Intel D58XSO (Smackover)

[switch_abx]

Sandog, any idea when the Intel Board D58XSO “Smackover” will become available? This one will be with the new Intel Core i7 (aka Nehalem) and "Tygersburg" chipset. QuickPath instead of FSB, I think Southbridge and Northbridge integrated into one, support for Intel VT-d etc. etc.

Truly a revolutionary processor and chipset and not an evolutionary IMHO.

[Sandog]

In September sometime. This is when Intel is suppose to introduce both the first Nehalem chip (3.2 Extreme Edition CPU) and the board it will run on (X58 Bloomfield.)

[switch_abx]

Thanks for your feedback.

Bloomfield was the Intel development codename for the Nehalem-EP processor and not of the chipset AFAIK.

The chipset's development codename is the Tylersburg DT (36-S?) chipset AFAIK. I presume this chipset is called now the X58?

Or am I mixing up things?

[Sandog]

Yes I believe Bloomfield is the CPU name. Intel is suppose to introduce the new hardware in September but hardware availability might not happen at that time. I am still waiting for Intels P45 chipset board which was introduced in the beginning of June.

[Sandog]

Here's a nice pic of the X58.

Legit Reviews - Intel DX58SO 'Smackover' Motherboard Does Triple-Channel Memory - Intel SmackOver Details Come To Light

[Sandog]

Pic from above post (LegitReviews.)

intel_dx58so_smackover.jpg

This is the most beautiful mobo I have ever seen. Red fan headers, black power connectors, same chipset and regulator HS as the X48. Maybe its the neat rearanging of the mobo but this board looks killer. If you look close you will notice something never before on an Intel board, two 7-segment LED displays. Probably for post codes. I just hope Intel continues to support my lowly DP45SG board after this board comes out.

[ATINUT]

Sorry Sandog, after reading on Legit Revies, that the memory drops down to single channel mode only, when all the DIMM slots are full, this is one board I certainly wouldn't want.

Far as I'm concerned, Intell needs to go back to the drawing board, and allow for triple channel performance when all four DIMM slots are filled on this board.

[CK8-04]

Originally Posted by ATINUT

Sorry Sandog, after reading on Legit Revies, that the memory drops down to single channel mode only, when all the DIMM slots are full, this is one board I certainly wouldn't want.

Far as I'm concerned, Intell needs to go back to the drawing board, and allow for triple channel performance when all four DIMM slots are filled on this board.

I believe this is a limitation of the CPU's integrated memory controller and not the board. AMD's K8 CPUs also have limitations when four double sided DIMMs are used. In the initial single core CPUs, the memory speed would drop to 333MHz, however, in later revisions, the memory would be able to run at 400MHz but with a 2T command rate.

Additionally, how would triple channel be attained with four memory sticks? With 3 memory sticks, that's one channel per stick. Memory stick capacities are always powers of 2, so they can never be evenly divided by 3.

[ATINUT]

From what I've read on the the X58 chipset in the past few months as I recall, Intel said that each DIMM slot would have 3 channels, meaning 3 channels per DIMM, but apparently they decided not to implement that capability.

I wouldn't be suprised to see 3rd party X58 boards come out, with the ability to have all four DIMM slots filled with no performance loss.

[CK8-04]

Please send me a link to where you've read this. I'm very interested in learning how Intel was planning to achieve triple channel bandwidth from a single stick of memory.

I would certainly be surprised to see 3rd party boards come out with the ability to have four DIMM slots filled, let alone six, with double sided DIMMS and still retain maximum performance. Four unbuffered double sided DIMMs alone would put a pretty heavy load on the memory controller...

[ATINUT]

Originally Posted by CK8-04

Please send me a link to where you've read this. I'm very interested in learning how Intel was planning to achieve triple channel bandwidth from a single stick of memory.

I would certainly be surprised to see 3rd party boards come out with the ability to have four DIMM slots filled, let alone six, with double sided DIMMS and still retain maximum performance. Four unbuffered double sided DIMMs alone would put a pretty heavy load on the memory controller...

Look at the attached picture here, underneath the Kingston specs for the ram, it clearly states that the X58 chipset supports, 3 channels per socket, which isn't true if all four DIMM sockets are full on this board, which is why I wouldn't want this board.

Gigabyte's X58 board has 6 DIMM sockets on it.

[CK8-04]

That's 3 DDR3 memory channels per CPU socket, not memory socket. Multi-channel memory technology works like a HDD RAID. You can't magically achieve a three fold bandwidth increase from a single stick of memory just as you can't achieve a boost in performance with a single HDD. Look at how dual channel DDR works. If there is only one DIMM on the board, the memory subsystem operates in single channel mode because there is only one DIMM to channel data from. However, if there are two DIMMs on separate channels, the memory controller can then split data as if the two DIMMs were a RAID and channel the data between both DIMMs, increasing bandwidth.

I am fully aware that the Core i7 CPU supports a maximum of 6 DIMM slots. I don't know why and see no reason for Intel not to include the 2 extra slots other than to save a little bit of manufacturing cost and space on the board. Personally, I'd still get this board if I were building a Nehalem system because 1) it's an Intel board and that means quality components are used instead of the Realtek POS on Taiwanese boards and 2) I wouldn't ever want to use more than 3 DIMMs so the memory controller doesn't take too much of a beating.

I have a pretty good feeling that other boards will exhibit the same performance loss if 3 DIMMs are placed in 3 channels and one DIMM is placed randomly elsewhere.

[ATINUT]

Intel pulled something similar on their X48BT2. Their specs for the X48 chipset state that it supports 1600MHZ DDR3 memory on all DIMM slots, and yet they only support 1600MHZ DDR3 memory in only the first two DIMM slots. I looked up the specs on the X48 chipset on Intel's website, and the specs specifically state that the X48 chipset does support 1600MHZ DDR3 memory on all DIMM slots.

This is why 3rd party motherboards in some cases, actually look better than Intel's boards, because some of the third party manufacturers of boards know not to cheat their customers, and know how to properly program a motherboard bios to support all memory options, including on the X58 chipset being able to have all the DIMM slots full, without a significant performace decrease.

This is why, if I were to get an X58 board, I would be taking a look at Gigabyte, and other 3rd party manufacturers.

In this case, Intel's X58 board at this stage of the game is not going to be a good one, unless Intel does something to improve memory performance on the board.

[fvbounty]

I know this is just one review, and its on X48 chipset but Intels last board didn't look to good to this reviewer!!!!!

[H] Enthusiast - Intel DX48BT2 Motherboard

[ATINUT]

Originally Posted by fvbounty

I know this is just one review, and its on X48 chipset but Intels last board didn't look to good to this reviewer!!!!!

[H] Enthusiast - Intel DX48BT2 Motherboard

I saw that review. I haven't been impressed with the X48BT2 and now the upcoming X58SO.

If I do anything this fall, such as getting a quad core, get a Q6600 quad core, which is still available for purchase at retail on the net, and in local shops and stick with my trusty 975XBX2.