Jesus, please stop pulling figures out your arse. Have you seen what kind of GPU power $200 +/- $50 buys these days?!?!? Take your pick - 9600GT, 8800GT, 8800GTS (G92 and G80), 3870, 3850, etc.
Exactly. Nothing rings quite as true as "time is money" in IC manufacturing. Hell, we're even cutting as many tests as possible (and scaling down the ones we do run) during testing just to shave fractions of seconds off time on ATEs.
One of the big things in VLSI is not reinventing the wheel. Lots of companies have labs dedicated to reverse engineering competitor's products. Eg. TI acquired Burr-Brown precisely because they couldn't reverse engineering / produce a particular data converter;)
David Bishop (Head of Microstructure Physics Research Dept., AT&T Bell Labs)
"In today's world, if you have a customer with a problem, he doesn't really care if it's a physics, engineering, or software problem. He just wants it to be fixed."
I beg to differ. It really seems to need shader power as using hacked Very High settings results in slowdowns in the ice cave and final carrier levels on an overclocked 8800GTS 320MB (backed by a 3.6GHz C2D E6600 and 4GB RAM).
Don't even need to resort to that. Just keep an eye out for the sometimes insane after-rebate deals on HDDs, optical drives, motherboards, RAM, vid cards, etc.
That's true. Plus even an enthusiast rig isn't all that expensive. Some rough workings :
ABIT IP35 ($80), C2D E6420 ($190), 4GB DDR2-667 ($40), 8800GT 512MB (~$250), SomeCase (~$50), Corsair HX520W PSU (~$70), SomeHDD~250GB (~$60), SomeBurner (~$30), SB X-Fi XtremeMusic (~$55), 22" LCD ($179), SomeSpeakers (~$100), Logitech G15 / Saitek Eclipse II (~$40), SomeMouse (~$40)
Grand Total = $1184 for a powerhouse rig (once you OC that E6420).
Let's save some green : E4500 (-$100), Ultra X-Finity 500W PSU (-$55, these go cheap pretty often and are solid PSUs) = ~$1000.
Guessed as much, although I think you would have trouble pushing Crysis at 1920x1200 at even medium settings (trouble = maintaining ~30fps). That game eats even an 8800Ultra for lunch:(
That said, you are missing out on dual/quad-cores (overall system is more responsive) and have a dead-end in terms of graphics upgradability (maybe okay if you're waiting on the overdue refresh from nVIDIA and ATi). What with the dirt cheap prices of DDR2-667 (search around for the $8.95/1GB HP stick deal) and (relatively) sky-high price of DDR1, it might be a worth it for you to switch to say an overclocked E4300/E4500 setup. Dead easy to hit 3GHz with half-decent air cooling.
I'm waiting for neural implants. Then we can finally do away with tiny-a*s screens and input devices.
But God forbid that we should get g0atse or tubg*rl transmitted right into our brains *shudders*
In applications like MRI where you need not just high-intensity but also high-precision, you'll typically have a main superconducting coil and a couple of ancillary coils to iron out the kinks in the field pattern (eg. tesseral and shim coils).
Also depends on the frequency of operation; get high enough and the losses due to skin and proximity effects become unbearable (stuff like Litz wires and appropriate coil turns arrangement can help to mitigate this).
One more thing to worry apart with such high fields involved is whether the darned thing is going to fly apart when you hit the ON button. Lorentz forces are a bitch (I've looked at a group's work where they generate a 50T field using a single-turn micro-coil. They kinda cheat though - it's only a 40ns pulse and the single-turn coil literrally blows up after a couple of repetitions).
Come on. You have to admit that they occasionally have kick ass deals, like during the last BF. I have yet to see a 22" WS LCD with a multitude of inputs (VGA, DVI, component, composite) for $189 without rebates. And where would I get my "cheap" Verbatim dual-layer media? (quite often available for ~$1.50/disc, sometimes less if they have their 12% customer appreciation coupons out) There was also a nice deal for external WD 750GB HDDs (internally SATA SE16s) about 2 months back.
Heck, it may be that the branch at 50014 is just more decent that average.
Aaack... forgot to mention amps as well. Classical stuff needs one heck of a large dynamic range (~120dB IINM). Just designing op amps to reach that is a wicked process, especially with all the digital folks wanting smaller min feature sizes (bad lambda degradation) and lower supply voltages (excess bias voltage, where art thou?).
Should probably also mention quantization noise. Another factor is data converter performance. As you approach the Nyquist rate, the effective number of bits (ENOB) will drop off due to unwanted transients (toggling current sources / cap banks, etc.) Just because it has 24-pins, don't mean that has a 24 ENOB. Heck, at modern voltage levels (~1.2V), farting on the data converter would produce a mountain of LSBs of error (1.2V / 2^24).
We should all thank the folks who buy the top-end card. Who do you think effectively subsidizes the value cards for us? It's especially sweet when you can get a mid-range / mid-high-end card that comes close to an unoverclocked top-end card (eg. the 9800 Pro, GeForce4 Ti4400, etc.)
Slashdot Mirror
Free space would be quite a pain. You'd need to collimate the beam, worry about acceptance angles, mode field diameters, etc.
I'd like their prefer some t/F ratio effects test (together with alpha and beta).
Jesus, please stop pulling figures out your arse. Have you seen what kind of GPU power $200 +/- $50 buys these days?!?!? Take your pick - 9600GT, 8800GT, 8800GTS (G92 and G80), 3870, 3850, etc.
Exactly. Nothing rings quite as true as "time is money" in IC manufacturing. Hell, we're even cutting as many tests as possible (and scaling down the ones we do run) during testing just to shave fractions of seconds off time on ATEs.
One of the big things in VLSI is not reinventing the wheel. Lots of companies have labs dedicated to reverse engineering competitor's products. Eg. TI acquired Burr-Brown precisely because they couldn't reverse engineering / produce a particular data converter ;)
David Bishop (Head of Microstructure Physics Research Dept., AT&T Bell Labs) "In today's world, if you have a customer with a problem, he doesn't really care if it's a physics, engineering, or software problem. He just wants it to be fixed."
If you'd like, I could sell you a S3 Virge DX PCI for $200. Don't mind the 3D deceleration ;)
I beg to differ. It really seems to need shader power as using hacked Very High settings results in slowdowns in the ice cave and final carrier levels on an overclocked 8800GTS 320MB (backed by a 3.6GHz C2D E6600 and 4GB RAM).
Yes it does. Greedo shot first! :P
Because I actually need the speeed? Core 2 E6420 @ 3.4GHz here.
Don't even need to resort to that. Just keep an eye out for the sometimes insane after-rebate deals on HDDs, optical drives, motherboards, RAM, vid cards, etc.
Any you would be nuts to get that X1950Pro when the 8800GT assr*pes it for not much more ;)
That's true. Plus even an enthusiast rig isn't all that expensive. Some rough workings :
ABIT IP35 ($80), C2D E6420 ($190), 4GB DDR2-667 ($40), 8800GT 512MB (~$250), SomeCase (~$50), Corsair HX520W PSU (~$70), SomeHDD~250GB (~$60), SomeBurner (~$30), SB X-Fi XtremeMusic (~$55), 22" LCD ($179), SomeSpeakers (~$100), Logitech G15 / Saitek Eclipse II (~$40), SomeMouse (~$40)
Grand Total = $1184 for a powerhouse rig (once you OC that E6420).
Let's save some green : E4500 (-$100), Ultra X-Finity 500W PSU (-$55, these go cheap pretty often and are solid PSUs) = ~$1000.
Guessed as much, although I think you would have trouble pushing Crysis at 1920x1200 at even medium settings (trouble = maintaining ~30fps). That game eats even an 8800Ultra for lunch :(
That said, you are missing out on dual/quad-cores (overall system is more responsive) and have a dead-end in terms of graphics upgradability (maybe okay if you're waiting on the overdue refresh from nVIDIA and ATi). What with the dirt cheap prices of DDR2-667 (search around for the $8.95/1GB HP stick deal) and (relatively) sky-high price of DDR1, it might be a worth it for you to switch to say an overclocked E4300/E4500 setup. Dead easy to hit 3GHz with half-decent air cooling.
I'm waiting for neural implants. Then we can finally do away with tiny-a*s screens and input devices.
But God forbid that we should get g0atse or tubg*rl transmitted right into our brains *shudders*
Must have swapped out a lot of guts. PCIe, DDR2, (maybe) SLi, beefier PSU, etc. ;)
In applications like MRI where you need not just high-intensity but also high-precision, you'll typically have a main superconducting coil and a couple of ancillary coils to iron out the kinks in the field pattern (eg. tesseral and shim coils).
Also depends on the frequency of operation; get high enough and the losses due to skin and proximity effects become unbearable (stuff like Litz wires and appropriate coil turns arrangement can help to mitigate this).
One more thing to worry apart with such high fields involved is whether the darned thing is going to fly apart when you hit the ON button. Lorentz forces are a bitch (I've looked at a group's work where they generate a 50T field using a single-turn micro-coil. They kinda cheat though - it's only a 40ns pulse and the single-turn coil literrally blows up after a couple of repetitions).
Dead easy. Magneto-optic effects (Faraday, Kerr, Zeeman, circular birefringence, etc.)
The group I'm involved in is doing some work in applying the Faraday rotation effect :
http://jwtioh.bluesonic.net/files/04202947.pdf
http://jwtioh.bluesonic.net/files/01704539.pdf
Never tried Gordian Knot? http://www.doom9.org/
They do bother trying. Why do you think WHCL exists?
Come on. You have to admit that they occasionally have kick ass deals, like during the last BF. I have yet to see a 22" WS LCD with a multitude of inputs (VGA, DVI, component, composite) for $189 without rebates. And where would I get my "cheap" Verbatim dual-layer media? (quite often available for ~$1.50/disc, sometimes less if they have their 12% customer appreciation coupons out) There was also a nice deal for external WD 750GB HDDs (internally SATA SE16s) about 2 months back.
Heck, it may be that the branch at 50014 is just more decent that average.
Aaack... forgot to mention amps as well. Classical stuff needs one heck of a large dynamic range (~120dB IINM). Just designing op amps to reach that is a wicked process, especially with all the digital folks wanting smaller min feature sizes (bad lambda degradation) and lower supply voltages (excess bias voltage, where art thou?).
Should probably also mention quantization noise. Another factor is data converter performance. As you approach the Nyquist rate, the effective number of bits (ENOB) will drop off due to unwanted transients (toggling current sources / cap banks, etc.) Just because it has 24-pins, don't mean that has a 24 ENOB. Heck, at modern voltage levels (~1.2V), farting on the data converter would produce a mountain of LSBs of error (1.2V / 2^24).
We should all thank the folks who buy the top-end card. Who do you think effectively subsidizes the value cards for us? It's especially sweet when you can get a mid-range / mid-high-end card that comes close to an unoverclocked top-end card (eg. the 9800 Pro, GeForce4 Ti4400, etc.)
Agreed. That's why you wait for awesome deals like back in summer (PNY 8800GTS 320MB for ~$205).