8-core motherboard for hackintosh?

deleted_account · March 15, 2008

Some CEB Form Factor followers may find some interst in this new 8-Core Motherboard:

ASUS Z7S WS

Product Specifications:

Manufacturer's Website:

http://www.asus.com/products.aspx?l1=3&amp...amp;modelmenu=1

HotHardware Preview:

http://www.hothardware.com/News/Asus_Z7S_W...oard_Sneak_Peek

Team Scream · March 15, 2008

I dont own a real piece of modern hardware yet, let alone an 8 core.... been delaying for one reason or another and seems like finances keep getting in the way. Plus the fact that I'm going to have to buy EVERYTHING at once if I go 8 core.

Threw my PCI FX5200 into a friends hack running kalyway 10.5.2 last night.... 3 monitors easy. swapped third monitor from port 3 to port 4..... hit detect displays, pops right up. So at looks like i'll only really need to buy one new video card.....

I hope to be throwing something pretty spiffy together here before too long... but i've been hoping and planning quite a while, so no guarantee on timeframe... but when I get something rockin, I'll be sure to post bench's.

And yeah, we need to figure out how to boot SAS or U320 SCSI off of a LSI FusionMPT supported controller....... hehe.

Oh, another thing to consider when buying ram..... dual rank FB-DIMM's give higher total throughput....

AND filling each memory channel with 2+ FB-DIMM's also improves throughput.... although latency goes up a hair (but then scales better)

Anyone bitching about the "horrible latency problems" with FB-DIMM's hasn't really looked at the issue... UNLESS they're talking latency while at nearly idle. YES, at idle, the FB-DIMM's have a higher latency. Start pulling some memory bandwidth.... even 25% or so, and DDR2 has worse latency than FB-DIMM at that point... and FB-DIMM just makes bigger and bigger gains from there on out. (and has twice the total bandwidth in a balls to the walls config)

I'm still debating on if I should build a dirt cheap 2/4 core desktop board and get everything else I'll need... then just swap in the 771 board, xeons and FB-DIMM's and do it in two stages.... of if I should just put my face to the grindstone and build what I really want anyways....

(And where are the memory riser 16 slot 5400 chipset boards with 1600/800 support? also easier to cool....) BAH!

I hear ya Vlad,

Very nice information in those last 2 posts of yours my friend.

I noticed a significant increase in real world performance when I filled every available memory slot on my board, it just seems like heavy renders take quite a bit less time than they did when I only had 4 gigs in this beast.

I love this machine, and I literally have NOT turned it off or rebooted in weeks, perhaps almost 2 months since I got it sorted.

I am stable as hell with every pro app I use, it sleeps and wakes up and really what more could I ask for.

vladthebad · March 15, 2008

That asus board doesn't look too bad for CEB/SSI but I'd rather have AT LEAST 8 FB-DIMM sockets to take advantage of dual rank modules, and 2 modules per channel at some point... besides as I see it, FB-DIMM pricing is likely to continue to drop on 1gb/2gb/4gb modules.

also..... Was doing a bit of in depth reading of intel documents (and IBM Redbooks for xSeries stuff... but I digress) Look at the little wonder I found in the ESB southbridge manual: (Applicable to 5000x/5400 series chipsets)

EFI Text and Keyboard at Boot Time

EFI is designed to support text and keyboard redirection over COM port. No specific

additions are needed in EFI to be supported.

Why yes..... EFI BABY.... hmmmmmm. Wonder if that could possibly solve some problems.

chuckeroo · March 15, 2008

I hope this isn't too ridiculous of question:

I was able to acquire 2 E8400s (socket 775) for a really good price. Are there any mobos out there

that can handle both of these processors that will work well with osx86? The chipset of some of the

more successful boards (p35/ich9r) or the 5000 chipsets from earlier in this thread would be great.

This is my first build, and I realize that most of these dual cpu boards are server boards so they use

the xeon 771 format.

Is there anything like this out there?

vladthebad · March 16, 2008

There are no dual 775 boards..... I do not believe there will EVER be any dual socket 775 boards, because as far as I know, there is no control logic in the front side bus to label the sockets and direct data correctly.

The pentium 3, socket 370... let you have dual processors on the same board with normal chips... BUT BOTH SOCKETS SHARED THE SAME FSB.

Now..... suppose that someone DOES find a way to make a dual 775 board..... look at how long the FSB link would have to be... they'd never get it to clock. The older Xeons that were P4 based and shared FSB between multiple sockets had this exact problem. The desktop P4's had far faster FSB's than the Xeons, because of the physical realities of mounting two sockets. One possible solution to this, would be to put BOTH sockets.... onto a riser card... with 1 socket on each side of the riser card. This would shorten the FSB links considerably.. but since you now have to put northbridge there on the riser as well... now the riser is getting unreasonably big...... now add memory to the riser... (or pay a price in speed most likely) Even then, you'll still have to pass PCIe connections to your graphics slots and your link to southbridge off of the riser..... Another thing you could do would be put the power management on the board below the riser, and use something like a pair of dedicated 20 pin molex plugs to pass the low voltage power to the riser card.... You're riser card is going to likely have to be an 8 layer board or so... total cost of such a rediculous buildout I'm guessing would be in the $500-800 for the motherboard range.... and you gain what?...up to 4 more cores fighting for DESKTOP BOARD memory bandwidth? 4 more cores fighting over the same front side bus? and don't even THINK about trying to overclock such a board. Consider it factory overclocked to hell and back just getting it to work at normal speed.

Now.... let me talk about the other reason why a dual 775 board makes no sense:

You can buy Socket 771 chips for about the same price as socket 775 chips....

EACH socket has its own front side bus.

You get 4 channels of memory bandwidth instead of 2.

You get up to 16 FB-DIMM slots at DDR2-800 speed

or 32 FB-DIMM slots at up to DDR2-667 speed (haven't seen any news of 800mhz FB-DIMM support in 32 slot solutions.

4 gigs ram per slot = 64 gig or 128 gig..

If you're willing to blow HUGE HUGE money on ram.... you can buy 8 gig sticks for about $5k each. (Must install in pairs.... so 10k a pair)

(If you want to use normal desktop DDR2, you can use the 5100 chipset.... 6 slots(8 slots unofficially as seen on a few boards), 4 gigs per slot = 24 gig memory support, but only dual channel) (The 5100 boards normally are using ICH-9 southbridge) (With the 5100 chipset, you get 3 x8 slots from northbridge and 6 x1 links from southbridge so less overall bandwidth as well)

In regards to the whole FB-DIMM thing..... the latency issue is VERY MUCH overblown by gamers and overclockers for several reasons. A: they're not comparing same speed memory, and in most situations they're comparing single threaded performance. FURTHERMORE... the part they normally don't tell you, is that memory latency increases RADICALLY as you make more and more memory requests and really start pulling major memory bandwidth AND... once you get above a reasonable access rate... the FB-DIMM solution offers FAR MORE bandwidth, with a lower latency... (ONCE you get over the hump.... if you aren't really thrashing memory in multiple threads, you may not get over the hump. The vast majority of synthetic benchmarks don't get over the latency hump... and most games will never get there either. FB-DIMM's aren't the lowest latency.... but start really multi-tasking until your dual channel DDR2 machine really starts lagging... then compare. (Or just start doing some high end video editing/rendering or audio tasks, or major database work...)

(Another note: The Phenom's with the TLB patch enabled in bios actually suffer worse memory latency than some of the FB-DIMM configs across the board.... and no one seems to totally {censored}, moan and whine about that..... Frankly, I think a lot of the FB-DIMM bashing has been done by people that were put off by the early pricing of FB-DIMM, and by people that don't need or use much ram. AND by tons of review sites that are 90+ focused on GAMING. If you only need 4 gig, FB-DIMM offers little in the way of advantages other than radical bandwidth... but people who need radical bandwidth normally need FAR MORE than 4gb of ram.)

In the end though, all the memory bandwidth in the world can't save you from going to hard drive.... although an insane amount of ram can certainly help... as can a wicked fast drive setup.

PLUS you get the added benefit of workstation class support.

ECC memory (1,2,3 and 4 bit error correction on 5400 chipset if configured correctly)

I believe the 5400 series also has something similar to IBM's chipkill that will stop using a chunk of ram that repeatedly is showing errors.

Memory Mirroring (RAID1 for RAM... if you have ram failure, you don't lose any data, and the machine keeps crunching... a few IBM workstations also feature how swap memory. YES HOT SWAP MEMORY! REPLACE OR UPGRADE RAM WHILE MACHINE IS RUNNING!!!!!!)

PCIe

PCI-X

HOT PLUG SUPPORT is an option for PCIe/PCI-X

Lastly: The practical side.

The E8400's you have each have 1 penryn die under the heat spreader... with 2 cores at 3ghz and 6mb L2 cache, at a rated FSB of 1333.. with a 65 watt TDP.

Put both together.... under 1 heat spreader... and you have a QX9650... with 2 penryn dies under the heat spreader, for a total of 4 cores at 3ghz, and 6mb L2 cache per pair of cores (12mb total L2) at a rated FSB of 1333 with a 130 watt TDP.

The E8400's are roughly 250$ each

The QX9650 is roughly 1000$ (Why build an insanely custom, fully unsupported dual 775 board when the quad core does everything your pair of E8400's does for $500 more.)

OR... go Xeon on socket 771 with a workstation/server board:

Xeon E5450, 2 penryn dies under the same heat spreader, for a total of 4 cores at 3ghz and 6mb L2 cache per pair of cores (12mb total L2) at a rated FSB of of 1333 with a 80 watt tdp

Xeon E5450 $1000 (Same chip as QX9650, but on socket 771, hair cheaper, with 70 watts less heat.)

If the heat isn't an issue, you can save another $100 and go with an X5450, 120 watt TDP

Or for the slightly more budget minded,

$550 or so buys you a PAIR of E5410... 4 cores each, 8 cores total, 2.33ghz, 12mb L2, 1333 FSB, 80 watt TDP.

(pinmod to 1600 FSB, and you get 2800mhz if you're insane)

or under $500 buys a pair of E5405's, same as E5410's but at 2 ghz... get 2.4ghz with pinmod to 1600 FSB (for those Xeon overclockers out there)

Really there is no super huge extra cost in going to socket 771... Figure an extra $250-300 for the motherboard, beyond that, its the small costs that add up.

E-ATX case (in most cases), EPS12v power supply with correct plugs, two aftermarket heatsink/fan setups intead of one (if you don't want the loudest computer in the dorm building)

Bit more expencive ram.... and 4 sticks for quad channel.... etc

Now, 5100 chipset isn't such a bad deal if all you want is cores. You still get a seperate FSB for each processor, you get to use cheap DDR2 ram, only 2 sticks needed.... and ICH9 southbridge.

http://www.newegg.com/Product/Product.aspx...N82E16813151163

Thats not a bad example of a nice 5100 board.

Personally, I prefer the 5400 chipset.... more memory bandwidth by far, more workstation/server class features, and the fact that it happens to be what apple uses in the mac pro.... so should be more compatible overall as versions change and things get updated.

Like many of my posts, this one has grown rather long due to my need to convey information I feel is important. If someone wants to summarize or quote me and maybe sticky something about why there are no dual socket 775 boards, that'd be great....

Until then, back to the dual socket goodness...

chuckeroo · March 16, 2008

Thanks for the extremely thorough reply

I got the two E8400's for a total of $200, so I was just hoping to make good use of both of them. Since it's not

possible, I'll have to settle for overclocking this sucker to 4 Ghz I suppose, and do something else with the other.

Thanks!

Team Scream · March 16, 2008

Thanks for the extremely thorough reply

I got the two E8400's for a total of $200, so I was just hoping to make good use of both of them. Since it's not

possible, I'll have to settle for overclocking this sucker to 4 Ghz I suppose, and do something else with the other.

Thanks!

I say build 2 identical machines and set up a mini render farm for your video workload, and let those E8400's spread their wings a bit and crunch mad data for you!

RobertNesta · March 18, 2008

Hi I'm new to the board, and new to trying to build an 8 core quad hackintosh, never mind on the cheap.

So what do you guys think of this motherboard, in conjunction with a pair of 5410's? (i've read the whole thread, but i can't remember if someone talked about it...i read it late last night)

http://www.tyan.com/product_board_detail.aspx?pid=560

BTW< the knowledge that some of you have is amazing, and I love that you spread it around. Information collectives are where it's at, so just let me start by saying thank you very much for your contributions to my increased knowledge.

Cheers

EDIT: found a thread talking specifically about this board. its here if anyone's interested and a little slow off the mark like me

http://forum.insanelymac.com/index.php?showtopic=85238

vladthebad · March 19, 2008

Thats the board I happen to like.... it SHOULD run fine. Same northbridge chipset as mac pro... same southbridge chipset as mac pro..... etc.

My advice is to NOT try and cheap out in a few places......

Don't cheap out on power supply.... dual socket Xeon boards with FB-DIMM drink power... and they can be pretty finicky with power supplies. (And make sure the wiring is long enough... big board, big case, may need longer wiring... especially for hard drives)

Another thing to consider.... is looking at the pricing of 800mhz FB-DIMM modules instead of 667.

That gives you the option to later swap up to 1600 mhz FSB Xeon's and not have to replace your ram.

deleted_account · March 19, 2008

What are the official hardware parts of the Mac Pro? I can never find the specifics other than the general (Intel 5400 chipset, 800 MHz FB-DIMM, up to 32gb, harpertown) If you find out any info, please share.

Are any of you guys familiar with the Intel® ESB2E ICH storage controller? I believe its part of the 5400 chipset, but not the same as the S5397.

When I consider a motherboard purchase, caveats I pretty much lookout for before purchasing motherboard parts would be to make sure the Chipset (Northbridge + Southbridge) and Controllers (SATA, Firewire, USB, LAN/Network, AUDIO) are compatible (Processors and memory will usually work OOB). What else do you guys usually look for prior to purchasing?

And 800 MHz would be a better investment vs. 667 MHz?

karterilla · March 19, 2008

gramarye

here is an lspci listing (one without the -n option the other with -n which gives Vendor ID and device ID) of an original Mac Pro Rev. 1.1. Not too differento from the 2008 boxes.

I've used an Ubuntu 7.10 CD booted with "alt" pressed at boot: the mac identify the CD as Windoze...

I do also have the very same output (but not right away) of the new 2.8 Mac.

If I'm right 2.8 and 3.0 share same mobo but the 3,2 chipset seems to be different.

Anyway here is what you find inside an "old" Mac Pro:

00:00.0 Host bridge: Intel Corporation 5000X Chipset Memory Controller Hub (rev 30)

00:02.0 PCI bridge: Intel Corporation 5000 Series Chipset PCI Express x8 Port 2-3 (rev 30)

00:03.0 Non-VGA unclassified device: Intel Corporation 5000 Series Chipset PCI Express x4 Port 3 (rev 30)

00:04.0 PCI bridge: Intel Corporation 5000X Chipset PCI Express x16 Port 4-7 (rev 30)

00:05.0 Non-VGA unclassified device: Intel Corporation 5000 Series Chipset PCI Express x4 Port 5 (rev 30)

00:06.0 Non-VGA unclassified device: Intel Corporation 5000 Series Chipset PCI Express x4 Port 6 (rev 30)

00:07.0 Non-VGA unclassified device: Intel Corporation 5000 Series Chipset PCI Express x4 Port 7 (rev 30)

00:08.0 System peripheral: Intel Corporation 5000 Series Chipset DMA Engine (rev 30)

00:10.0 Host bridge: Intel Corporation 5000 Series Chipset FSB Registers (rev 30)

00:10.1 Host bridge: Intel Corporation 5000 Series Chipset FSB Registers (rev 30)

00:10.2 Host bridge: Intel Corporation 5000 Series Chipset FSB Registers (rev 30)

00:11.0 Host bridge: Intel Corporation 5000 Series Chipset Reserved Registers (rev 30)

00:13.0 Host bridge: Intel Corporation 5000 Series Chipset Reserved Registers (rev 30)

00:15.0 Host bridge: Intel Corporation 5000 Series Chipset FBD Registers (rev 30)

00:16.0 Host bridge: Intel Corporation 5000 Series Chipset FBD Registers (rev 30)

00:1b.0 Audio device: Intel Corporation 631xESB/632xESB High Definition Audio Controller (rev 09)

00:1c.0 PCI bridge: Intel Corporation 631xESB/632xESB/3100 Chipset PCI Express Root Port 1 (rev 09)

00:1c.1 PCI bridge: Intel Corporation 631xESB/632xESB/3100 Chipset PCI Express Root Port 2 (rev 09)

00:1c.2 PCI bridge: Intel Corporation 631xESB/632xESB/3100 Chipset PCI Express Root Port 3 (rev 09)

00:1c.3 PCI bridge: Intel Corporation 631xESB/632xESB/3100 Chipset PCI Express Root Port 4 (rev 09)

00:1d.0 USB Controller: Intel Corporation 631xESB/632xESB/3100 Chipset UHCI USB Controller #1 (rev 09)

00:1d.1 USB Controller: Intel Corporation 631xESB/632xESB/3100 Chipset UHCI USB Controller #2 (rev 09)

00:1d.2 USB Controller: Intel Corporation 631xESB/632xESB/3100 Chipset UHCI USB Controller #3 (rev 09)

00:1d.3 USB Controller: Intel Corporation 631xESB/632xESB/3100 Chipset UHCI USB Controller #4 (rev 09)

00:1d.7 USB Controller: Intel Corporation 631xESB/632xESB/3100 Chipset EHCI USB2 Controller (rev 09)

00:1e.0 PCI bridge: Intel Corporation 82801 PCI Bridge (rev d9)

00:1f.0 ISA bridge: Intel Corporation 631xESB/632xESB/3100 Chipset LPC Interface Controller (rev 09)

00:1f.1 IDE interface: Intel Corporation 631xESB/632xESB IDE Controller (rev 09)

00:1f.2 IDE interface: Intel Corporation 631xESB/632xESB/3100 Chipset SATA IDE Controller (rev 09)

00:1f.3 SMBus: Intel Corporation 631xESB/632xESB/3100 Chipset SMBus Controller (rev 09)

01:00.0 PCI bridge: Intel Corporation 6311ESB/6321ESB PCI Express Upstream Port (rev 01)

01:00.1 PIC: Intel Corporation 6311ESB/6321ESB I/OxAPIC Interrupt Controller (rev 01)

01:00.3 PCI bridge: Intel Corporation 6311ESB/6321ESB PCI Express to PCI-X Bridge (rev 01)

02:00.0 PCI bridge: Intel Corporation 6311ESB/6321ESB PCI Express Downstream Port E1 (rev 01)

02:01.0 PCI bridge: Intel Corporation 6311ESB/6321ESB PCI Express Downstream Port E2 (rev 01)

02:02.0 PCI bridge: Intel Corporation 6311ESB/6321ESB PCI Express Downstream Port E3 (rev 01)

05:00.0 Ethernet controller: Intel Corporation 80003ES2LAN Gigabit Ethernet Controller (Copper) (rev 01)

05:00.1 Ethernet controller: Intel Corporation 80003ES2LAN Gigabit Ethernet Controller (Copper) (rev 01)

08:00.0 VGA compatible controller: ATI Technologies Inc R580 [Radeon X1900 XT] (Primary)

10:0b.0 FireWire (IEEE 1394): Texas Instruments TSB82AA2 IEEE-1394b Link Layer Controller (rev 01)

lspci -n

00:00.0 0600: 8086:25c0 (rev 30)

00:02.0 0604: 8086:25f7 (rev 30)

00:03.0 0000: 8086:25e3 (rev 30)

00:04.0 0604: 8086:25fa (rev 30)

00:05.0 0000: 8086:25e5 (rev 30)

00:06.0 0000: 8086:25e6 (rev 30)

00:07.0 0000: 8086:25e7 (rev 30)

00:08.0 0880: 8086:1a38 (rev 30)

00:10.0 0600: 8086:25f0 (rev 30)

00:10.1 0600: 8086:25f0 (rev 30)

00:10.2 0600: 8086:25f0 (rev 30)

00:11.0 0600: 8086:25f1 (rev 30)

00:13.0 0600: 8086:25f3 (rev 30)

00:15.0 0600: 8086:25f5 (rev 30)

00:16.0 0600: 8086:25f6 (rev 30)

00:1b.0 0403: 8086:269a (rev 09)

00:1c.0 0604: 8086:2690 (rev 09)

00:1c.1 0604: 8086:2692 (rev 09)

00:1c.2 0604: 8086:2694 (rev 09)

00:1c.3 0604: 8086:2696 (rev 09)

00:1d.0 0c03: 8086:2688 (rev 09)

00:1d.1 0c03: 8086:2689 (rev 09)

00:1d.2 0c03: 8086:268a (rev 09)

00:1d.3 0c03: 8086:268b (rev 09)

00:1d.7 0c03: 8086:268c (rev 09)

00:1e.0 0604: 8086:244e (rev d9)

00:1f.0 0601: 8086:2670 (rev 09)

00:1f.1 0101: 8086:269e (rev 09)

00:1f.2 0101: 8086:2680 (rev 09)

00:1f.3 0c05: 8086:269b (rev 09)

01:00.0 0604: 8086:3500 (rev 01)

01:00.1 0800: 8086:3504 (rev 01)

01:00.3 0604: 8086:350c (rev 01)

02:00.0 0604: 8086:3510 (rev 01)

02:01.0 0604: 8086:3514 (rev 01)

02:02.0 0604: 8086:3518 (rev 01)

05:00.0 0200: 8086:1096 (rev 01)

05:00.1 0200: 8086:1096 (rev 01)

08:00.0 0300: 1002:7249

10:0b.0 0c00: 104c:8025 (rev 01)

Hope it helps

deleted_account · March 19, 2008

Thanks for sharing karterilla,

Hmm...I'm a bit more interested in the Intel 5400 Seaburg chipset however. But this is very interesting stuff.

vladthebad · March 20, 2008

only 1 southbridge really works with the 5400 chipset.... pretty much same one used on the old chipset, major changes were all handled in northbridge.

Go look on intels site... the 631/632 ESB (or ESB2, same thing) is pretty much the only current beefy server southbridge..... the 5100 chipset uses the desktop ICH-9 southbridges most of the time...

Frankly, the ESB is a MONSTER of a southbridge.... For an idea how monster it is:

Intel® 632xESB I/O Controller Hub 1284 Flip Chip - Ball Grid Array (PBGA)

1284 pins/balls/etc

vs

ICH9: 676 mBGA

Side by Side:

ICH9:

HDA (High Def Audio)

6 SATA ports

12 USB 2.0 ports

6 PCIe x1

10/100/1000 NIC

ALL connected to northbridge by a 2 GB/sec link.

6321ESB:

HDA

6 SATA ports

IDE channel

8 USB 2.0 ports

4 PCIe x1 (or 1 PCIe x4)

2 PCIe x4 (or 1 PCIe x8)

2 10/100/1000 NIC's

PCI-X133 64 bit bus

PCI bus

over an x4 ESI link and x8 PCIe to northbridge (ESI = 1GB/sec each direction, x8 PCIe = 2GB/sec each direction so total of 6GB/sec total bandwidth (3GB/sec each direction) (All boards I've seen detailed specs on only do x4ESI + x4 so 4GB/sec) Don't know if thats an error in documentation or not.)

Give me ESB2 or 6321ESB or whatever intel is calling it these days ANYDAY....

With Nehalem moving the memory controller into the CPU, as well as some PCIe bandwidth for graphics that was previously the domain of northbridge... I wonder what is going to happen to northbridge... will it be removed from most systems to save cost... and dual socket systems might use a monster northbridge with more PCIe (say PCIe 2.0 x32 to each processor using their "graphics" links... and then PCIe 2.0 x64 to 3 x16 slots and x16 to southbridge. (or 3 x16 + x16(x8 from northbridge+x8 from southbridge) and a monster snoop filter/L4 cache? (like IBM's custom 4/8 socket chipset). (PLX currently makes a 48 lane PCIe 2.0 switch... I'm sure intel could easily out do them... most likely in a smaller transistor budget than the current 5400B)

deleted_account · March 20, 2008

Thanks again for sharing this vladthebad, I like seeing the comparison really, it helps me understand it a lot better as well. Now time to dd! :\

karterilla · March 20, 2008

graymarye

this is the 2.8GHz early 2008 chipset:

00:00.0 Host bridge: Intel Corporation Memory Controller Hub (rev 20)

00:01.0 PCI bridge: Intel Corporation PCI Express Port 1 (rev 20)

00:05.0 PCI bridge: Intel Corporation PCI Express Port 5 (rev 20)

00:09.0 PCI bridge: Intel Corporation PCI Express Port 9 (rev 20)

00:0f.0 System peripheral: Intel Corporation DMA/DCA Engine (rev 20)

00:10.0 Host bridge: Intel Corporation FSB Registers (rev 20)

00:10.1 Host bridge: Intel Corporation FSB Registers (rev 20)

00:10.2 Host bridge: Intel Corporation FSB Registers (rev 20)

00:10.3 Host bridge: Intel Corporation FSB Registers (rev 20)

00:10.4 Host bridge: Intel Corporation FSB Registers (rev 20)

00:11.0 Host bridge: Intel Corporation Unknown device 4031 (rev 20)

00:15.0 Host bridge: Intel Corporation FBD Registers (rev 20)

00:15.1 Host bridge: Intel Corporation FBD Registers (rev 20)

00:16.0 Host bridge: Intel Corporation FBD Registers (rev 20)

00:16.1 Host bridge: Intel Corporation FBD Registers (rev 20)