This is interesting. Why hire young? Whats wrong with older, more experienced programmers? I can see the draw towards youthful exuberance, but no one in my company under the age of 40 touches our main programming components.
Hiring young is primarily due to using internships/coops as evaluation periods - if there was a way to do the same type of thing for experienced hires, I'd jump on it, but people who have been in the workforce for any extended period of time are unlikely to agree to come work with us for 6 months, and we'll offer you a job at the end if we think you're worth it.
Poaching brings another question, if you don't mind. I was poached, and it changed my opinion a bit on company to employee loyalty and vice versa. Is there a concern, that if your newly poached employee could be poached right back out again by a different company?
Unfortunately, there's no reason that they couldn't be poached right back out by a different company. On the other side, that goes for ALL the programmers you have, so you should already be investing in ways to do whatever you can to improve employee retention, boost work morale, etc.
Honestly - imo, you are incredibly fortunate to hire excellent experienced people based off interviews (our hit rate is about 25% good, 50% passable, 25% poor)
the 2 best strategies for having a high hit rate with your new hires:
1. hire young - bring people in as interns/coops and use their term as a 6 month interview - this can give you a great insight into their potential
2. poach - has anyone else in your organization worked somewhere else? find out if there are any excellent people from previous jobs looking for work
Does an IQ of 102 really provide that much of an advantage over someone with an IQ of 100?
In a standard IQ test, since the standard deviation is 15 points, a 2 point increase from the mean would indicate an additional 3-4% of the population you were smarter than - I'd consider that pretty significant...
single DLP chip, it's almost the same as the color wheel - you alternate red, green, blue
The LED is able to switch a little faster than you could get with the color wheel, which should pretty much eliminate any rainbow effect
To me, the big selling point on the LED is the much lower power usage (thus lower cooling requirements), and the fact that the LED displays should get close to 20k hour lifetime, compared to 3-4k for the current bulbs
DLP does this with a color wheel rotating anywhere from 10,000 to 30,000 times per second
FYI, it's more on the order of 100 rotations/sec
The reason the DMD does not suffer burn-in is the fact that the hinges that the mirrors are mounted on do not suffer from the same type of repetitive stress wearout that a larger hinge would, not anything due to the color wheel - top end LCDs and DLPs actually have one light source per primary color
USIV is the replacement for the USIII, upgrading an USIII system is as easy as calling Sun and getting them to swap in the USIVs (very similar to replacing a single core athlon with a same-socket dual core athlon, except the motherboards get swapped out, not just the processors)
Unfortunately, the article is talking about the USIIIi+ being cancelled, which was the upgrade for the USIIIi
In the server space, Niagara based systems are the replacement for the USIIIi, but there is not really a replacement on the workstation side (they could always try to wring more speed out of the USIIIi, or go in with some kind of opteron solution)
Electronic components conduct electricity better (or worse, in the case of many semiconductors) at lower temperatures, so even in a humidity-controlled environment, you could end up melting certain components.
better conductivity would lead to most parts running on less power than they would at high temp...
manufacturers over ride methodology (I.E. a prius allows a certain pattern of engaging and disengaging the parking brake to over ride the security system and other systems will be disabled by simply removing a specific fuse from underhood)
The over ride pattern for every RFID enabled car I have ever looked at, in addition to opening the passenger door, pushing the brake twice and honking, required that a master key be in the ignition (and usually the car switched on) during the sequence...
otherwise, it'd be fairly simple to open a car, program a new master key, and drive off
I've actually put a bit of thought into building a system like this - I'm going to try to cover the front quarters with small cameras in the headlights (well, actually under the headlight lens), cover the back with a couple of cameras mounted in the back grille, and have one more somewhere in the interior that gets a shot of the gauges
I figured I could have it only save ~20-30 minutes of footage on each camera, and simply have a switch on the dashboard that I can use to stop capture
Since ticketmaster handles the ticket sales for a lot of the festivals (well, the ones in the states anyway) you'll still be stuck paying 15-25% surcharges on all the tickets
Maybe you should get your tickets earlier as then there are no last minute fees. Shipping and handling charges make sense, especially as venue pickup and international deliveries make these costs variable.
I just attended the coachella valley music and arts festival - I ordered my tickets 2.5 months in advance, and paid no shipping charges - however, I ended up paying $386 for 2 tickets that were listed at $160/each - fees made up 17% of the cost of my tickets
Perhaps you should attend Celine Dion concerts instead of rock concerts.
Couldn't agree more - at 2 of the last 3 concerts I've been to, I've thought that it need to be a lot louder.
If you queue early you'll have an excellent view from the very front row (assuming a general admission show, which most rock concerts are). The bigger the crowd, the least I expect from venue facilities. Perhaps you should stay clear of large public events though, if crowds piss you off.
I don't think you could really do a show at a staduim as a general admission show... In any case, those floor tickets will now be going for 20x what they were previously due to this auctioning scheme
There are still many, many shows to be seen in the $5-$15 range. Perhaps not mainstream artists, but you can't then compare to Oingo Boingo who haven't had a chart hit ever.
A lost of this depends on where you live. I live in Houston, my younger sister lives in Lubbock - we went and saw the exact same bands on 2 consecutive nights - she paid $5, I paid $25. In the past 3 years, I think I've only been to 2 concerts with ticket prices under $20 (and it was pushed to well over $20 after fees).
I think we can all agree that TicketBastard is just trying to suck the last few dollars out of the consumers that can afford it...
If the on chip SRAMs work, you should get something like 90% fallthrough from that point (i.e. if SRAMs yield 50%, you should have something like 45% overall product yield)
On a related note, every major chip manufacturer has redundant resources embedded in the SRAMs so that they can repair chips that are not nominally good (trust me, the probability of making a nominally good part is VERY low on any fairly modern process)
The process designation refers to the the distance between the source and drain in the FETs (transistors) on a processor. Keep in mind that this distance is by no means the smallest thing in the processor - the actual gate oxide layer is tiny by comparison, with Intel's 65nm process having only 1.2nm of the stuff. That's less than 11 atoms thick.
Not quite correct - the process designator refers to the size of a minimum-width metal line, the physical gate length is usually significantly smaller
130nm chips that I have either worked on or seen have gate widths ranging from 50nm - 90nm
At 90nm, this has shrunk to 35-60nm or so
system ram is already dram, which is also single transistor memory - not nearly as much size savings, and if the dynamic power requirement rises at all, system manufacturers would probably be slow to adopt
probably some form of all the following in the test/debug block
JTAG
SCAN/ATPG controller (can be inplemented in the JTAG controller)
DMA controller
BIST controller
EFuses
Ring Oscillators, although those are probably spread all over the chip, not just in test/debug
If you want more info, I can look around for some resources.
Actually, it can explode - don't actually do this, if you have a good enough power supply and capacitor you can get an explosion that's about as powerful as a hand grenade
considering the fact that static (leakage) power should be the real killer in microprocessors is the dynamic power - for wireless/dsp/ucontroller type applications, it could be pretty huge
honestly, the article has so few details it's impossible to tell what they're really doing, but i am pretty sure that most companies out there already use RTP on there gate oxide...
Electrically programmable fuses make this very simple - when the part is tested at wafer/multiprobe, you simply blow in the ID when you are blowing in all your repair solutions - I can guarantee IBM is blowing an ID into the parts anyway for general yield/return tracking purposes.
This ID can probably be accessed through the JTAG port, or accessed internally - the data is going to be in a certain format (Lot #, wafer #, x coord, y coord, or something similar) that would be easy to verify...
You could also make it so reading the id from one place and writing it to another was part of the reset sequence on the chip...
WRT getting the serialid out of the processor, you should be able to read it out through a simple JTAG instruction
you can buy DC power converters that are very efficient - IIRC, buck converters can be >>90% efficient, and you can get very clean power out of them (it's not really any harder than gettiing clean power out of a rectifier), and with IGBT's, you can push the switching frequency fairly high.
Altough you would need larger wiring, removing one full AC -> DC -> AC cycle could easily pay for itself in power savings.
FYI, low powered RFID devices don't need any battery, so they should last pretty much forever - the transponders are powered by induction from the scanner
MTTF on the circuit is probably in the dozens (if not the hundreds) of years
Have you ever tried to get in a dorm if/when the power happens to go out?
[I was locked once locked out of my dorm for a few hours in freezing weather when the power went out halfway through my freshman year]
Slashdot Mirror
It's a more common practice than you might think
This is interesting. Why hire young? Whats wrong with older, more experienced programmers? I can see the draw towards youthful exuberance, but no one in my company under the age of 40 touches our main programming components.
Hiring young is primarily due to using internships/coops as evaluation periods - if there was a way to do the same type of thing for experienced hires, I'd jump on it, but people who have been in the workforce for any extended period of time are unlikely to agree to come work with us for 6 months, and we'll offer you a job at the end if we think you're worth it.
Poaching brings another question, if you don't mind. I was poached, and it changed my opinion a bit on company to employee loyalty and vice versa. Is there a concern, that if your newly poached employee could be poached right back out again by a different company?
Unfortunately, there's no reason that they couldn't be poached right back out by a different company. On the other side, that goes for ALL the programmers you have, so you should already be investing in ways to do whatever you can to improve employee retention, boost work morale, etc.
Honestly - imo, you are incredibly fortunate to hire excellent experienced people based off interviews (our hit rate is about 25% good, 50% passable, 25% poor)
the 2 best strategies for having a high hit rate with your new hires:
1. hire young - bring people in as interns/coops and use their term as a 6 month interview - this can give you a great insight into their potential
2. poach - has anyone else in your organization worked somewhere else? find out if there are any excellent people from previous jobs looking for work
Does an IQ of 102 really provide that much of an advantage over someone with an IQ of 100?
In a standard IQ test, since the standard deviation is 15 points, a 2 point increase from the mean would indicate an additional 3-4% of the population you were smarter than - I'd consider that pretty significant...
single DLP chip, it's almost the same as the color wheel - you alternate red, green, blue
The LED is able to switch a little faster than you could get with the color wheel, which should pretty much eliminate any rainbow effect
To me, the big selling point on the LED is the much lower power usage (thus lower cooling requirements), and the fact that the LED displays should get close to 20k hour lifetime, compared to 3-4k for the current bulbs
DLP does this with a color wheel rotating anywhere from 10,000 to 30,000 times per second
FYI, it's more on the order of 100 rotations/sec
The reason the DMD does not suffer burn-in is the fact that the hinges that the mirrors are mounted on do not suffer from the same type of repetitive stress wearout that a larger hinge would, not anything due to the color wheel - top end LCDs and DLPs actually have one light source per primary color
USIV is the replacement for the USIII, upgrading an USIII system is as easy as calling Sun and getting them to swap in the USIVs (very similar to replacing a single core athlon with a same-socket dual core athlon, except the motherboards get swapped out, not just the processors)
Unfortunately, the article is talking about the USIIIi+ being cancelled, which was the upgrade for the USIIIi
In the server space, Niagara based systems are the replacement for the USIIIi, but there is not really a replacement on the workstation side (they could always try to wring more speed out of the USIIIi, or go in with some kind of opteron solution)
Electronic components conduct electricity better (or worse, in the case of many semiconductors) at lower temperatures, so even in a humidity-controlled environment, you could end up melting certain components. better conductivity would lead to most parts running on less power than they would at high temp...
manufacturers over ride methodology (I.E. a prius allows a certain pattern of engaging and disengaging the parking brake to over ride the security system and other systems will be disabled by simply removing a specific fuse from underhood)
The over ride pattern for every RFID enabled car I have ever looked at, in addition to opening the passenger door, pushing the brake twice and honking, required that a master key be in the ignition (and usually the car switched on) during the sequence...
otherwise, it'd be fairly simple to open a car, program a new master key, and drive off
they used all the good names 20 years ago...
I've actually put a bit of thought into building a system like this - I'm going to try to cover the front quarters with small cameras in the headlights (well, actually under the headlight lens), cover the back with a couple of cameras mounted in the back grille, and have one more somewhere in the interior that gets a shot of the gauges
I figured I could have it only save ~20-30 minutes of footage on each camera, and simply have a switch on the dashboard that I can use to stop capture
Since ticketmaster handles the ticket sales for a lot of the festivals (well, the ones in the states anyway) you'll still be stuck paying 15-25% surcharges on all the tickets
Maybe you should get your tickets earlier as then there are no last minute fees. Shipping and handling charges make sense, especially as venue pickup and international deliveries make these costs variable.
I just attended the coachella valley music and arts festival - I ordered my tickets 2.5 months in advance, and paid no shipping charges - however, I ended up paying $386 for 2 tickets that were listed at $160/each - fees made up 17% of the cost of my tickets
Perhaps you should attend Celine Dion concerts instead of rock concerts.
Couldn't agree more - at 2 of the last 3 concerts I've been to, I've thought that it need to be a lot louder.
If you queue early you'll have an excellent view from the very front row (assuming a general admission show, which most rock concerts are). The bigger the crowd, the least I expect from venue facilities. Perhaps you should stay clear of large public events though, if crowds piss you off.
I don't think you could really do a show at a staduim as a general admission show... In any case, those floor tickets will now be going for 20x what they were previously due to this auctioning scheme
There are still many, many shows to be seen in the $5-$15 range. Perhaps not mainstream artists, but you can't then compare to Oingo Boingo who haven't had a chart hit ever.
A lost of this depends on where you live. I live in Houston, my younger sister lives in Lubbock - we went and saw the exact same bands on 2 consecutive nights - she paid $5, I paid $25. In the past 3 years, I think I've only been to 2 concerts with ticket prices under $20 (and it was pushed to well over $20 after fees).
I think we can all agree that TicketBastard is just trying to suck the last few dollars out of the consumers that can afford it...
there is a FPU per core and the power for niagara2 is still supposed to be remarkably low
I think there are 2 things you should realize.
If the on chip SRAMs work, you should get something like 90% fallthrough from that point (i.e. if SRAMs yield 50%, you should have something like 45% overall product yield)
On a related note, every major chip manufacturer has redundant resources embedded in the SRAMs so that they can repair chips that are not nominally good (trust me, the probability of making a nominally good part is VERY low on any fairly modern process)
The process designation refers to the the distance between the source and drain in the FETs (transistors) on a processor. Keep in mind that this distance is by no means the smallest thing in the processor - the actual gate oxide layer is tiny by comparison, with Intel's 65nm process having only 1.2nm of the stuff. That's less than 11 atoms thick.
Not quite correct - the process designator refers to the size of a minimum-width metal line, the physical gate length is usually significantly smaller
130nm chips that I have either worked on or seen have gate widths ranging from 50nm - 90nm
At 90nm, this has shrunk to 35-60nm or so
system ram is already dram, which is also single transistor memory - not nearly as much size savings, and if the dynamic power requirement rises at all, system manufacturers would probably be slow to adopt
probably some form of all the following in the test/debug block
JTAG
SCAN/ATPG controller (can be inplemented in the JTAG controller)
DMA controller
BIST controller
EFuses
Ring Oscillators, although those are probably spread all over the chip, not just in test/debug
If you want more info, I can look around for some resources.
and hook it up backwards...
nothing bad will happen...really
Actually, it can explode - don't actually do this, if you have a good enough power supply and capacitor you can get an explosion that's about as powerful as a hand grenade
Now if someone could create a extremely high density reprogrammable chip (500M gates) then all bets are off.
Good luck building an UST-1 in on a 500M gate ASIC - the chip is close to that xtor count as designed...
considering the fact that static (leakage) power should be
the real killer in microprocessors is the dynamic power - for wireless/dsp/ucontroller type applications, it could be pretty huge
honestly, the article has so few details it's impossible to tell what they're really doing, but i am pretty sure that most companies out there already use RTP on there gate oxide...
Electrically programmable fuses make this very simple - when the part is tested at wafer/multiprobe, you simply blow in the ID when you are blowing in all your repair solutions - I can guarantee IBM is blowing an ID into the parts anyway for general yield/return tracking purposes.
This ID can probably be accessed through the JTAG port, or accessed internally - the data is going to be in a certain format (Lot #, wafer #, x coord, y coord, or something similar) that would be easy to verify...
You could also make it so reading the id from one place and writing it to another was part of the reset sequence on the chip...
WRT getting the serialid out of the processor, you should be able to read it out through a simple JTAG instruction
you can buy DC power converters that are very efficient - IIRC, buck converters can be >>90% efficient, and you can get very clean power out of them (it's not really any harder than gettiing clean power out of a rectifier), and with IGBT's, you can push the switching frequency fairly high.
Altough you would need larger wiring, removing one full AC -> DC -> AC cycle could easily pay for itself in power savings.
FYI, low powered RFID devices don't need any battery, so they should last pretty much forever - the transponders are powered by induction from the scanner
MTTF on the circuit is probably in the dozens (if not the hundreds) of years
Have you ever tried to get in a dorm if/when the power happens to go out?
[I was locked once locked out of my dorm for a few hours in freezing weather when the power went out halfway through my freshman year]
oops...
:P
guess that's why I always run it on a test case before I run it on the full file