have umpa-lumpas in their factory as well?
You obviously didn't read my post correctly. Go back and re-read it.
I actually said "if you're running from a battery". I didn't say the fibre card was running from a battery. I was talking in general terms, about power consumption and heat.
Anyone that knows anything about energy would realise that heat energy is wasted energy, which is why in battery powered systems you don't want to chips running hot.
And stop being so arrogant.
"Precisely the opposite of what you are claiming I said"
What are you saying which I said was the opposite of what you claimed?
I have re-read my first post and it made no claim about anything you said at all. You are mistaken.
Go back and re-read my original post.
You said in your first post, and I paste it here to remind you: " A more likely reason is to improve reliability of operation as the chip is unstable at higher temperatures (and, at extremes, it will improve longevity)."
My response to you was quite clearly, in my first sentence: "Chips are not less stable at higher temperatures."
I then went on to explain flaws in your original statement.
I standby what I said. And I used to be a former chip designer. On CMOS, GaAs technologies.
Heatsinks are used to cool chips, but they're used to bring the temperature down to ensure the chip works within its design specification.
Saying that heat makes a chip unreliable doesn't present the full picture.
During simulation of a device before manufacture, different delay multipliers are used on the propagation delay of the logic gates to simulate changes in supply voltage and temperature.
The power dissipation of a device can be calculated and then a decision made as to what kind of heat sink to fit to ensure the required junction temperature is not exceeded.
If you, as the user, or a board designer, want to then run that chip at a higher frequency, which results in greater heat generation, then yes, the propagation delays of the gates will be higher than the original design and the circuit will become unstable. The basic issue here, is you're using the chip in a way which is outside of it's original design parameters, and it can happen at any temperature.
I know someone who fried a full sized egg on an old P3, it took about 5 minutes but it cooked it
Emulex engineers ?
- don't know the difference between molecules and atoms.
- data does not 'flow' . Electrons flow !
i switched it of after that.
Besides, electrons become more mobile as heat increases.
If that is the quality of engineers at Emulex i'll take my breakfast at Qlogic anytime.
physics fail @vincent himpe
yes, electrons gain more kinetic energy at higher temperatures, but will be scattered more frequently from electron phonon interactions (as temperature will increase the KE of the crystal lattice also), effectively reducing their mean free path within the bulk of the conductor and increasing resistance.
Electrons don't flow either, but electron-hole pairs do...
How long until...
I can get that nifty egg-based cooling system for my PC? Then I shall laugh at the other nerds with their old-fashioned water-based cooling systems.
My understanding of board failure ...
Surely the operating temperature of a single chip / component is not an issue. Isn't it high temperature differentials between components resulting in thermo-mechanical stress on the board that leads to failure?
Byte and switch
Did anyone else notice that the chip which was supposedly dissipating 190F of heat was an AMCC PHY not the Qlogic comms processor that they then proceeded to "fry" and egg on?
I wonder what PHYs are hidden under the Emulex heatsink? There seems to be a certain amount of heat escaping from under their heatsink ;)
I need to get out more :(
42 degrees is enough to scramble your backup diskettes... according to a data fire-safe advert that I think I remember.
Heat is work, so the hotter the computer equipment gets, the more work it's doing. You should buy the hottest-running servers you can find. Although electrical resistance increases with heat, too. Unless that's different for semiconductors.
"Hitler Has Given the Jews a City" was a great ad, too, for his relocation service. He had good people doing public relations. Wait. That's not right. Evil people. That's it. He had evil people.
Did he say???
Did he say "that's where the temperature comes from"?
Also, I recall a story about a pizza baking on an old ZX spectrum that was used in a pizzeria.
But, surely the real test would be to cook the egg yoke, since the protein molecules in the white unravel at a lower temperature?
Same could be said of ATI cards as they all run very hot!
Will it blend?
I'm not convinced.
I say we take the whole thing with a pinch of salt ...