Unfortunately, while the NVS series bare the Quadro branding, NVidia does not support the professional/scientific feature sets on those chips. So, features like the unified back buffer, etc. are not available. Essentially, the NVS450 is a card with two GeForce 8400 chips and a PCI-E to PCI-E bridge. It's kinda lame.
NVidia marketing material suggests that the NVS line is intended for business users who need to support many displays without any advanced rendering.
While you're right, I imagine the NVS450 costs more than a pair of GTX220 or GT650 cards; he'd be better served with your suggestion than the NVS card. Personally, I suspect his desktop is a Dell or HP professional workstation as they generally ship with NVS graphics as the entry-level video solution. I doubt he specifically chose the card.
Also, the best solution to his dilemma, IMHO, is the Matrox DualHead2Go or TripleHead2Go. I know it seems like having more GPUs would be a better solution, but I think less GPUs means less overhead in synchronization, mutexes, locks, etc. That's just a hypothesis... no data to back it up.
All power rails appear to be exposed. While they are on the back, this could be a significant safety (personnel and/or fire) issue. Considering that you can up to 500A @ 12.5V DC running through the zone power rails, and potentially more for the main cabinet DC power rails, exposed seems like a bad idea.
Your assertion that you'd save "way more" by switching to SSD storage is assuming that the spindle disks are the main consumer of current.
According to WD, the WD20EARX draws 5.3W during read/write, 3.3W during idle, and 0.7W standby/seep(which, admittedly is a rare situation in datacenters.) (from the WD20EARX datasheet)
According to Intel, the Intel 910 series SSD draws up to 25W while active and 8W while idle. The Intel 520 series SSD draws 850mW active and 600mW idle. (from the Intel 520 series product specifications.) I don't know if those numbers for the 910 are a typo, because it seems weird that they'd exceed a mechanical drive.
Either way, my point is that the WD's have a power ratio of 2.65W/TB and the Intel 520 SSDs have a power ratio of 1.78W/TB. Which means that switching to SSDs will save you 33% on your storage power needs. Thing is, because the SSDs have less capacity per SATA port, once you factor in the extra necessary RAID controllers, SATA cards or SATA port expanders, the percent power saving will drop. Admittedly, I have no idea by how much.
I guess, my point is to challenge the popularly regarded idea that mechanical harddrives are extremely power hungry. While CPU efficiency has improved considerably in recent years, I hold that CPUs and associated electronics consume a much larger portion of a server's power than commonly believed.
Also, at idle, the WD consumes 1.65W/TB and the intel consumers 1.45W/TB. Then again, it's not a fair comparison because the SSD can switch between idle and active far more quickly than the mechanical drive. So, once you consider more of aspects of the situation, things become less clearly cut.
The facebook hinged storage server must be using their new 21" rack because they (from images) appear to have an arranged the drives in three rows of 5 drives. The 3.5" drive formfactor is 4" wide, meaning that the enclosure must be at least 20" wide to accomodate five drives per row. Also, using their new rack concept, their servers don't include and AC power supply. So, it's not exactly as space efficient when you factor in the 2U power supply at the bottom. With one PS and one 30drive facebook server, you're at 30drives for 4U or an efficiency of 7.5drives/U. One PS and two 30drive servers, you're at 60drives for 6U and an efficiency of 10drives/U. One PS and three 30drive servers, 90 drives on 8U and an efficiency of 11.25 drives/U. At four servers on one PS unit, you've got 120 drives occupying 10U for an efficiency of 12 drives/U. So, once you have four servers together with the associated PS, you finally reach the efficiency of a thumper.
The thumper (Sun x4500 and x4540) had 48 3.5" HDD's, 2 (x4500) or 3 (x4540) 800W/1600W (110VAC or 220VAC) power supplies and an adorable, itty-bitty dual opteron server. 48 drives occupying 4U is an efficiency of 12drives/U.
To be fair, while the 4server, 1power supply configuration only equals the storage density of the thumper, it has better server/cpu/nic density.
As an aside, the full rack setups appear to have three power supply units. Assuming/guessing 42U per rack with 6U devoted to PS, it leaves 36U divided into three bays of 12U. So, with a PS and five facebook hingy servers occupying 10U and sporting 150 drives at 15 drives/U, you finally outdo the thumper.
To be clear, I do believe that there are benefits to the proposed new rack size, but I don't think it's a clear improvement. Personally, I think the thumper design was brilliant. The only purpose of this reply was to point out that it's not as simple as 15 drives per U.
On a separate note, the ability to fit 5 drives side-by-side in a 21" rack is the best justification I've seen, so far, for widening to 21".
Wow, I prefix too many of my comments with insecure clauses devoid of information and only serving to indirectly apologize to the reader for supplying information I think is important for them to understand despite my worry that I'm trying their patience. If you read all of this reply including even this sudden instrospective insight to my character; then, thank you. I'm flattered.
Nope, they increase a "U" to be 48mm from 44.45mm. This is now called an OU. So, 1OU = 48mm or an increase of 8% compared to a regular U. They claim that this 3.55mm increase will "increases airflow, improving air economization; it also allows for better for cable and thermal management and efficient use of space." Personally, I question wether the increase in airflow, cable management, and efficient use of space will be significant. I'd be very keen to see a good example of how these new 48mm rack units will improve cable management.
Also, the bus bars depicted in the photos appear to be incredibly vulnerable to accidental short circuiting.
Have you read the email shown in the image from the first link(threatpost.com)? It's dated 2003 and it's describing how to optimize the thread local storage local descriptors introduced to linux around that time. If the source code is related to that, then it's likely irrelevant at this point. A lot has happened in the past 9 years.
Uhh, Hi. Yeah. Sorry, The 12cm/13cm band is 2.3-2.31Ghz and 2.39-2.45Ghz. It doesn't overlap the entire ISM allocation. People are still welcome to use channel 11 on their wifi routers... or use those fancy 5GHz radios.
Why are you replacing the motherboards yourself? With my T61p, when something in it died - motherboard needed replacing - I called up IBM and told em it's broke. Something like 19 hours later, DHL has a box for me at my door to ship the laptop out in. I put the laptop in the box and call up DHL to schedule picking up the shipping box. The same DHL guy is back 15 minutes later and takes the box. 23 hours later, same DHL delivery guy is back on my doorstep with my repaired laptop. This was with the standard warranty option when buying the laptop. My mind was blown on just how quickly it got fixed. Apparently got shipped from MI to Memphis, repaired, and shipped back in less than 24 hours.
Does this level of support not exist anymore? Otherwise, why are you replacing the motherboards in house - especially if you don't have spare parts readily available. Also, your complaint about the 44 screws? I mean, come'on, tieing the laces on my shoes takes 10 times longer than using velcro, but it's really kinda not a very big deal.
The difference between chernobyl's RBMK design and and our operating relics is already rather significant. Also, we have organizations in the US, such as the United States Navy, which are at the forefront of safe reactor design and operation.
Noone's surprised. Marvell's had a track record of faulty, ill-documented, or bug-plagued parts. Sometimes, I wonder why they still bother. I suppose, someone has to make Realtek look good.
Only 1, and he did it because he could keep the loud drive cabinet in the basement and use the drives from his bedroom. It was decent gear, too, LSI PCI-E 4GBps card with two multimode transceivers to a JBOD cabinet in the basement.
Just FYI, 32bit Intel processors from the Pentium Pro generation and forward (with the exception of most, if not all of the Pentium-M's) have 36 physical address pins or more?
Many, but not all, chipsets have a facility for breaking the physical address presentation of the system RAM into a configurably-sized contiguous block below the 4GB limit and then making the rest available above the 4GB limit. If you're curious, the register (in intel parlance) is often called TOLUD (Top of Low Useable DRAM).
Yes, furthermore, given modern OS designs on x86 architecture, a process cannot utilize more than 2gb (windows without/3gb boot option) or 3gb (linux, most BSDs, windows with/3gb and apps specially built to use the 3/1 instead of 2/2 split.)
However, that limitation does not preclude you from having a machine running eight processes using 2GB of physical memory each.
The processor feature is called PAE (Physical Address Extension). It works, basically, by adding an extra level of processor pagetable indirection.
Incidentally, I have a quad P3-700 (It's a Dell PowerEdge 6450) propping a door open that could support 8GB of RAM if you had enough registered, ECC PC-133 SDRAM to populate the sixteen dimm slots.
Anyways, here's a snippet from the beginning of a 32 bit machine running Linux which has 4GB of RAM: [ 0.000000] BIOS-provided physical RAM map: [ 0.000000] BIOS-e820: 0000000000000000 - 0000000000097c00 (usable) [ 0.000000] BIOS-e820: 0000000000097c00 - 00000000000a0000 (reserved) [ 0.000000] BIOS-e820: 00000000000e8000 - 0000000000100000 (reserved) [ 0.000000] BIOS-e820: 0000000000100000 - 00000000defafe00 (usable) [ 0.000000] BIOS-e820: 00000000defb1e00 - 00000000defb1ea0 (ACPI NVS) [ 0.000000] BIOS-e820: 00000000defb1ea0 - 00000000e0000000 (reserved) [ 0.000000] BIOS-e820: 00000000f4000000 - 00000000f8000000 (reserved) [ 0.000000] BIOS-e820: 00000000fec00000 - 00000000fed40000 (reserved) [ 0.000000] BIOS-e820: 00000000fed45000 - 0000000100000000 (reserved) [ 0.000000] BIOS-e820: 0000000100000000 - 000000011c000000 (usable)
The title of that list should really be "Physical Address Space map." Either way, notice that the majority of the RAM is available up until 0xDEFAFE00 and the rest is available from 0x100000000 to 0x11c000000 - a range that's clearly above the 4GB limit.
Yes, it's running a bigmem kernel... But that's what bigmem kernels are for.
Oh, incidentally, even windows 2000 supported PAE. The bigger problem is the chipset. Not all of them support remapping a portion of RAM above 4GB.
The difference is not how the data is run through the cipher. The difference is that there are two completely different algorithms. Think of RC4 as a truck and RC5 as an airplane. Yes, they're both vehicles, but they operate quite differently. Incidentally, an RC2 also exists.
Read the wikipedia articles on stream and block ciphers if you'd like more details.
Furthermore, WEP had a 24bit initialization vector intended to make brute-forcing keys harder. It was, unfortunately, flawed. Furthermore, as you must well know, in addition to the 64 bit (40bit key, 24 bit IV) mode of operation, there was a 128 bit (104bit key, 24 bit IV) mode.
Lastly, the point of the challenges was to get a real measurement of how hard those algorithms were to break. Essentially, they were offered as an experiment. Surely you can understand the value of the results.
Uhh, wrong wrong wrong. WEP, the nonsense encryption in 802.11 cards you described, uses RC4, not RC5. RC4 is a stream cipher, RC5 is a block cipher. Further, the distributed.net RC5-* efforts are trying to win a challenge set forth by RSA Labs, described here: http://www.rsa.com/rsalabs/node.asp?id=2100
The banner on the above page indicates that the challenges are no longer active.
> Yes, who would do business with such an entity. Probably about as many as would trust their business hosting to a company who declares its home page to be XHTML 1.1 but then serves it as text/html. Not to mention the 88 validation errors.
They were taking nude pictures to document any possible signs of a struggle. Try subduing a woman whom is convinced you're going to kill or rape her - you're likely going to end up with scratches, bruises, or marks.
> And yes, I sell valves. Relief valves, control valves, block valves, cryogenic valves, high temperature valves, steam valves. All kinds of valves. All kinds of materials.:) C02 is no big deal to hold underground. It can be done easily.
I think.. I think you're my new hero.
What about electrically acuated valves? For hobby/geek projects?
Slashdot Mirror
Unfortunately, while the NVS series bare the Quadro branding, NVidia does not support the professional/scientific feature sets on those chips. So, features like the unified back buffer, etc. are not available. Essentially, the NVS450 is a card with two GeForce 8400 chips and a PCI-E to PCI-E bridge. It's kinda lame.
NVidia marketing material suggests that the NVS line is intended for business users who need to support many displays without any advanced rendering.
While you're right, I imagine the NVS450 costs more than a pair of GTX220 or GT650 cards; he'd be better served with your suggestion than the NVS card. Personally, I suspect his desktop is a Dell or HP professional workstation as they generally ship with NVS graphics as the entry-level video solution. I doubt he specifically chose the card.
Also, the best solution to his dilemma, IMHO, is the Matrox DualHead2Go or TripleHead2Go. I know it seems like having more GPUs would be a better solution, but I think less GPUs means less overhead in synchronization, mutexes, locks, etc. That's just a hypothesis... no data to back it up.
That sounds a lot like Michael Crichton's 1972 novel, "The Terminal Man." http://en.wikipedia.org/wiki/The_Terminal_Man
*snip*Here are the concerns I have with it:
All power rails appear to be exposed. While they are on the back, this could be a significant safety (personnel and/or fire) issue. Considering that you can up to 500A @ 12.5V DC running through the zone power rails, and potentially more for the main cabinet DC power rails, exposed seems like a bad idea.
*snip*
That appears to be an illustrative picture. An image from a different article of an "in production" or "active testing" rack shows grounded shields around the bus bars. This is the wired.com article I'm referring to. The picture is somewhere in the bottom third.
Your assertion that you'd save "way more" by switching to SSD storage is assuming that the spindle disks are the main consumer of current.
According to WD, the WD20EARX draws 5.3W during read/write, 3.3W during idle, and 0.7W standby/seep(which, admittedly is a rare situation in datacenters.) (from the WD20EARX datasheet)
According to Intel, the Intel 910 series SSD draws up to 25W while active and 8W while idle. The Intel 520 series SSD draws 850mW active and 600mW idle. (from the Intel 520 series product specifications.) I don't know if those numbers for the 910 are a typo, because it seems weird that they'd exceed a mechanical drive.
Either way, my point is that the WD's have a power ratio of 2.65W/TB and the Intel 520 SSDs have a power ratio of 1.78W/TB. Which means that switching to SSDs will save you 33% on your storage power needs. Thing is, because the SSDs have less capacity per SATA port, once you factor in the extra necessary RAID controllers, SATA cards or SATA port expanders, the percent power saving will drop. Admittedly, I have no idea by how much.
I guess, my point is to challenge the popularly regarded idea that mechanical harddrives are extremely power hungry. While CPU efficiency has improved considerably in recent years, I hold that CPUs and associated electronics consume a much larger portion of a server's power than commonly believed.
Also, at idle, the WD consumes 1.65W/TB and the intel consumers 1.45W/TB. Then again, it's not a fair comparison because the SSD can switch between idle and active far more quickly than the mechanical drive. So, once you consider more of aspects of the situation, things become less clearly cut.
The facebook hinged storage server must be using their new 21" rack because they (from images) appear to have an arranged the drives in three rows of 5 drives. The 3.5" drive formfactor is 4" wide, meaning that the enclosure must be at least 20" wide to accomodate five drives per row. Also, using their new rack concept, their servers don't include and AC power supply. So, it's not exactly as space efficient when you factor in the 2U power supply at the bottom. With one PS and one 30drive facebook server, you're at 30drives for 4U or an efficiency of 7.5drives/U. One PS and two 30drive servers, you're at 60drives for 6U and an efficiency of 10drives/U. One PS and three 30drive servers, 90 drives on 8U and an efficiency of 11.25 drives/U. At four servers on one PS unit, you've got 120 drives occupying 10U for an efficiency of 12 drives/U. So, once you have four servers together with the associated PS, you finally reach the efficiency of a thumper.
The thumper (Sun x4500 and x4540) had 48 3.5" HDD's, 2 (x4500) or 3 (x4540) 800W/1600W (110VAC or 220VAC) power supplies and an adorable, itty-bitty dual opteron server. 48 drives occupying 4U is an efficiency of 12drives/U.
To be fair, while the 4server, 1power supply configuration only equals the storage density of the thumper, it has better server/cpu/nic density.
As an aside, the full rack setups appear to have three power supply units. Assuming/guessing 42U per rack with 6U devoted to PS, it leaves 36U divided into three bays of 12U. So, with a PS and five facebook hingy servers occupying 10U and sporting 150 drives at 15 drives/U, you finally outdo the thumper.
To be clear, I do believe that there are benefits to the proposed new rack size, but I don't think it's a clear improvement. Personally, I think the thumper design was brilliant. The only purpose of this reply was to point out that it's not as simple as 15 drives per U.
On a separate note, the ability to fit 5 drives side-by-side in a 21" rack is the best justification I've seen, so far, for widening to 21".
Wow, I prefix too many of my comments with insecure clauses devoid of information and only serving to indirectly apologize to the reader for supplying information I think is important for them to understand despite my worry that I'm trying their patience. If you read all of this reply including even this sudden instrospective insight to my character; then, thank you. I'm flattered.
Nope, they increase a "U" to be 48mm from 44.45mm. This is now called an OU. So, 1OU = 48mm or an increase of 8% compared to a regular U. They claim that this 3.55mm increase will "increases airflow, improving air economization; it also allows for better for cable and thermal management and efficient use of space." Personally, I question wether the increase in airflow, cable management, and efficient use of space will be significant. I'd be very keen to see a good example of how these new 48mm rack units will improve cable management.
Also, the bus bars depicted in the photos appear to be incredibly vulnerable to accidental short circuiting.
Have you read the email shown in the image from the first link(threatpost.com)? It's dated 2003 and it's describing how to optimize the thread local storage local descriptors introduced to linux around that time. If the source code is related to that, then it's likely irrelevant at this point. A lot has happened in the past 9 years.
Uhh, Hi. Yeah. Sorry, The 12cm/13cm band is 2.3-2.31Ghz and 2.39-2.45Ghz. It doesn't overlap the entire ISM allocation. People are still welcome to use channel 11 on their wifi routers... or use those fancy 5GHz radios.
Wouldn't placing such a device on your car constitute a tresspass to land?
Why are you replacing the motherboards yourself? With my T61p, when something in it died - motherboard needed replacing - I called up IBM and told em it's broke. Something like 19 hours later, DHL has a box for me at my door to ship the laptop out in. I put the laptop in the box and call up DHL to schedule picking up the shipping box. The same DHL guy is back 15 minutes later and takes the box. 23 hours later, same DHL delivery guy is back on my doorstep with my repaired laptop. This was with the standard warranty option when buying the laptop. My mind was blown on just how quickly it got fixed. Apparently got shipped from MI to Memphis, repaired, and shipped back in less than 24 hours.
Does this level of support not exist anymore? Otherwise, why are you replacing the motherboards in house - especially if you don't have spare parts readily available. Also, your complaint about the 44 screws? I mean, come'on, tieing the laces on my shoes takes 10 times longer than using velcro, but it's really kinda not a very big deal.
At least the nuclear solution isn't anywhere near as contaminating and destructive to the environment as coal or oil. Think of it as a lesser evil.
The difference between chernobyl's RBMK design and and our operating relics is already rather significant. Also, we have organizations in the US, such as the United States Navy, which are at the forefront of safe reactor design and operation.
VMware has supported 3D acceeleration pass through for years. Works fantastic.
Noone's surprised. Marvell's had a track record of faulty, ill-documented, or bug-plagued parts. Sometimes, I wonder why they still bother. I suppose, someone has to make Realtek look good.
I suck, I meant 4 Gbps.
Only 1, and he did it because he could keep the loud drive cabinet in the basement and use the drives from his bedroom. It was decent gear, too, LSI PCI-E 4GBps card with two multimode transceivers to a JBOD cabinet in the basement.
naw, you mean LSD. Came out of Berkeley around the same time as BSD. Note, not everyone is convinced this is a coincidence.
Just FYI, 32bit Intel processors from the Pentium Pro generation and forward (with the exception of most, if not all of the Pentium-M's) have 36 physical address pins or more?
Many, but not all, chipsets have a facility for breaking the physical address presentation of the system RAM into a configurably-sized contiguous block below the 4GB limit and then making the rest available above the 4GB limit. If you're curious, the register (in intel parlance) is often called TOLUD (Top of Low Useable DRAM).
Yes, furthermore, given modern OS designs on x86 architecture, a process cannot utilize more than 2gb (windows without /3gb boot option) or 3gb (linux, most BSDs, windows with /3gb and apps specially built to use the 3/1 instead of 2/2 split.)
However, that limitation does not preclude you from having a machine running eight processes using 2GB of physical memory each.
The processor feature is called PAE (Physical Address Extension). It works, basically, by adding an extra level of processor pagetable indirection.
Incidentally, I have a quad P3-700 (It's a Dell PowerEdge 6450) propping a door open that could support 8GB of RAM if you had enough registered, ECC PC-133 SDRAM to populate the sixteen dimm slots.
Anyways, here's a snippet from the beginning of a 32 bit machine running Linux which has 4GB of RAM:
[ 0.000000] BIOS-provided physical RAM map:
[ 0.000000] BIOS-e820: 0000000000000000 - 0000000000097c00 (usable)
[ 0.000000] BIOS-e820: 0000000000097c00 - 00000000000a0000 (reserved)
[ 0.000000] BIOS-e820: 00000000000e8000 - 0000000000100000 (reserved)
[ 0.000000] BIOS-e820: 0000000000100000 - 00000000defafe00 (usable)
[ 0.000000] BIOS-e820: 00000000defb1e00 - 00000000defb1ea0 (ACPI NVS)
[ 0.000000] BIOS-e820: 00000000defb1ea0 - 00000000e0000000 (reserved)
[ 0.000000] BIOS-e820: 00000000f4000000 - 00000000f8000000 (reserved)
[ 0.000000] BIOS-e820: 00000000fec00000 - 00000000fed40000 (reserved)
[ 0.000000] BIOS-e820: 00000000fed45000 - 0000000100000000 (reserved)
[ 0.000000] BIOS-e820: 0000000100000000 - 000000011c000000 (usable)
The title of that list should really be "Physical Address Space map." Either way, notice that the majority of the RAM is available up until 0xDEFAFE00 and the rest is available from 0x100000000 to 0x11c000000 - a range that's clearly above the 4GB limit.
Yes, it's running a bigmem kernel... But that's what bigmem kernels are for.
Oh, incidentally, even windows 2000 supported PAE. The bigger problem is the chipset. Not all of them support remapping a portion of RAM above 4GB.
The difference is not how the data is run through the cipher. The difference is that there are two completely different algorithms. Think of RC4 as a truck and RC5 as an airplane. Yes, they're both vehicles, but they operate quite differently. Incidentally, an RC2 also exists.
Read the wikipedia articles on stream and block ciphers if you'd like more details.
Furthermore, WEP had a 24bit initialization vector intended to make brute-forcing keys harder. It was, unfortunately, flawed. Furthermore, as you must well know, in addition to the 64 bit (40bit key, 24 bit IV) mode of operation, there was a 128 bit (104bit key, 24 bit IV) mode.
Lastly, the point of the challenges was to get a real measurement of how hard those algorithms were to break. Essentially, they were offered as an experiment. Surely you can understand the value of the results.
Uhh, wrong wrong wrong. WEP, the nonsense encryption in 802.11 cards you described, uses RC4, not RC5. RC4 is a stream cipher, RC5 is a block cipher. Further, the distributed.net RC5-* efforts are trying to win a challenge set forth by RSA Labs, described here: http://www.rsa.com/rsalabs/node.asp?id=2100
The banner on the above page indicates that the challenges are no longer active.
oh my god... you weren't joking.
> Yes, who would do business with such an entity. Probably about as many as would trust their business hosting to a company who declares its home page to be XHTML 1.1 but then serves it as text/html. Not to mention the 88 validation errors.
teknopurge, you just got served.
They were taking nude pictures to document any possible signs of a struggle. Try subduing a woman whom is convinced you're going to kill or rape her - you're likely going to end up with scratches, bruises, or marks.
> And yes, I sell valves. Relief valves, control valves, block valves, cryogenic valves, high temperature valves, steam valves. All kinds of valves. All kinds of materials. :) C02 is no big deal to hold underground. It can be done easily.
I think.. I think you're my new hero.
What about electrically acuated valves? For hobby/geek projects?
There's only 44,000 myspace profiles, only 44,000 with private pictures, or he only could grab from 44,000 profiles before he got shut off or bored?