From: Peter Arremann loony@loonybin.org
Better believe it - cause that's what they are doing :-)
Actually, I can believe it (I just wish that had *1* good option). Tyan's focus is cost and economies-of-scale. You'll find the same of SuperMicro, among a few others.
When you start looking at HP and Sun's Opteron designs, then the mainboard cost is only a portion of the system cost. So they are more focus on delivering the optimal solution. Because doing so only affects the end price less than 10%, whereas it might affect the mainboard cost 50%+.
Tyan doesn't lay out that many different boards from scratch... just look at most of their stuff (S2892, S2885, S2882) - the CPU/memory area is identical, so all they did was take the old design, mod around with the chipset and PCI/PCI-X/PCI Express slots and they got a new one...
Actually, other than EMF and the support microelectronics, it's simply a matter of deciding how many traces you want, which may result in more PCB layers.
Tyan is really focused on cost. They have an older, "cheap" 2-way Opteron mainboard that doesn't even put DDR channels on CPU #2.
And several manufacturers are even worse than Tyan. Especially the non-server focused vendors. SuperMicro and Tyan are better than most.
But still not the quality in mainboard design you'll get from select Tier-1 OEMs like HP and Sun.
-- Bryan J. Smith mailto:b.j.smith@ieee.org
Tyan is really focused on cost. They have an older, "cheap" 2-way Opteron mainboard that doesn't even put DDR channels on CPU #2.
And several manufacturers are even worse than Tyan. Especially the non-server focused vendors. SuperMicro and Tyan are better than most.
But still not the quality in mainboard design you'll get from select Tier-1 OEMs like HP and Sun.
So what is the bottom line? None of the common board makers such as Asus, Iwill, Tyan, Supermicro and Gigabit make a "good" board? So what should the poor server builder do?
On Thu, Jun 23, 2005 at 05:54:49PM -0400, Ed Clarke wrote:
So what is the bottom line? None of the common board makers such as Asus, Iwill, Tyan, Supermicro and Gigabit make a "good" board? So what should the poor server builder do?
Build 3 or 4 decent servers for the price of one Sun ...
On Thu, 2005-06-23 at 15:13 -0700, Alan Hodgson wrote:
On Thu, Jun 23, 2005 at 05:54:49PM -0400, Ed Clarke wrote:
So what is the bottom line? None of the common board makers such as Asus, Iwill, Tyan, Supermicro and Gigabit make a "good" board? So what should the poor server builder do?
Build 3 or 4 decent servers for the price of one Sun ...
I personally buy Asus and/or Tyan boards all the time. I am still running several Dual Pentium III Tyan and Asus servers that I bought 4-7 years ago. (And even one Quad Pentium Pro board)
They are still DNS, mail, and http servering just fine.
Are there (and were there at the time) better boards out there ... certainly. But, they performed just fine for me for a very long time ... though most now have moved into secondary rolls as e-mail gateways, firewall boxes, low volume web servers or DNS servers.
On Thursday 23 June 2005 17:54, Ed Clarke wrote:
So what is the bottom line? None of the common board makers such as Asus, Iwill, Tyan, Supermicro and Gigabit make a "good" board? So what should the poor server builder do?
Lets be realistic here... all I/O on one CPU only makes a difference for the most IO intensive apps (talking several hundret MB/sec or more)... Memory on only one CPU is a bad thing though if you can avoid it. I ran a Tyan board though several benchmarks - anything from db2 to http... All tests were run with 2 GB - the difference was 4 modules attached to both cpus, or 2 modules on just one... All DDR266 registered. The difference was between 0 and 11% on DB2 ... With the data AMD provides I expected it to be much less than that but maybe IBM is doing something funky there...
Anyway, long story short, try to avoid boards that attach all memory to one cpu - unless they cost so much less that you can add another gig or two in memory... I/O on the other hand, while not ideal, will have much lesser effect on your performance.
Peter.
On Thu, 2005-06-23 at 17:54 -0400, Ed Clarke wrote:
So what is the bottom line? None of the common board makers such as Asus, Iwill, Tyan, Supermicro and Gigabit make a "good" board? So what should the poor server builder do?
I never said they don't make a "good" board. I just said that they don't make "ideal" ones.
I personally like the Tyan S2895 now that more and more PCIe x4 and x8 storage controllers are hitting the market. You don't have to use PCIe channels for video, and it already has a NIC on CPU #2. It's not much more than the S2892 at all.
_Otherwise_, PCIe is _not_ a necessity on a server, especially if _all_ I/O is attached to CPU #0. You still have PCI-X channels.
In fact, I would argue that you're not going to lose much at all by going with an older, $300 dual-Opteron mainboard that has an AMD8131 HyperTransport tunnel. *NOTE* that's "3" (AMD8131 = dual-PCI-X channels) not "5" (AMD8151 = AGP 3.0) nor "1" (AMD8111 = legacy PCI).
So instead of working about the latest'n greatest Opteron mainboard, see if you an find a $300 dual-Opteron mainboard with the AMD8131 that supports dual-core. Heck, dual-core is probably over-rated unless you're really crunching a lot of data, because the memory channels are _not_ doubled.
Bryan J. Smith wrote:
So instead of working about the latest'n greatest Opteron mainboard, see if you an find a $300 dual-Opteron mainboard with the AMD8131 that supports dual-core.
Okay, there's the challenge kiddies! Who can come up with an actual in-stock SKU that I can pay real $$$ for? I love Bryan's knowledge and information. Really, I do! But my choices *end* with the motherboards a retailer can put into my UPS driver's hands.
Bryan, I wish I could force Tyan, Asus and all the rest to make your perfect design. But I can't. I'm sorry. ;)
On Thursday 23 June 2005 22:18, Kirk Bocek wrote:
Bryan J. Smith wrote:
So instead of working about the latest'n greatest Opteron mainboard, see if you an find a $300 dual-Opteron mainboard with the AMD8131 that supports dual-core.
Okay, there's the challenge kiddies! Who can come up with an actual in-stock SKU that I can pay real $$$ for? I love Bryan's knowledge and information. Really, I do! But my choices *end* with the motherboards a retailer can put into my UPS driver's hands.
Bryan, I wish I could force Tyan, Asus and all the rest to make your perfect design. But I can't. I'm sorry. ;)
The board: Asus K8N-DL http://www.asus.com.tw/products.aspx?l1=9&l2=39&l3=174&model=456...
The dual core support: In bios 1004 http://support.asus.com/faq/faq.aspx?SLanguage=en-us&model=K8N-DL
The place to buy: ZipZoomFly.com among others - $274.99 (free shipping) http://www.zipzoomfly.com/jsp/ProductDetail.jsp?ProductCode=240430 or NewEgg $270.00 (plus shipping) http://www.newegg.com/Product/Product.asp?Item=N82E16813131059&CMP=OTC-p... both in stock :-)
Only two issues with the board - memory support is supposedly not so great and Asus themselves claims the board doesn't exist :-) Go to their webpage and go through the links - no dual opteron boards... Go to the search box, enter the name of the board and you'll get all you need though...
Peter.
On Thu, 2005-06-23 at 23:06 -0400, Peter Arremann wrote:
The board: Asus K8N-DL http://www.asus.com.tw/products.aspx?l1=9&l2=39&l3=174&model=456...
Um, that's _not_ a server board at all. That's a workstation board with virtually no I/O. It's designed for pure computational need and _no_ I/O at all. That's not me being "anal" -- I mean, there is _no_ I/O on that thing!
You've got the PCIe x16 for video, and a PCIe x1 for maybe an audio, NIC or low-end storage card. Unless you get a PCIe x4 or x8 storage card and put it in that PCIe x16 slot, you're going to be sharing _all_ I/O on a legacy, 133MBps PCI slot.
And it would be nice to have a server NIC that is connected via PCI-X, and not a "client" designed NIC that is in the nForce4 chipset.
You might as well have an old, 100MHz Pentium processor mainboard. I'm dead serious here, the 100MHz Pentium would have no trouble handling the legacy PCI interconnect that system is giving you. It is a _workstation_ mainboard (and assumes your workstation doesn't need much I/O ;-).
The place to buy: ZipZoomFly.com among others - $274.99 (free shipping) http://www.zipzoomfly.com/jsp/ProductDetail.jsp?ProductCode=240430 or NewEgg $270.00 (plus shipping) http://www.newegg.com/Product/Product.asp?Item=N82E16813131059&CMP=OTC-p... both in stock :-)
For $20-50 more, you can have a mainboard with an AMD8131 and dual PCI-X 1.0 slots. That's 533MBps per 64-bit @ 66MHz card (such as a 3Ware Escalade 8506-4 or 8506-8), or 1066MBps for newer 64-bit @ 133MHz PCI-X cards (like an LSI Logic MegaRAID 300X-8).
Only two issues with the board - memory support is supposedly not so great and Asus themselves claims the board doesn't exist :-) Go to their webpage and go through the links - no dual opteron boards... Go to the search box, enter the name of the board and you'll get all you need though...
This is _not_ a server board at all.
You'd be far better off getting an old dual-Pentium III with a ServerWorks IIILE or IIIHE chipset for $100-200, and some old 800-933MHz P3 processors for $50/each, while reusing some old PC100-133 memory. You'll at least have a 533MBps (64@66) PCI slot or two for storage and NIC.
You're just asking for trouble putting everything on a shared, legacy 133MBps PCI bus in a server. Unless you start using the PCIe slots, although the GbE NIC in the nForce4 doesn't have much SRAM, although the typical Brocomm designs in the PCI-X versions typically do have about 8x the receive SRAM buffer (which is needed for servers).
This is _not_ me being "anal." I install _low-cost_ networks all-the- time. I do it on a dime. And this board for $270 is _not_ what you want in a server. Trust me on this! Splurge $20-50 more and get something with at least an AMD8131.
Wow - you surely do type fast :-)
Anyway, I was only half serious with that answer - anything less than a Tyan S2882 would be a waste when using dual core chips, especially if you figure out the cost difference in the board in comparison to the cost of the CPUs. But give me some credit - after all it was you who threw out the $300 price mark and besides that, the original poster said he had no real IO requirements.
On Friday 24 June 2005 00:48, Bryan J. Smith wrote:
On Thu, 2005-06-23 at 23:06 -0400, Peter Arremann wrote:
The board: Asus K8N-DL http://www.asus.com.tw/products.aspx?l1=9&l2=39&l3=174&model=456... u=1
Um, that's _not_ a server board at all. That's a workstation board with virtually no I/O. It's designed for pure computational need and _no_ I/O at all. That's not me being "anal" -- I mean, there is _no_ I/O on that thing!
Yep :-)
You've got the PCIe x16 for video, and a PCIe x1 for maybe an audio, NIC or low-end storage card. Unless you get a PCIe x4 or x8 storage card and put it in that PCIe x16 slot, you're going to be sharing _all_ I/O on a legacy, 133MBps PCI slot.
And that would be a smart thing to do if you got stuck with a board like that... In a server you waste the PCIe slot for graphics - put an ancient 8MB ati rage in there and you're set.
And it would be nice to have a server NIC that is connected via PCI-X, and not a "client" designed NIC that is in the nForce4 chipset.
You might as well have an old, 100MHz Pentium processor mainboard. I'm dead serious here, the 100MHz Pentium would have no trouble handling the legacy PCI interconnect that system is giving you. It is a _workstation_ mainboard (and assumes your workstation doesn't need much I/O ;-).
You keep repeating that word workstation a lot... What exactly is the difference between a workstation and a server? The high end is easy... Server - domainable, cell board based with redundant power grid and and and... but the lower you go, the more similar they become... First you loose the redundant power grids, then the cell board based approach... Next goes the blacklisting of components and finally somewhere you lost your service processor.. What's left is a Sun 420R server... oops - the Ultra-80 workstation I mean... Oh darn - except the case, they are the same... Well, maybe sun doesn't know any better.. I bet HP does... unless the J6xxx workstations and the Ax00 servers are the same too... Darn... but the J6750s make excellent rackmountable workstations :-) IBM? low end P5 p5[25]0 - they don't even bother calling it two different models - you can just get it configured as server or workstation... (btw, not defending that asus board with this - that board really is very limited in IO)
This is _not_ me being "anal." I install _low-cost_ networks all-the- time. I do it on a dime. And this board for $270 is _not_ what you want in a server. Trust me on this! Splurge $20-50 more and get something with at least an AMD8131.
Completely agreed - spend the extra money if you want to build a serious server...
Peter.
On Thu, 2005-06-23 at 19:18 -0700, Kirk Bocek wrote:
Okay, there's the challenge kiddies! Who can come up with an actual in-stock SKU that I can pay real $$$ for? I love Bryan's knowledge and information. Really, I do! But my choices *end* with the motherboards a retailer can put into my UPS driver's hands. Bryan, I wish I could force Tyan, Asus and all the rest to make your perfect design. But I can't. I'm sorry. ;)
First off, I _know_, that's why I _dropped_ it! Give me some credit. ;->
Secondly, I _then_ said find an _older_, $300 dual-Opteron mainboard with the AMD8131 dual PCI-X HyperTransport tunnel. I've yet to see a mainboard with an AMD8131 that didn't also have dual-DDR channels to _each_ CPU, so that's the best, single recommendation I can make.
For example, call up Monarch Computer (who sells both components or complete, tested systems from those components -- your choice) and see what they have. It looks like Monarch is selling some older boards re- branded with their own BIOS that supports dual-core for $300-400.
http://www.monarchcomputer.com/Merchant2/merchant.mv? Screen=CTGY&Store_Code=M&Category_Code=OMB
Tell them you want: A) Mainboard with dual-DDR (128-bit) to _each_ CPU B) An AMD8131 for _dual_ PCI-X
Many of the boards at the above link for $300-400 are _exactly_that_! And several are dual-core supporting. Again, it looks like Monarch is taking older/discontinued mainboards of the older AMD8xxx generation and putting a new, dual-core capable BIOS on them.
From a server standpoint, you're going to use PCI-X channels, and
whether you have an older AMD8xxx generation or a new, nForce Pro generation, you're _still_ going to get PCI-X via an AMD8131 -- same difference.
-- Bryan
P.S. If you think you're going to get a dual-Opteron mainboard for under $300, it's near impossible. The only ones that are under $300 typically lack an AMD8131 -- i.e., _no_ PCI-X channels. You'd be better off getting a $100 Athlon64 nForce mainboard with an Athlon64x2 4200+ and a PCIe x4 or PCIe x8 storage controller instead.
On Friday 24 June 2005 00:37, Bryan J. Smith wrote:
P.S. If you think you're going to get a dual-Opteron mainboard for under $300, it's near impossible. The only ones that are under $300 typically lack an AMD8131 -- i.e., _no_ PCI-X channels. You'd be better off getting a $100 Athlon64 nForce mainboard with an Athlon64x2 4200+ and a PCIe x4 or PCIe x8 storage controller instead.
Here we go again with that UPS guy though... other than the Intel 2ch scsi controller and emulex fibre cards there isn't much out there. And all those options fit into a nice $40K server - but if you're talking a couple hundret for the board and cpu, they're overkill... Until you got a sata raid controller or so available for pci express, you're better off if a more expensive board and then save the money on some cheaper storage..
Peter.
On Fri, 2005-06-24 at 01:49 -0400, Peter Arremann wrote:
Here we go again with that UPS guy though... other than the Intel 2ch scsi controller and emulex fibre cards there isn't much out there.
_Exactly_! PCIe storage and NIC controllers are lacking. Which means we're left with PCI-X and legacy 64-bit PCI cards for now.
And all those options fit into a nice $40K server - but if you're talking a couple hundret for the board and cpu, they're overkill... Until you got a sata raid controller or so available for pci express, you're better off if a more expensive board and then save the money on some cheaper storage..
Exactly. You're already spending $1,500 on the CPUs and Registered ECC memory. So why not spend a few bucks more on the mainboard, and throw in a $300 storage card for the $400 your spending on disks to ensure performance?
Whey people are willing to spend >>$1,000 and then skimp on a few hundred bucks more is beyond me. You'd be better off not going Opteron/Registered at all, and saving all that dough.
On Fri, 2005-06-24 at 01:13 -0500, Bryan J. Smith wrote:
Whey people are willing to spend >>$1,000 and then skimp on a few hundred bucks more is beyond me. You'd be better off not going Opteron/Registered at all, and saving all that dough.
The other thing is the NIC. People will spend thousands of dollars, then put in a desktop NIC.
In the days of 100Mbps, it wasn't too bad. But in the days of GbE, it's unbelievable.
I remember back when NetGear first introduced the 533MBps PCI GA620 with only a 512KB SRAM cache for under $500. Man, I had never seen a GbE with less than 1MB (and typically 2MB+). Now that's back when it was glued (and not a single MAC+SRAM IC), but the wait state wasn't any worse.
But now we're in the age of MAC+SRAM in 1 IC.
Mainboards are coming with RealTek GbE MACs that have 2KB of SRAM -- yes _two_kilobytes_! You know what that means? It can store 1 Ethernet frame (and forget Jumbo frames) and that's all! ;-ppp Most of the Intel PCIe and AMD HyperTransport MACs are only 8-32KB as well. Okay for a client -- especially with 802.3x -- but detrimental for GbE on a server in _any_ capacity.
Even Intel's NICs are MACs with typically 16-64KB, with only 1 model has 256KB SRAM. The Broadcoms on Opterons are typically 64-96KB SRAM/port, although they do have a MAC for a NIC with 256KB in the single MAC+SRAM IC.
First thing I do when I walk into a client that says they're having performance issues with GbE is find out what their server NIC is. 9 times out of 10, it's a cheap, desktop NIC or embedded MAC -- let alone a pre-802.3x one.
802.3x doesn't solve the problem for servers (or will the newer sub- committees either), but at least it does let the system tell every transmitting node to shut-up until its ready again. Otherwise you get a compounding effect that makes good'ole broadcast storms look tame. At least the wire is the problem there, not the server itself.
P.S. If you think you're going to get a dual-Opteron mainboard for under $300, it's near impossible. The only ones that are under $300 typically lack an AMD8131 -- i.e., _no_ PCI-X channels. You'd be better off getting a $100 Athlon64 nForce mainboard with an Athlon64x2 4200+ and a PCIe x4 or PCIe x8 storage controller instead.
Do you have any information on PCIe storage controllers?
I have googled and gone here and there but not found anything.
Do you know of any card that supports more than 4 SATA ports without a dumb fakeraid tacked on? Just a pure SATA card with more than 4 ports.
On Friday 24 June 2005 03:24, Feizhou wrote:
Do you have any information on PCIe storage controllers?
Fibre Emulex: http://www.emulex.com/products/fc/index.html QLogic: http://www.qlogic.com/products/fc_san_hostadapers.asp
Scsi Intel: http://www.intel.com/design/servers/RAID/srcu42e/ LSI: http://www.lsilogic.com/products/megaraid/megaraid_320_2e.html
SATA Tekram/Areca: http://www.tekram.com/raidcrd.asp
In all - not really worth it right now in my opinion... might be faster than PCI-X but all the options above are fairly new - therefore untested and probably more difficult to set up than you'd expect.
I have googled and gone here and there but not found anything.
Do you know of any card that supports more than 4 SATA ports without a dumb fakeraid tacked on? Just a pure SATA card with more than 4 ports.
Unfortunately that will be virtually impossible to find... If you want more than 4 ports, a real raid card will probably be your only option. Some fakeraid controllers allow you to completely disable the firmware though - I'm sure you've read through it - but just in case you didn't... http://linuxmafia.com/faq/Hardware/sata.html
Peter.
Peter Arremann wrote:
On Friday 24 June 2005 03:24, Feizhou wrote:
Do you have any information on PCIe storage controllers?
[list of controllers]
In all - not really worth it right now in my opinion... might be faster than PCI-X but all the options above are fairly new - therefore untested and probably more difficult to set up than you'd expect.
Thanks for the info.
I have googled and gone here and there but not found anything.
Do you know of any card that supports more than 4 SATA ports without a dumb fakeraid tacked on? Just a pure SATA card with more than 4 ports.
Unfortunately that will be virtually impossible to find... If you want more than 4 ports, a real raid card will probably be your only option.
:(
i hope somebody makes a card with that broadcom chip that supports 8 SATA ports without add any dumb fakeraid.
http://www.broadcom.com/products/Enterprise-Small-Office/Storage-Solutions/B...
Any 4 port SATA only controllers around? That was hard for me to find too.
Some fakeraid controllers allow you to completely disable the firmware though
- I'm sure you've read through it - but just in case you didn't...
Thanks.
Any 4 port SATA only controllers around? That was hard for me to find too.
The link has led me to some, thanks.
Some fakeraid controllers allow you to completely disable the firmware though - I'm sure you've read through it - but just in case you didn't... http://linuxmafia.com/faq/Hardware/sata.html
Thanks.
-
Alan Hodgson -
Bryan J. Smith -
Bryan J. Smith <b.j.smith＠ieee.org> -
Ed Clarke -
Feizhou -
Johnny Hughes -
Kirk Bocek -
Peter Arremann