Re: Intel D945GSE vs Zotac ION ITX (was: Support for Zotac MB with nVidia ION chipset)
Robert Bonomi bon...@mail.r-bonomi.com wrote: One fairly well-known super computer class architecture from the mid 1960s ran without *any* error checking in the CPU *or* main memory. Dr. Seymour Cray analyzed things and concluded the significant extra component count for just doing 'parity' checking, let alone ECC made for a net _reduction_ in overall system reliability, *IF* the machine was run under very tightly controlled operating conditions -- the big ones being extremely stable power and a very limited temperature range. So, he specified the design to tight tolerances, and ran truely 'naked' hardward. Scary, but true. And, it worked. CDC-6600 and/or 7600, I presume? The flaw in that reasoning is that, while an unchecked machine may indeed be faster and/or have a somewhat better MTBF, the symptom of a failure may well be silently incorrect results. If reliable production results are what's valued, as opposed to time between detected failures while running diagnostics*, a checked or corrected design wins hands down. This was also a machine where, at any given moment, a fair part of the data in the CPU was 'in the wires' (in transit from one part of the CPU to another), and significant parts of the wiring harness had to be of _just_the_right_length_ (speed-of-light considerations) for the box to work. Second- (or third?) hand war story from the manufacturing dept: Occasionally the instructions would call for pin so-and-so to be connected to pin thus-and-such with, say, a 6 wire -- when the pins in question were 8 apart! The source of the story claimed that the standard practice in such cases was to use the shortest wire that would reach, and let the QA dept worry about the fallout. * A diagnostic is a program that runs when the hardware is malfunctioning -- R. F. Rosin. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: Intel D945GSE vs Zotac ION ITX (was: Support for Zotac MB with nVidia ION chipset)
Hi Dan, Peter, Thank you for your enlightening replies. On Mon, Apr 05, 2010 at 10:40:35AM +1000, Peter Jeremy wrote: On 2010-Apr-04 23:54:55 +0200, Jeremie Le Hen jere...@le-hen.org wrote: Yeah, you are right. I should have mentionned that I do not want necessarily a high-performance NAS, it's for home use so my premary concern is the low power consumption. This is why I want an Atom-based motherboard. Unfortunately, FreeBSD has some issues with low memory handling that make FreeBSD/i386 a bad choice for ZFS. You would be far better off running amd64 with as much RAM as you can fit onto the board. (And this is one case where you want amd64 even if you don't have 4GB RAM). Note that this doesn't mean you can't use an Atom - some Atoms include EM64T - you just need to check. IMO, the biggest disadvantage of using an Atom in a ZFS NAS is the lack of ECC support on the Atom. ZFS can detect bitflips in the I/O sustem but you can still get screwed by a bitflip in RAM. I'm still not sure about which motherboard to buy actually. After some additional reading, my leaning seems to go towards Intel's one as it is less expensive and consumes half the power of the Zotac's one (13W with a HDD [2] vs. 25W [3]). I'd recommend against buying anything with the Atom combined with a 945. Whilst the Atom is low-power, the 945 isn't. That is also an older motherboard using an older, superseded Atom. I suggest you look for motherboards built around the new Pinetrail Atoms (which _do_ support EM46T and hence can run amd64). Supermicro make a number of potentially suitable boards: http://www.supermicro.com/products/motherboard/ATOM/ICH9/X7SPA.cfm this is pricier but supports remote management - other options at: http://www.supermicro.com/products/motherboard/Atom/ If you want an Intel MB, search for BOXD510MO I think I will go for the Supermicro X7SPA-H. It seems a better hardware although it consumes more power. Nonetheless I'm a little worried by what you said about the lack of ECC. Computers has been used for years before ECC came out and obviously they worked :). Do you really think it might happen to be a problem? Would an Intel board would compensate for this? Dan, have you ever experienced weird problems that could be explained by bitflips? For the records, I've found an interesting and very recent post about someone running OpenSolaris on this Supermicro motherboard [1]. He uses a thumbdrive for the operating system and with four drives connected onto it, the whole system sucks 41 watts when idle (27 without any HDD, which is twice as the Intel D945GSE, but I guess this is the price for better performance). Now, it seems that OpenSolaris unfortunately has some problems with this hardware, but according to Dan, FreeBSD runs correctly on it. Thanks again for your help. Regards, [1] http://sorenragsdale.livejournal.com/19875.html -- Jeremie Le Hen Humans are born free and equal. But some are more equal than others. Coluche ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: Intel D945GSE vs Zotac ION ITX (was: Support for Zotac MB with nVidia ION chipset)
On Mon, Apr 5, 2010 at 1:20 PM, Jeremie Le Hen jere...@le-hen.org wrote: Nonetheless I'm a little worried by what you said about the lack of ECC. Computers has been used for years before ECC came out and obviously they worked :). Do you really think it might happen to be a problem? Would an Intel board would compensate for this? Dan, have you ever experienced weird problems that could be explained by bitflips? Personally, I haven't had any issues, but then again on the ZFS scale of things, both my current pool size (2 TB) and projected pool size when I add more disks (6 TB) is pretty small. If this was a heavily used machine with a 10 TB pool or bigger, I would definately give strong consideration to ECC. For the records, I've found an interesting and very recent post about someone running OpenSolaris on this Supermicro motherboard [1]. He uses a thumbdrive for the operating system and with four drives connected onto it, the whole system sucks 41 watts when idle (27 without any HDD, which is twice as the Intel D945GSE The power draw (from the wall) for the Supermicro X7SPA-H without any disks attached is as following: 26W - During boot. 24W - IDLE at console 28W - Full load This is with a 80+ rated Corsair 400CX PSU. Sadly, I did not have the opportunity to measure the power draw with powerd enabled. The D945GSE is unsuitable for use as a ZFS NAS due to it's severe feature limitations when compared against the X7SPA-H, of biggest concern would be the limitation of RAM, followed by the amount of native SATA ports, followed by the fact that you only get a PCI-E x1 (both physical formfactor and speed-wise) slot for expansion, while most controller cards are either 4x or 8x, meaning they simply wouldn't physically fit into the slot. Singlecore 1,6Ghz Diamondville Atom VS Dualcore 1,66Ghz Pineview Atom 1 RAM socket supporting a max of 1GB VS 2 RAM sockets supporting a max of 4GB (note that X7SPA-H uses SO-DIMMs, not regular DIMMs) 2 SATA ports vs 6 SATA ports 1 Realtec NIC vs 2 x Intel NIC PCI-E x1 Slot VS PCI-E x4 Slot (in x16 form factor) for expansion - Sincerely, Dan Naumov ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: Intel D945GSE vs Zotac ION ITX (was: Support for Zotac MB with nVidia ION chipset)
On 2010-Apr-05 12:20:12 +0200, Jeremie Le Hen jere...@le-hen.org wrote: Nonetheless I'm a little worried by what you said about the lack of ECC. Computers has been used for years before ECC came out and obviously they worked :). Not really. Most early computers had fairly extensive error detecting hardware. Early microprocessors didn't because the novelty of getting an entire on a CPU on a chip was enough. Most 486 based PCs supported parity RAM but maufacturers and end-users found they could save pennies by leaving the parity bits off. ECC support was a requirement for building servers with microprocessors and some support has trickled down to the desktop. It hasn't been really popular because wider memory costs more and most people want the fastest, cheapest system possible to make their games render faster. Occasional glitches don't matter. With the current generation of CPUs, Intel appear to have made a marketing decision to not support ECC on their desktop CPUs - if you want ECC, you need to user a server-grade CPU (with a much greater profit margin). AMD have gone the other way and have have ECC support in all their x64 chips except mobile ones. You are still at the mercy of motherboard manufacturers who decide to not include the tracks between the DIMM sockets and the CPU. Do you really think it might happen to be a problem? There's no way to know. Definitely, the added error checking in ZFS have resulted in a number of ZFS kept reporting errors and I found I actually had bad hardware even though I've been using it for years reports. Would an Intel board would compensate for this? No. The memory controller is embedded in the Atom and doesn't support ECC. If you decide to go the ECC path, you need to pick a different CPU. -- Peter Jeremy pgpVvDasCr3Le.pgp Description: PGP signature
Re: Intel D945GSE vs Zotac ION ITX (was: Support for Zotac MB with nVidia ION chipset)
From owner-freebsd-questi...@freebsd.org Mon Apr 5 16:34:40 2010 Date: Tue, 6 Apr 2010 07:34:05 +1000 From: Peter Jeremy peterjer...@acm.org To: Jeremie Le Hen jere...@le-hen.org Cc: freebsd-questions@freebsd.org, freebsd-hardw...@freebsd.org Subject: Re: Intel D945GSE vs Zotac ION ITX (was: Support for Zotac MB with nVidia ION chipset) --n8g4imXOkfNTN/H1 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On 2010-Apr-05 12:20:12 +0200, Jeremie Le Hen jere...@le-hen.org wrote: Nonetheless I'm a little worried by what you said about the lack of ECC. Computers has been used for years before ECC came out and obviously they worked :). Not really. Most early computers had fairly extensive error detecting hardware. Depends on what machines you're talking about. One fairly well-known supe-r computer class architecture from the mid 1960s ran without *any* error checking in the CPU *or* main memory. Dr. Seymour Cray analyzed things and concluded the significant extra component count for just doing 'parity' checking, let alone ECC made for a net _reduction_ in overall system reliability, *IF* the machine was run under very tightly controlled operating conditions -- the big ones being extremely stable power and a very limited temperature range. So, he specified the design to tight tolerances, and ran truely 'naked' hardward. Scary, but true. And, it worked. This was also a machine where, at any given moment, a fair part of the data in the CPU was 'in the wires' (in transit from one part of the CPU to another), and significant parts of the wiring harness had to be of _just_the_right_length_ (speed-of-light considerations) for the box to work. Incidentally, this computer COULD NOT ADD two numbers together. Literally!! It performed addition by 'complement and subtract'. Yeah, it -sounds- silly, but there were valid architectural reasons for it. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: Intel D945GSE vs Zotac ION ITX (was: Support for Zotac MB with nVidia ION chipset)
Hi, --- Cc: me when replying, as I'm not subscribed. --- I cross-post this reply to freebsd-hardware@ since the result of my little study around Atom-based motherboard may be of interest for readers of this ML too. On Sun, Apr 04, 2010 at 03:00:09PM -0600, Tim Judd wrote: On 4/4/10, Jeremie Le Hen jere...@le-hen.org wrote: Hi, --- Please Cc: me when replying, I'm not subscribed. --- I plan to purchase a Zotac motherboard with a embedded ATOM processor. It uses an NVidia chipset. http://www.zotacusa.com/zotac-ionitx-f-e-atom-n330-1-6ghz-dual-core-mini-itx-intel-motherboard.html My intent is to build a small NAS with ZFS and NFS/CIFS. I'd like to know if anyone successfully ran FreeBSD on this motherboard and what performance could be achieved, especially if ZFS is used. I checked the archives without luck. A NAS w/ ZFS, NFS and CIFS/SMB, doesn't need any feature of the ion chipset. Why are you electing for this board if you're not running any graphical environment? And ZFS is memory hungry, the Atom is a i386-like chip, so you'd have too much overhead with ZFS. I think you've elected the wrong board for your purposes. Will FreeBSD run on it? yes. I have freebsd on another atom N-series ASUS box. Yeah, you are right. I should have mentionned that I do not want necessarily a high-performance NAS, it's for home use so my premary concern is the low power consumption. This is why I want an Atom-based motherboard. By the way, I found an post on OpenSolaris forums where the author achieves something like 35MB/s on a ZFS filesystem through CIFS using an Intel Atom-based motherboard [1]. This is enough for the use I intend to have. Zotac mobo is better than Intel D954GSE because it provides a wireless interface - although I couldn't figure out which chipset yet, so I don't know if it's corretly supported on FreeBSD - and three S-ATA connectors. I'm still not sure about which motherboard to buy actually. After some additional reading, my leaning seems to go towards Intel's one as it is less expensive and consumes half the power of the Zotac's one (13W with a HDD [2] vs. 25W [3]). I can live with two S-ATA connectors and I can plug a wireless interface on the available PCI connector if I ever need it. Regards, [1] http://mail.opensolaris.org/pipermail/opensolaris-discuss/2009-June/048214.html [2] http://www.homeserverhacks.com/2009/06/hands-on-whs-build-with-intel-d945gsejt.html [3] http://www.anandtech.com/show/2765/12 -- Jeremie Le Hen Humans are born free and equal. But some are more equal than others. Coluche ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
RE: Intel D945GSE vs Zotac ION ITX (was: Support for Zotac MB with nVidia ION chipset)
Just a small comment regarding Atom suitability for a home NAS: feel free to completely ignore people saying that ZFS overhead is too much for an Atom to handle efficiently, they have no idea what they are talking about. I am using a Supermicro X7SPA-H board (Atom D510) and I an easily achieving ~85mb/s transfers over Samba to and from the machine. 85mb/s is also the best these drives will do and my CPU is nowhere near maxed during these transfers, so with better disks I would be easily saturating gigabit, while still having plenty of available CPU time. What you want is a good disk controller and fast and reliable disks, 2gb RAM is enough, but with 4gb ram you can basically safely enable prefetch for a very noticable boost in sequential pattern reads. Below are some numbers from my personal Atom NAS system: === bonnie -s 8192 ---Sequential Output ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- MachineMB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 8192 29065 68.9 52027 39.8 39636 33.3 54057 95.4 105335 34.6 174.1 7.9 dd if=/dev/zero of=test1 bs=1M count=8192 8589934592 bytes transferred in 111.300481 secs (77177875 bytes/sec) (73,6mb/s) dd if=/dev/urandom of=test2 bs=1M count=8192 dd if=test2 of=/dev/zero bs=1M 8589934592 bytes transferred in 76.031399 secs (112978779 bytes/sec) (107,74mb/s) === This is a ZFS mirror of 2 x 2tb WD Green drives with 32mb cache with the automatic headparking disabled via WDIDLE3. The drives are very cheap and hence, are the bottleneck in my case. - Sincerely, Dan Naumov ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re: Intel D945GSE vs Zotac ION ITX (was: Support for Zotac MB with nVidia ION chipset)
On 2010-Apr-04 23:54:55 +0200, Jeremie Le Hen jere...@le-hen.org wrote: Yeah, you are right. I should have mentionned that I do not want necessarily a high-performance NAS, it's for home use so my premary concern is the low power consumption. This is why I want an Atom-based motherboard. Unfortunately, FreeBSD has some issues with low memory handling that make FreeBSD/i386 a bad choice for ZFS. You would be far better off running amd64 with as much RAM as you can fit onto the board. (And this is one case where you want amd64 even if you don't have 4GB RAM). Note that this doesn't mean you can't use an Atom - some Atoms include EM64T - you just need to check. IMO, the biggest disadvantage of using an Atom in a ZFS NAS is the lack of ECC support on the Atom. ZFS can detect bitflips in the I/O sustem but you can still get screwed by a bitflip in RAM. I'm still not sure about which motherboard to buy actually. After some additional reading, my leaning seems to go towards Intel's one as it is less expensive and consumes half the power of the Zotac's one (13W with a HDD [2] vs. 25W [3]). I'd recommend against buying anything with the Atom combined with a 945. Whilst the Atom is low-power, the 945 isn't. That is also an older motherboard using an older, superseded Atom. I suggest you look for motherboards built around the new Pinetrail Atoms (which _do_ support EM46T and hence can run amd64). Supermicro make a number of potentially suitable boards: http://www.supermicro.com/products/motherboard/ATOM/ICH9/X7SPA.cfm this is pricier but supports remote management - other options at: http://www.supermicro.com/products/motherboard/Atom/ If you want an Intel MB, search for BOXD510MO Note that I'm not sure how well FreeBSD's X.org supports the Pinetrail yet. There have been some commits but I don't know if support is complete. -- Peter Jeremy pgpC4FYAbzinT.pgp Description: PGP signature
