When SB drops, what motherboard to get for 2600k?

So basically, "don't worry about 16x/8x?"

Correct, in this bracket 8x/8x is normal and isn't really a noticeable performance hit. The NF200 is almost snake oil because of the increased latency. For two cards in Crossfire or SLI it's not worth it.

If you must have 16x/16x you go with the higher end socket, for Sandy Bridge that's LGA 2011. Previously it was LGA 1366 with the x58 chipset.
 

That's about right for Asus. I don't need SLI, I already have a HD 6970 that I don't plan to crossfire unless I can get a really good deal on a 6990.

I just need a couple of PCI Express 1x ports for my NIC & a Sound Card. If the basic P8P67 isn't more than $140 I'll probably get that. Otherwise I'll probably go MSI for a basic board.

http://usa.asus.com/product.aspx?P_ID=Qx3PdnZI9Pq9BcIU

EDIT, the P8P67 DELUXE isn't bad at all. It just depends on how much they want for it.

http://usa.asus.com/product.aspx?P_ID=FpufhQASBFHNvccl
 
The Evo and the Pro look like they will be mid-range in terms of price yet still offer most of the main features (namely 8x8x PCIe for SLI/CFX)...looking like top picks at the moment...
 
so if i don't want crossfire/sli gigabyte p67-a ud3p will satisfy my needs correct?
 
I'm getting the impression that the P8P67 EVO may not be available in the US at launch, since it is the only P8P67 model that doesn't show up for sale in Google Shopping.
 
I'd like to get the one with the nvidia add on chip to enable full 16x bandwidth for 2 cards, which would mean a gigabyte UD7 for futureproofing. I don't have the same faith as other people that 8X will be good enough for an SLI setup going out a few years into the future, which is as often as I upgrade. Really I should just wait for 2011, but that remains to be seen. Most will get the gigabyte UD4 for the basic SLI capability (16X/8X) ... I don't blame them since there's a 150 dollar price difference.

Sorry, but your understanding of how the nForce 200 MCP works is off base. First off, it doesn't add PCI-Express lanes. It works like a multiplexor or expander. It has it's own lanes, but it interfaces through the system via the chipset's existing PCI-Express lanes. For example: if you've got an X58 chipset based board you have 36 PCI-Express lanes. The nForce 200 MCP uses all available lanes to communicate to the chipset and thus the rest of the system. So your bottleneck is still the chipset's native PCI-Express lanes. No matter what, you've only got 36 lanes worth of bandwidth. Period, end of story. The only thing the nForce 200 MCP does for you is to add flexibility to the configuration. Where a board without it would typically have a fixed slot configuration, the nForce 200 MCP allows the flexibility of a more dynamic setup. So instead of being fixed to 8x8x4, you could have 8x8x8x8, 16x16x16x4, 16x8x8, or whatever you'd like to have. This flexibility comes at a cost in the form of heat, power, added latency, PCB real estate space and of course increases the cost of the board.

Right now graphics cards aren't really saturating PCI-Express 2.0 x8 slots. Therefore there would be no benefit or almost no benefit to a motherboard which had 64 PCI-Express lanes natively. A "psuedo PCI-Express expander" is worth even less.

Not likely since x58 is triple channel and P67/H67 is dual channel. LGA 2011 that comes out later this year will be triple channel.

Fixed that for you.
 
I think the maximus IV or the maximus IV gene will both be good boards. Never had one board asus board.
 
It's just dual channel? Damn, I thought it was duel channel. Gave each of my DIMMs a flintlock pistol and had them stand back to back and take ten paces...
 
Will there be any 6 ram slot boards like most of the x58's were?
Afaict desktop DDR3 is limited to two modules per channel so on a desktop board for a dual channel socket you will only see four slots.

Registered ECC DDR3 can support 3 modules per channel. For LGA1156 there were server boards and processors that did support 3 modules per channel I have no idea if the same will apply for sandy bridge.
 
Sorry, but your understanding of how the nForce 200 MCP works is off base. First off, it doesn't add PCI-Express lanes. It works like a multiplexor or expander. It has it's own lanes, but it interfaces through the system via the chipset's existing PCI-Express lanes. For example: if you've got an X58 chipset based board you have 36 PCI-Express lanes. The nForce 200 MCP uses all available lanes to communicate to the chipset and thus the rest of the system. So your bottleneck is still the chipset's native PCI-Express lanes. No matter what, you've only got 36 lanes worth of bandwidth. Period, end of story. The only thing the nForce 200 MCP does for you is to add flexibility to the configuration. Where a board without it would typically have a fixed slot configuration, the nForce 200 MCP allows the flexibility of a more dynamic setup. So instead of being fixed to 8x8x4, you could have 8x8x8x8, 16x16x16x4, 16x8x8, or whatever you'd like to have. This flexibility comes at a cost in the form of heat, power, added latency, PCB real estate space and of course increases the cost of the board.

Right now graphics cards aren't really saturating PCI-Express 2.0 x8 slots. Therefore there would be no benefit or almost no benefit to a motherboard which had 64 PCI-Express lanes natively. A "psuedo PCI-Express expander" is worth even less.

so inotherwords, these nforce controllers might come in handy for someone who wants to do quad SLI, but they are going to be limited by the number of pci express lanes the chipset supports. It's like taking lots of water and trying to pour it into a funnel, the funnel lets you add the water faster but the weak link is the narrow part of the funnel where the water drains through. No matter what, you'll have to cram all that data back through those 20 lanes. So those 4 cards would effectively share 20 lanes, giving each 5 lanes. These people should wait for socket 2011. Am I right on this? That seems fairly limiting for a futureproofing build, I may just wait for socket 2011 or get a 2500k with a cheap motherboard and resell em.
 
Last edited:
Afaict desktop DDR3 is limited to two modules per channel so on a desktop board for a dual channel socket you will only see four slots.

Registered ECC DDR3 can support 3 modules per channel. For LGA1156 there were server boards and processors that did support 3 modules per channel I have no idea if the same will apply for sandy bridge.

Not quite. How many RAM slots a given board can have is determined by the memory controller. LGA1156 boards typically support only 4 slots and two channels. Optionally however some boards actually broke out one memory channel into four slots. Either you can use a single double-sided or dual ranked DIMM in the channel or you could use two separate single-ranked or single-sided DIMMs in two slots. This is more common in server / workstation boards were single-ranked DIMMs are more common. In the desktop world we don't see too much of this.

Here is a crude view of how it works:

Channel A: DS Channel B1: DS Channel B2: NP
or
Channel A: DS Channel B1: SS Channel B2: SS

The "NP" stands for not populated or otherwise empty. So again you can use a dual ranked or double sided DIMM in channel A and and use the other two slots for single ranked or single sided modules, or use four double sided DIMMs and discard the remaining slots in channel B. It's an either or thing. In fact I confirmed this through Supermicro's web site here.

LGA1366 processors can support more channels and more modules because it was designed to do so. You can use a maximum of 6 4GB modules for a total of 24GB of RAM compared to P55 / LGA1156's 16GB of RAM using 4 4GB modules. LGA1366 processors simply support 2 double sided / dual ranked memory slots in three separate channels for a total of six.

so inotherwords, these nforce controllers might come in handy for someone who wants to do tri SLI or quad SLI, but someone would be foolish to do this with sandy bridge because they are going to be limited by the number of pci express lanes the chipset supports. It's like taking lots of water and trying to pour it into a funnel, the funnel lets you add the water faster but the weak link is the narrow part of the funnel where the water drains through. No matter what, you'll have to cram all that data back through those 20 lanes or however many there are. These people should wait for socket 2011. Am I right on this? That seems fairly limiting for a futureproofing build, I may just wait for socket 2011.

No. It's not necessary either way. The nForce 200 MCP doesn't really get you more bandwidth because the chipsets lanes are always a choke point because that's how they interface with the chipset. Benchmarks and tests on boards equipped with nForce 200MCPs and similar boards without show almost identical performance. In most cases nForce 200 MCP equipped boards run slower due to higher latency than boards without. In rare circumstances the nForce 200MCP equipped boards are slightly faster, but the margin is again miniscule.

In other words all the nForce 200MCP's do is make for clever marketting, meet requirements for SLI certification and they do allow for the slots to be slightly more dynamic about how many lanes which can be allocated like the old switch cards on some early SLI boards did but without the hassle of flipping them. They don't do much for you though. In fact many people will argue that you are really better off without the nForce 200MCP's. That's pretty much my stance on the matter.
 
Last edited:
ok, thanks for giving me an education. ha. Now I have to tell everyone who I recommended the top end motherboards too: forget about them.. get the UD-4 instead, or even something cheaper. They will be fine for a basic SLI setup for now. It should be cheap enough to get the 2500k now, with a low end motherboard, then upgrade those 2 components to socket 2011 when the time comes. Now I know what to order... thanks for clarifying. I would say that the nforce chip made sense a few years ago, but we are much closer to saturating the pci express 2.0 specs especially with only 20 lanes doing something like quad SLI... The 40 lanes of PCI Express 3 should come in handy then.

so in essence, nforce 200 is outdated at this point and rather useless.. you're not gonna have the bandwidth to run 3-4 cards well on socket 1155 no matter what.
 
Last edited:
If you want 3-Way or CrossfireX in a 3 card setup you really need to go for an X58 based setup at this time. There is just no way of getting around the chipsets and processors PCI-Express lane limitations at this time.
 
Sorry, but your understanding of how the nForce 200 MCP works is off base. First off, it doesn't add PCI-Express lanes. It works like a multiplexor or expander. It has it's own lanes, but it interfaces through the system via the chipset's existing PCI-Express lanes. For example: if you've got an X58 chipset based board you have 36 PCI-Express lanes. The nForce 200 MCP uses all available lanes to communicate to the chipset
Afaict the nf200 can only use 16 lanes to communicate with the chipset or CPU. It then provides 32 lanes for cards.

IIRC nvidia added some tricks to the NF200 to allow broadcasting to both cards and of course the cards won't always be using their full bandwidth all the time. A bridge allows one card to use the full x16 while the other card isn't using it while just bifrucating the lanes won't allow that.

Be aware that there are two ways to connect a NF200 to a LGA1156/LGA1155 CPU. Generally the sensible thing to do is to connect all the lanes to the NF200 but for some reason EVGA seem to only connect 8 lanes to it (which provides more slots but worse performance).

Toms hardware did a comparision of the impact of the NF200 for P55 users and concluded it was benificial with two cards (provided it was connected up in the sensible way) and pretty much vital with three.

http://www.tomshardware.com/reviews/p55-crossfire-nf200,2537.html
 
Afaict the nf200 can only use 16 lanes to communicate with the chipset or CPU. It then provides 32 lanes for cards.

IIRC nvidia added some tricks to the NF200 to allow broadcasting to both cards and of course the cards won't always be using their full bandwidth all the time. A bridge allows one card to use the full x16 while the other card isn't using it while just bifrucating the lanes won't allow that.

Be aware that there are two ways to connect a NF200 to a LGA1156/LGA1155 CPU. Generally the sensible thing to do is to connect all the lanes to the NF200 but for some reason EVGA seem to only connect 8 lanes to it (which provides more slots but worse performance).

Toms hardware did a comparision of the impact of the NF200 for P55 users and concluded it was benificial with two cards (provided it was connected up in the sensible way) and pretty much vital with three.

http://www.tomshardware.com/reviews/p55-crossfire-nf200,2537.html

Well I was thinking of my EVGA X58 3X SLI Classified when I posted that. It uses two nForce 200MCP's which each take 16 PCI-Express lanes. X58 has 36 lanes but 32 of them are taken for the nForce 200MCPs. The others are used for onboard devices or PCI-Express x1 slots. In any case our testing showed no improvement or and in fact typically shows a slight reduction in performance using the nForce 200MCPs. They aren't much if any help.

EDIT:

I remember this article you linked to on Tom's. Essentially P55 benefits from it because the nForce 200MCP works around the Lynnfield's PCI-Express controller's limitations. Essentially Lynnfield can only host two PCI-Express devices. The nForce 200MCP basically allows more devices to get multiplexed through it and as far as the CPU is concerned, the nForce 200MCP is but one device. The nForce 200MCP uses it's own logic to handle traffic to each device which is connected to it rather than the processor doing so. The X58 chipset on the other hand does not need this, nor benefit from nForce 200MCP's.
 
Last edited:
You may be interested to note that expc is listing the UD7 for $312 CAD.

http://www.expc.ca/i-16726-GIGABYTE_GA_P67A_UD7.html

why do you guys think the ud4 is priced at $290 and the ud5 is priced lower at $248 on their (expc.ca) website?
http://www.expc.ca/i-16724-GIGABYTE_GA_P67A_UD4.html
http://www.expc.ca/i-16725-GIGABYTE_GA_P67A_UD5.html

I compared all three products on gigabytes website and I'm having trouble differentiating a $40 difference. If anything I'd think the ud5 would be more expensive due to more usb 3.0 ports and handy CMOS switches. http://www.gigabyte.com/products/comparison/list.aspx?ck=2&pids=3648,3647,3646
 
why do you guys think the ud4 is priced at $290 and the ud5 is priced lower at $248 on their (expc.ca) website?
http://www.expc.ca/i-16724-GIGABYTE_GA_P67A_UD4.html
http://www.expc.ca/i-16725-GIGABYTE_GA_P67A_UD5.html

I compared all three products on gigabytes website and I'm having trouble differentiating a $40 difference. If anything I'd think the ud5 would be more expensive due to more usb 3.0 ports and handy CMOS switches. http://www.gigabyte.com/products/comparison/list.aspx?ck=2&pids=3648,3647,3646

Gouging? Most places have the UD4 around $200. Maybe it's a pricing error.
 
so is that where the price of the ud5 should be based on early reports ?
i'm getting excited about sb. 2600k at 3.4 out of the box sounds wiked. and if early reports are anything to go by, this chip will overclock like a beast.
 
Trying to decide between asus P8P67 PRO/EVO...
I was about to go with PRO, but EVO, looks like to have a better cooling, though isnt it supposed to have no northbridge? so whats cooling these heatpipe? i'm little confused...
 
Trying to decide between asus P8P67 PRO/EVO...
I was about to go with PRO, but EVO, looks like to have a better cooling, though isnt it supposed to have no northbridge? so whats cooling these heatpipe? i'm little confused...

It may be an nForce 200MCP.
 
Going to go Gigabyte this time around, my last two ASUS motherboards died on me. :(
 
I'm now hella confused.

I am currently on a Core Duo 2, and have been holding out for the new Sandy Bridge CPUs since they were first announced, was all excited that I've skipped the last few iterations of Sockets and could finally settle on the LGA1155 but now I'm seeing mention of this LGA2011 in Q3 2011 ><.

Will my CPU be forward compatible? My RAM obviously won't because I'd be going from dual channel to triple/quad? I'm looking at a single GPU setup at the moment but was going to move to dual (Crossfire) after I recovered from the initial $$$ sink haha.

Game, Video Editing are on my to-do list.
 
I'm now hella confused.

I am currently on a Core Duo 2, and have been holding out for the new Sandy Bridge CPUs since they were first announced, was all excited that I've skipped the last few iterations of Sockets and could finally settle on the LGA1155 but now I'm seeing mention of this LGA2011 in Q3 2011 ><.

Will my CPU be forward compatible? My RAM obviously won't because I'd be going from dual channel to triple/quad? I'm looking at a single GPU setup at the moment but was going to move to dual (Crossfire) after I recovered from the initial $$$ sink haha.

Game, Video Editing are on my to-do list.

1155 will be the mainstream socket, and will continue to be for a couple years at least. It will work just fine for dual card crossfire and gaming (which generally doesn't utilize more than 4 cores). 2011 is a high-end socket, with support for more cores and more memory and PCIe bandwidth. Both of those sockets will be available concurrently, with the 1155 just being available a bit sooner. Expect both of them to be discontinued around the same time, as well.

Right now, the top 1155 chip is expected to sit around $350. The cheapest 2011 chip will likely be over $500, going up to the $1000 range. For a gaming build, with only dual card SLI/xfire, 1155 is plenty fine.
 
Right now, the top 1155 chip is expected to sit around $350. The cheapest 2011 chip will likely be over $500, going up to the $1000 range. For a gaming build, with only dual card SLI/xfire, 1155 is plenty fine.

BS - s2011 is replacement for s1366 and I dont see any reason why would its (non-Xeon) CPUs start that high. Its much more likely that their prices will start at around 300$ just like s1366 CPUs.
 
I just want to see a liget review that shows a real detail yes we all know this is going to kick some ass but how much ..... and where the hex and octa cores ... cause i want 8 true core maybe 12! with HT .... ummm hummm!
 
I just want to see a liget review that shows a real detail yes we all know this is going to kick some ass but how much ..... and where the hex and octa cores ... cause i want 8 true core maybe 12! with HT .... ummm hummm!

Hex & Octo Core CPUs will be on LGA 2011. Unless there is a very good deal I'm going to hold out until Ivy Bridge for an HT Octo-core monster CPU. A 2600k should be plenty for me for the next two years.
 
It may be an nForce 200MCP.

I looked over asus site, and I don't think there's a 200 under there. The specs list "2 x PCIe 2.0 x16 (single at x16 or dual at x8/x8 mode)". I think gigabyte is doing the same thing. The UD4 has an extra heatsink with a heat pipe. I'm not sure it really does anything at all.

I think I'm going with the p8p67 pro. I like the UD4, too, but the lack of EFI kind of kills it for me.
 
Last edited:
I think I'm going with the p8p67 pro. I like the UD4, too, but the lack of EFI kind of kills it for me.

i'm deciding between these two boards and right now i'm leaning towards asus - EFI is a big selling point to me. my only concern with this board is the proximity of the heatsinks to the cpu socket. i'm wondering if they will interfere with large aftermarket coolers. and i'm fighting with the fact that the Gigabyte board looks better...absolutely meaningless but it's still a nagging thought :D
 
Back
Top