Of course they're not quite the same as an octuple core AMD64; they're fairly simple in-order-execution UltraSparc cores depending on chip-multithreading to keep throughput up on highly concurrent workloads, which is somewhat different from the rather complex deeply pipelined out-of-order cores with good single-process throughput we're used to.
Hmm, I use fb2k for most of my audio conversion (and everything else for that matter), and it has no trouble maintaining metadata across whatever formats you throw at it, and it's certainly not the only tool capable of this. What are you using, command line decoders and encoding from the resulting WAV or something?
"The mere belief that a God doesn't exist, is a religion"
Pretty dilute meaning of the word "religion" you're using there. What do you base this on?
Much here depends on your interpretation of the nature of the belief. I consider myself an atheist, but I still don't really have a strong conviction along the lines of "there is no God"; how the hell should I know? It's not a concept I can disprove.
Equally, though, it's not a concept I see any vaguely convincing argument for, either. And let's face it, we're not talking about just one belief here; you say "a God", which is a pretty vague concept, but one that's far removed from most religions. It's a hop of astronomical proportions to go from there to a God who takes a personal interest in my specific beliefs, feelings and actions, and which really wouldn't mind a chat and some appreciation, and maybe a few temples (which you might want to donate something towards), and who also likes to talk more directly to a few authority figures who can tell us exactly what pleases him/her/it, and to never mind the thousands of others following the same patterns because *ours* is really *real*.
I fail to see how one can make such a leap; what little I do see points rather more strongly to evolutionary psychology and the exploitation of our flawed perceptions and critical thinking skills by memes and people more than any deeper truth about the nature of the universe.
Maybe I'm wrong, and this tendancy to believe in *some* kind of Godlike being does reflect reality. Maybe there is some God out there who is displeased by my doubt, and would like to see me stoned to death for engaging in premarital sex or using one of his names as an expletive or eating pork that hasn't been killed *just* so, but I don't see why I should give this significantly more credence than the possibility that the Invisible Pink Unicorn is going to trample me to/after death for suggesting that she likes pepperoni more than ham.
That's what hardware wear leveling is for. The company I was talking to about solid state drives claimed one of theirs would last decades even with constant writes, although we're talking on the order of $1k for a few GB here.
"but then they'd be buying the dual core opteron's since they have more cache than the amd 64 x2"
Nope, the higher end X2 at each clock rate comes with 1M L2 per core, just like any Opteron. You won't be getting X2's on dual CPU boards though:)
"dot to mention you can get a 'pair' of the lowest dual core opteron for less money than a single one of the fastest dual core opterons. price-perfomrance is incredibly askew... two $750 chips that have over 160% the operation capabilities of a $1,600-$2,000 chip?"
Right; yields of lower clocked dual core parts are higher than those of high clock single cores, so if you need the faster single threaded performance, you better be prepared to pay for it. Some people do, and thus are prepared to pay for it:)
"the opteron 280 is really only for those with a load of cash to blow. and they might as well be getting the 880 pay $3,000 per cpu and be able to go 4-way or 8-way, with a suitable backplane"
Going from £1k to £3k per CPU and from £300 to £1000+ per motherboard is a fairly big leap, even for those with a load of cash to blow. Most would be better off buying 2 or 3 dual CPU systems than a single quad.. those that are left generally *do* have loads of cash to blow:)
Seagate NL35 are near-line versions of their consumer drives, with MTBF's of around 1M hours.
Of course with a failure rate on your existing drives ~15x what you might expect, I'd consider getting one through a different vendor/courier, although 24/7 writes isn't exactly the sort of duty cycle you might expect for a cheap consumer drive either.
"As for dual dual-core CPUs, that's also not a stunt, and for the same reason in that it can provide actual performance advantages. Nonstop (fast) DVD-to-DivX encoding while playing F.E.A.R with all the bells and whistles, without buying a separate PC"
More like nonstop database serving to twice as many users, while taking up half the rack space and being 5x cheaper than an equivilent quad CPU system. Not everything's about gaming, even with high end graphics cards (think: medical imaging, CAD, artists, etc)
Quite. It's not that minimalist though.. well, it's minimalist in the same way vim is minimalist; the standard interface config makes the baby Jesus detonate supernovae near budding civilizations, but once he configures it properly he makes sure that the resulting debris cloud looks really pretty.
Also it's really fast, has kickass database features, ninja tagging skills, and contains 80% less evil by volume than iTunes and its proprietary DRM/ReplayGain-alike/lossless format. It just needs a cross-platform counterpart and a less SETI-upsetting default config.
"iTunes is the only desktop music player worth getting excited over."
Why is iTunes worth getting excited about? It supports fewer formats and standards than most other players, it's slower, more annoying to use, and less flexible than many. Really, what's exciting about it?
Also, iTunes makes the baby Jesus stab puppies in the eyes, so this better be good.
Well, no, the Broadcom is the less broken port, and the SP uses it, which would be good if the FreeBSD bge driver supported the pass-through stuff the SP uses. The patch designed to enable it seems non-functional, but it may just be a 64bit issue. Linux/Solaris et all support it fine, so there's plenty of reference material for a better patch; it's mainly some register tweaking.
We've also experienced some odd CMOS Checksum errors which seem indicitive of a BIOS bug or so. Also apparantly the rails are a bit on the flimsy side, so try to avoid resting unracked servers on top of an X2100;)
Yup, although we got the wrong cable with the original shipment as well as the first replacement, so we've not had a chance to try it until today.
Port 2 is the Broadcom, which sadly doesn't work well with FreeBSD since the bge driver doesn't let the SP see the network, but there's a patch I'm going to try shortly which should do the trick. It doesn't seem possible to talk to the SP via serial console, but with ipmitool/FreeIPMI it's quite usable.
The FreeBSD nve driver does kinda support it; it detects the port and you can actually make it behave vaguely like a network interface up to the point at which it dies horribly in a variety of interesting ways, seemingly due to problems with the binary blob nVidia provide; work is ongoing to try to fix this as well as another project to port the reverse-engineered forcedeth Linux driver, so it's not without hope.
It's a Tyan nForce 4, based on what looks like a rackmount-optimized version of this. Nothing mindblowing, but they're well built and are a decent price; we got one to try out as a FreeBSD 6 appserver, and ordered 3 more as a result.
Only caveats are some harmless ACPI notices during bootup, the predictably mostly-useless nForce 4 ethernet (the other's a perfectly fine Broadcom), and the not-quite-working nForce SATA hotswap; standard Tyan nForce 4 + FreeBSD fare.
Opera 9/Win32 here, and it actually starts off fine; it just gets progressively slower the more you move, and the visibility display on the left isn't cleared properly. Memory consumption keeps going up each frame too, so I guess something's leaking.
And in what way does better transfer rates not make it easier to exchange files? If it takes 30s to fill my thumbdrive instead of 3 minutes, well, yes, I would kind of care, especially when I'm about to shoot out the door and remember I was going to bring a GB or two of photos for a friend to see. I don't want to be thinking "damn, I don't have time", and hell yeah I'd be willing to pay a bit more to not have to do so.
My HD's for the past few years have pushed 50M/s STR. I'm sorry you can't see why I might want my portable storage to at least vaguely keep up.
Yes it does, provided you're running the latest version (1.70 iirc), although you may need to play with some of the compatibility options (in the Properties dialog for the executable). Certainly works for me. Maybe it doesn't like you; I, for one, can certainly see where it might be coming from;)
No.. that's just a user saying they use our service; I think we'd try including a link if we were gonna stoop to the level of advertising in a comment:P
3TB in 2 months though, yikes; that's about 5Mbps constant. I guess he has varied tastes, lots of free time, a big supply of DVD-R's, and perhaps a touch of OCD;)
Not necessarily. 6bpp TFT's quote 16.2 million colours because they employ temporal dithering -- switching pixels between shades rapidly so you perceive the value in between. If it's done well you'd probably be hard pressed to tell the difference.
You probably get more performance, if anything, since you're using faster unbuffered memory.. but that makes it tougher to add lots of memory to a system. Every S939 motherboard I've seen tops out at 4G (4*1G), dual sockets generally support up to 16G (8*2G); some even do 16*2G or more if you're willing to drop the memory speed and remortgage your house.
This seems to be the way AMD are balancing single socket dual core with dual socket configurations; drop the registered memory and max out at 4G so you can't just go build a 16G 165 or so.
One thing I haven't worked out yet is what makes an Opteron 175 any way different from an AMD64 X2 4400. The only distinction seems to come in the low power models such as those used by Sun; anyone noticed anything else?
Of course they're not quite the same as an octuple core AMD64; they're fairly simple in-order-execution UltraSparc cores depending on chip-multithreading to keep throughput up on highly concurrent workloads, which is somewhat different from the rather complex deeply pipelined out-of-order cores with good single-process throughput we're used to.
The ULV Pentium-M's quote a TDP of 5W :P
Apparantly nVidia are awaiting Page Attribute Table support before they can release a FreeBSD/amd64 driver.
Hmm, I use fb2k for most of my audio conversion (and everything else for that matter), and it has no trouble maintaining metadata across whatever formats you throw at it, and it's certainly not the only tool capable of this. What are you using, command line decoders and encoding from the resulting WAV or something?
"The mere belief that a God doesn't exist, is a religion"
Pretty dilute meaning of the word "religion" you're using there. What do you base this on?
Much here depends on your interpretation of the nature of the belief. I consider myself an atheist, but I still don't really have a strong conviction along the lines of "there is no God"; how the hell should I know? It's not a concept I can disprove.
Equally, though, it's not a concept I see any vaguely convincing argument for, either. And let's face it, we're not talking about just one belief here; you say "a God", which is a pretty vague concept, but one that's far removed from most religions. It's a hop of astronomical proportions to go from there to a God who takes a personal interest in my specific beliefs, feelings and actions, and which really wouldn't mind a chat and some appreciation, and maybe a few temples (which you might want to donate something towards), and who also likes to talk more directly to a few authority figures who can tell us exactly what pleases him/her/it, and to never mind the thousands of others following the same patterns because *ours* is really *real*.
I fail to see how one can make such a leap; what little I do see points rather more strongly to evolutionary psychology and the exploitation of our flawed perceptions and critical thinking skills by memes and people more than any deeper truth about the nature of the universe.
Maybe I'm wrong, and this tendancy to believe in *some* kind of Godlike being does reflect reality. Maybe there is some God out there who is displeased by my doubt, and would like to see me stoned to death for engaging in premarital sex or using one of his names as an expletive or eating pork that hasn't been killed *just* so, but I don't see why I should give this significantly more credence than the possibility that the Invisible Pink Unicorn is going to trample me to/after death for suggesting that she likes pepperoni more than ham.
Pay a bit extra and get ECC memory, enable Chipkill, set up Machine Check Exception handling/logging; there's your SMART for memory.
Cookie to whoever comes up with a list of ECC supporting S939 motherboards.
That's what hardware wear leveling is for. The company I was talking to about solid state drives claimed one of theirs would last decades even with constant writes, although we're talking on the order of $1k for a few GB here.
"but then they'd be buying the dual core opteron's since they have more cache than the amd 64 x2"
:)
:)
:)
Nope, the higher end X2 at each clock rate comes with 1M L2 per core, just like any Opteron. You won't be getting X2's on dual CPU boards though
"dot to mention you can get a 'pair' of the lowest dual core opteron for less money than a single one of the fastest dual core opterons. price-perfomrance is incredibly askew... two $750 chips that have over 160% the operation capabilities of a $1,600-$2,000 chip?"
Right; yields of lower clocked dual core parts are higher than those of high clock single cores, so if you need the faster single threaded performance, you better be prepared to pay for it. Some people do, and thus are prepared to pay for it
"the opteron 280 is really only for those with a load of cash to blow. and they might as well be getting the 880 pay $3,000 per cpu and be able to go 4-way or 8-way, with a suitable backplane"
Going from £1k to £3k per CPU and from £300 to £1000+ per motherboard is a fairly big leap, even for those with a load of cash to blow. Most would be better off buying 2 or 3 dual CPU systems than a single quad.. those that are left generally *do* have loads of cash to blow
Seagate NL35 are near-line versions of their consumer drives, with MTBF's of around 1M hours.
Of course with a failure rate on your existing drives ~15x what you might expect, I'd consider getting one through a different vendor/courier, although 24/7 writes isn't exactly the sort of duty cycle you might expect for a cheap consumer drive either.
"As for dual dual-core CPUs, that's also not a stunt, and for the same reason in that it can provide actual performance advantages. Nonstop (fast) DVD-to-DivX encoding while playing F.E.A.R with all the bells and whistles, without buying a separate PC"
More like nonstop database serving to twice as many users, while taking up half the rack space and being 5x cheaper than an equivilent quad CPU system. Not everything's about gaming, even with high end graphics cards (think: medical imaging, CAD, artists, etc)
Depends how hard you throw the book. At 3km/s a 1kg book would release energy practically equivilent to 1kg of TNT ;)
*tunes up his mass driver*
Quite. It's not that minimalist though.. well, it's minimalist in the same way vim is minimalist; the standard interface config makes the baby Jesus detonate supernovae near budding civilizations, but once he configures it properly he makes sure that the resulting debris cloud looks really pretty.
Also it's really fast, has kickass database features, ninja tagging skills, and contains 80% less evil by volume than iTunes and its proprietary DRM/ReplayGain-alike/lossless format. It just needs a cross-platform counterpart and a less SETI-upsetting default config.
"iTunes is the only desktop music player worth getting excited over."
Why is iTunes worth getting excited about? It supports fewer formats and standards than most other players, it's slower, more annoying to use, and less flexible than many. Really, what's exciting about it?
Also, iTunes makes the baby Jesus stab puppies in the eyes, so this better be good.
Well, no, the Broadcom is the less broken port, and the SP uses it, which would be good if the FreeBSD bge driver supported the pass-through stuff the SP uses. The patch designed to enable it seems non-functional, but it may just be a 64bit issue. Linux/Solaris et all support it fine, so there's plenty of reference material for a better patch; it's mainly some register tweaking.
;)
We've also experienced some odd CMOS Checksum errors which seem indicitive of a BIOS bug or so. Also apparantly the rails are a bit on the flimsy side, so try to avoid resting unracked servers on top of an X2100
Yup, although we got the wrong cable with the original shipment as well as the first replacement, so we've not had a chance to try it until today.
Port 2 is the Broadcom, which sadly doesn't work well with FreeBSD since the bge driver doesn't let the SP see the network, but there's a patch I'm going to try shortly which should do the trick. It doesn't seem possible to talk to the SP via serial console, but with ipmitool/FreeIPMI it's quite usable.
The FreeBSD nve driver does kinda support it; it detects the port and you can actually make it behave vaguely like a network interface up to the point at which it dies horribly in a variety of interesting ways, seemingly due to problems with the binary blob nVidia provide; work is ongoing to try to fix this as well as another project to port the reverse-engineered forcedeth Linux driver, so it's not without hope.
It's a Tyan nForce 4, based on what looks like a rackmount-optimized version of this. Nothing mindblowing, but they're well built and are a decent price; we got one to try out as a FreeBSD 6 appserver, and ordered 3 more as a result.
Only caveats are some harmless ACPI notices during bootup, the predictably mostly-useless nForce 4 ethernet (the other's a perfectly fine Broadcom), and the not-quite-working nForce SATA hotswap; standard Tyan nForce 4 + FreeBSD fare.
Opera 9/Win32 here, and it actually starts off fine; it just gets progressively slower the more you move, and the visibility display on the left isn't cleared properly. Memory consumption keeps going up each frame too, so I guess something's leaking.
And in what way does better transfer rates not make it easier to exchange files? If it takes 30s to fill my thumbdrive instead of 3 minutes, well, yes, I would kind of care, especially when I'm about to shoot out the door and remember I was going to bring a GB or two of photos for a friend to see. I don't want to be thinking "damn, I don't have time", and hell yeah I'd be willing to pay a bit more to not have to do so.
My HD's for the past few years have pushed 50M/s STR. I'm sorry you can't see why I might want my portable storage to at least vaguely keep up.
"Dungeon Keeper II doesn't run in Windows XP"
;)
Yes it does, provided you're running the latest version (1.70 iirc), although you may need to play with some of the compatibility options (in the Properties dialog for the executable). Certainly works for me. Maybe it doesn't like you; I, for one, can certainly see where it might be coming from
Ironically, my cron mailbox is currently 3.1GB, and I need a 64bit mutt to open it; 32bit mutt thinks it's -1158120246bytes in size ;)
No.. that's just a user saying they use our service; I think we'd try including a link if we were gonna stoop to the level of advertising in a comment :P
;)
3TB in 2 months though, yikes; that's about 5Mbps constant. I guess he has varied tastes, lots of free time, a big supply of DVD-R's, and perhaps a touch of OCD
It's mentioned on their PDF comparison chart; "16.7 million colours supported ... Yes".
Not necessarily. 6bpp TFT's quote 16.2 million colours because they employ temporal dithering -- switching pixels between shades rapidly so you perceive the value in between. If it's done well you'd probably be hard pressed to tell the difference.
You probably get more performance, if anything, since you're using faster unbuffered memory.. but that makes it tougher to add lots of memory to a system. Every S939 motherboard I've seen tops out at 4G (4*1G), dual sockets generally support up to 16G (8*2G); some even do 16*2G or more if you're willing to drop the memory speed and remortgage your house.
This seems to be the way AMD are balancing single socket dual core with dual socket configurations; drop the registered memory and max out at 4G so you can't just go build a 16G 165 or so.
One thing I haven't worked out yet is what makes an Opteron 175 any way different from an AMD64 X2 4400. The only distinction seems to come in the low power models such as those used by Sun; anyone noticed anything else?