Mac Pro "Sandy Bridge-E" Performance Analysis

[VirtualRain]

Stunted??? Maybe for the single core oriented "extreme" , "elite" gamer but on a budget crowd. For the workstation folks there is very little lost here. A dual package E5 2600 has 80 , yes eight zero , PCI-e lanes. Some of the current Xeon boards could throw all the embedded switches away. There is zero reason to hook to the low level I/O hub if you really have large I/O throughput issues.

The low level I/O hub is just that... a hub for the slower I/O. Turning it into a major intersection for double digit Gbps traffic is a flawed notion. The X58 went into the CPU package; not the I/O hub.


P.S. No I don't think that Apple will roll out a fully hooked up 80 lane monster dual package model. There is enough in 40 lanes to toss the switch in the current Mac Pro's oriented design though.

With PCIe 3 just around the corner, SATA3 SSDs becomming mainstream, and USB 3 starting to catch on, buying a system with this stunted chipset is not wise at this juncture. If this chipset was in the 2010 Mac Pro, that would be ok, but not in a machine that's due to launch on the cusp of 2012. I'll gladly wait for the next iteration. There's just not enough here to warrant an upgrade.

 

[jnpy!$4g3cwk]

The low level I/O hub is just that... a hub for the slower I/O. Turning it into a major intersection for double digit Gbps traffic is a flawed notion. The X58 went into the CPU package; not the I/O hub.


P.S. No I don't think that Apple will roll out a fully hooked up 80 lane monster dual package model. There is enough in 40 lanes to toss the switch in the current Mac Pro's oriented design though.

Apple should do an 80-lane version, though. You can't do 3 PCIe x16 slots without it, and that is what people will want - dual high-power x16 for graphics, and 1-2 x16 slots for high-speed network (e.g. 4X QDR IB or 40 GbE). Not to mention leftover lanes/bandwidth for 8-10 2.5" SAS/ SATA 3.1 6 Gbps disk/SSD slots.

 

[nanofrog]

Apple should do an 80-lane version, though. You can't do 3 PCIe x16 slots without it, and that is what people will want - dual high-power x16 for graphics, and 1-2 x16 slots for high-speed network (e.g. 4X QDR IB or 40 GbE). Not to mention leftover lanes/bandwidth for 8-10 2.5" SAS/ SATA 3.1 6 Gbps disk/SSD slots.

It would be nice generally speaking for graphics workstations, or for software that supports GPGPU processing, but I don't expect Apple to produce such a system as they don't offer much in the way of enterprise support.

All I could find in the Apple store:

• RAID card (which is garbage)
• Dual Channel Fibre Channel card
• Quad Channel Fibre Channel card

 

[deconstruct60]

Apple should do an 80-lane version, though. You can't do 3 PCIe x16 slots without it, and that is what people will want - dual high-power x16 for graphics, and 1-2 x16 slots for high-speed network (e.g. 4X QDR IB or 40 GbE).

With PCI-e v3.0 you can because the above amounts to 2 x8 slots and 1-2 x8 slots. In the latter case, if it is 1 store card then with v3.0 in 16-8-8-4(-4) setup this fits exactly. It is not lanes but bandwidth that people need. If somehow the E5 did slip back to v2.0 in deployment even 4 slots with no embedded switches would be an improvement.

In short, v3.0 reduces the need for x16 physical slots.

Furthermore there aren't any IB cards now at any speed so I extremely doubt 40Gb ones will show for new Mac Pro during its one year tour as "top of the line". I suspect it will be a while before any 40GbE cards show either since have to hook to a switch which few will be willing to pay for.

That task that would more likely heavily drive a 16x v3.0 set-up is something like a high end GPGPU card where constantly shuttling new data and new results back and forth at lower latencies. Again, none of those currently supported and widely deployed.

Not to mention leftover lanes/bandwidth for 8-10 2.5" SAS/ SATA 3.1 6 Gbps disk/SSD slots.

Chuckle... if you paid $10,000+ to access some disk over IB I seriously doubt you are going to screw around with a large number of local SATA drives. Pick one or the other. Both is getting into the blowing money just because you can range in the workstation context for 99% of workloads.

Nor is the Mac Pro going to be suited to being one of the ultimately fastest file servers you can by. (user the SATA for disk and IB to get data out to the clients). It will be fast enough for lots (95+%) of folks though.

The other problem is if Apple sticks with the CPU+ Memory + I/O hub daughter card design. 80 pins just for PCI-e is alot. For the single package version that's 40 dead pins on the card.

 

[deconstruct60]

With PCIe 3 just around the corner,

PCIe v3 is not coming to the I/O hub core chips any time soon. They just got v2. That's going to be the cap for several iterations.

SATA3 SSDs becomming mainstream,

Not really. They will have to sink so sub $100 and reasonable capacity to hit mainstream numbers.

and USB 3 starting to catch on,

Is currently implemented by a 1x PCI-e v2 link to the controller chip.
You can easily hook that up to the I/O hub now with this new chipset. Non-issue. If trying to hook 2-3 of them up that would be an issue, but there is no reason to do that.

If just prune four PCIe v3.0 lanes off for embedded internal connecdtions that are v2.0 speed if you just add a 4-to-8 switch to the board with minimal buffer you can hook up a 4x TB controller, and a 4x SATA III controller to the switch and still have 4 slots. TB and SATA III added without breaking sweat. It is the whole system that people buy... not the core chipset or CPU core.

This is better, IHMO, than trying to just either one of those into the core chipsets. With this board design you would have two independent SATA controllers. You could easily segregate the SSDs to one (apps , OS , and scratch) and the HDDs ( media originals and backups ) to the other.
That is going to be a more balanced system. You have board space because pulled the X58 off the board. It isn't like the Mac Pro has the space constraints of an iMac, mini, or laptops.

buying a system with this stunted chipset is not wise at this juncture.

What is stunted in the board design skills if they cant work out how to put the 40 v3.0 lanes to work effectdively. Not the chipset. The primarily things looping higher end SATA controller into the core chipset does is lower price and lowers the design skill necessary. Primarily that is quest on the user side for a cheaper box. The Mac Pro isn't going to get cheaper. Apple is going to find things of value to push into the Mac Pro to keep it in the $2000+ price range. Wailing about how Intel's chipset doesn't make for a cheaper box is a mismatch.

If the E5's are capable of v3 and Intel/Apple hamstring them to v2, that's what to wail about.

[quote I'll gladly wait for the next iteration. There's just not enough here to warrant an upgrade.[/QUOTE]

Waiting another year for an update is always going to bring a better box.

 

[deconstruct60]

As i mention before the way that apple can easily divide imac & mac pro. Is remove sp mac pro. And bring dp as standard like we seen on 1,1 2,1 3,1.

No they can't. Buried in here I think is a price drop for the DP models down into the SP range. I think that is wishful thinking.

There are two advantages the SP models have.
i. They are more affordable.
ii. They have higher clock rates.


There is little the DP versions can do about that first issue. Buying two E5's is always going to be more expensive than buying just one. Throw on top of that the price differential that Intel is extremely likely to make on the price gap between 1600's and 2600's and it is substantial. The notion that there isn't a $400 gap in value difference between iMac and likely new Mac Pro is suspect.

On the second issue, there are folks who don't have "high core count" problems. They are going happier with higher clock rates. Different boxes for different submarkets. That second submarket isn't going to go away if Apple drops the single package Mac Pro. Furthermore, the folks who are obsessed with maximizing the CPU price / total sytem price ratio aren't going to buy either the iMac or Mac Pro. They are non targeted customers. If nuking the SP model because they are unhappy that is a fundamentally flawed move.




The DP versions can only compete on clock by adding $1,000s to the price. Again budget constraints are goingto be a major impediment. Some users want higher clocks and fewer cores. Those folks will gravitate to the entry Mac Pro. On the synthetic and limited core benchmarks the 3.6GHz E5 1620 is going to post a significant gap on the 3.4GHz iMac model. Not to mention the extra 2MB of L3 cache and the two additional memory controllers in the 1620. For 4 concurrent core workloads that are so small that fit in 8M cache those won't make a difference, but for the others it will.


Furthermore, the notion that 6 cores is going to put you into the $3,000+ range is flawed. The E5 1640 is 6 cores and likely the middle of the line up. Currently, the middle SP model goes for $2,899. Even if add $100 for embedded graphics (to enable TB) that is $2,999. Far less than the $3,699 price point you cited. ( with that difference, you can buy a very good 3rd party SATA III SSD drive for that amount of money and deploy real performance gains.)

For the users that have high workload on the maximum cores clockrate will sag relative to the models where the core count is capped. The DP models will have an advantage in that could go to 6 active and still be in Tubro more is split that 3,3 across the two packages. Whereas the E5 1640 would be falling out of Turbo mode with that workload.

 

[deconstruct60]

That task that would more likely heavily drive a 16x v3.0 set-up is something like a high end GPGPU card where constantly shuttling new data and new results back and forth at lower latencies. Again, none of those currently supported and widely deployed.

There is a possible lower end GPGPU option that could be widespread among Mac Pros if they embed a nominal GPU on-board to enable Thunderbolt. For example if took the 6630M assembly out of the mini and attached it to the daughterboard. Then on the single package version, they could trim off x4 lanes (possibly shared ***) and hamper GPGPU performance. On the dual package version, they could attach 16x lanne from the second package 16x lanes to the 6630M and get better performance (both graphics and GPGPU). With either card though still would just have the 16-8-8-4 coming out to the expansion slots. (and a 4x to some embedded functionality on the base board: LOM ethernet, SATA III, TB, whatever. )


So yes, they could make use of some of the "extra" 40 to boost performance, but I doubt they are going to run any of the "extra" 40 out to optional PCI-e expansion slots. In order to save costs between the SP and DP models, the base board they both utilize will have to be the same. Trying to split the SP and DP designs limit both to volumes that are probably too low to justify continuing either. The Mac Pro needs higher volume numbers not lower ones.


*** If SP daughterboard had a switch so that the 4x expansion slot was shared with embedded GPU then there would be no impact if slot went unfilled. If someone fills the 4x slot then when doing the GPGPU caculations (and no transfers) then the expansion slot would get full time. When GPGPU transfers were done they'd get in each others way. However, lots of folks can get by with using just 3 slots. If 4 high concurrent usage slots is a big deal to someone, they can trade-up to a dual package model. No switch on that model... probably not even room on daughtercard even if wanted to.
[ Kind of assuming that if make Mac Pro shorter the box will get deeper and allow for deeper daughterboard with space for small mobile GPU and limited associated VRAM. The assembly that fits into the "extra" space in a mini should be kind of small. ]

 

[VirtualRain]

Waiting another year for an update is always going to bring a better box.

Yeah, there's no doubt about it, but there is another longer term cycle that influences my buying behavior that's related to overall platform and architecture improvements. It's my belief that the best time to buy a new system is when new platform technologies have just emerged. NOT just before they arrive.

In the past, the introduction of technologies like SATA replacing IDE, DDR2 replacing DDR or the integrated IMC in Nehalem replacing the Front Side Bus, are major architectural moves that signalled to me, it's time to upgrade my system. I don't buy new systems on a CPU bump. That's all I'm saying.

Now, you may disagree, but it seems clear that SATA3, PCIe3, and USB3 are all just around the corner... probably Ivy Bridge. To buy an SB-E system that's largely based on SATA2, PCIe2, and USB2 right now, seems foolish to me.

 

[blunti]

It's my belief that the best time to buy a new system is when new platform technologies have just emerged. NOT just before they arrive.


Now, you may disagree, but it seems clear that SATA3, PCIe3, and USB3 are all just around the corner... probably Ivy Bridge. To buy an SB-E system that's largely based on SATA2, PCIe2, and USB2 right now, seems foolish to me.

While I fully agree with you, let's not forget about Thunderbolt which will be a HUGE plus for a lot of people. Being able to work with uncompressed hd videos is surely tempting to a lot of people. (At least to those people who work in the film/tv industry)

While I know there are many ways to saturate the current bandwidth of Sata II, being able to work with huge files without the additional fuss is definitely a plus.

 

[Hellhammer]

While I fully agree with you, let's not forget about Thunderbolt which will be a HUGE plus for a lot of people. Being able to work with uncompressed hd videos is surely tempting to a lot of people. (At least to those people who work in the film/tv industry)

While I know there are many ways to saturate the current bandwidth of Sata II, being able to work with huge files without the additional fuss is definitely a plus.

But Mac Pro already offers alternative solutions, some of which are even faster than TB. You could buy for example a 6Gb/s eSATA card and get 6Gb/s per port. eSATA enclosures go for pennies compared to TB ones.

For some people, TB might be a big deal but with current product offerings and pricing, it really seems like a fancy version of mDP.

 

[initialsBB]

Just heard a rumor today from someone who was at IBC 2011 in Amsterdam. They said that a source close to Apple told them the Mac Pro is OVER.

Now this is just a rumor, but if it's true I hope to hell they have something else up their sleeve to replace it.

 

[blunti]

But Mac Pro already offers alternative solutions, some of which are even faster than TB. You could buy for example a 6Gb/s eSATA card and get 6Gb/s per port. eSATA enclosures go for pennies compared to TB ones.

For some people, TB might be a big deal but with current product offerings and pricing, it really seems like a fancy version of mDP.

While this is very true, let's not forget about field editing where plugging an external TB drive would be very very usefull. Instead of dragging a 40-50 pound Mac Pro around + Monitor, you can just take the MBP out of the sleeve, plug in the external drive and ready to go.

Very convenient I would say

 

[deconstruct60]

They said that a source close to Apple told them the Mac Pro is OVER.

While possible it doesn't seem likely. [ If folks have massively bailed into iMac (and Windows ) who were Mac Pro users previously and growth went negative I can see it. I suspect though that it is just flat with the longer than usual product cycle. ]

Unless it was a source on the Mac Pro team or on the Mac OS X team then I would be skeptical. Even folks inside of Apple don't know what Apple is doing. Someone who is outside ..... errrr .... that could be they just don't have any hooks into the relatively small subset of folks who work on it.

Now this is just a rumor, but if it's true I hope to hell they have something else up their sleeve to replace it.

It isn't like they need a different form factor. Apple isn't going to roll out another Mac team (the competition for resources is probably too high). The Mac Pro would get folded into the existing Mac lines. You'd probably see a matte iMac with some SSD/HDD combo option trotted out to put up the lower end.

Plus some minis paired with some of these

http://www.sonnettech.com/product/rackmacminixserver.html

to snag some of the server duties that only need 1-2 slots.

The higher end dual package models would just be left in a vacuum.

----------

Now, you may disagree, but it seems clear that ... USB3 are all just around the corner...

USB 3.0 is "just around the corner"? You should lay off the drugs. USB 3.0 is here. I can walk into common office supply stores and find USB 3.0 devices Go search for "USB 3.0" at www.walmart.com. The grossly artificial distinction you are making is whether it is in the core chipset or implemented in a discrete component. That has very little to do with having significantly arrived in deployed systems and peripherals. The 3rd generation USB 3.0 controllers are coming soon.

 

[VirtualRain]

Just heard a rumor today from someone who was at IBC 2011 in Amsterdam. They said that a source close to Apple told them the Mac Pro is OVER.

Now this is just a rumor, but if it's true I hope to hell they have something else up their sleeve to replace it.

Interesting. It wouldn't surprise me to see this on the front page at MR one day. But soon? I doubt it. Although TB solves a lot of pains with expansion that were previously only possible with a Mac Pro, and Apple has a clear focus on the consumer market, I have to think the Mac Pro is probably still very profitable for Apple at current prices with only a very modest level of investment required every 12-18 months to refresh it.

USB 3.0 is "just around the corner"? You should lay off the drugs. USB 3.0 is here. I can walk into common office supply stores and find USB 3.0 devices Go search for "USB 3.0" at www.walmart.com. The grossly artificial distinction you are making is whether it is in the core chipset or implemented in a discrete component. That has very little to do with having significantly arrived in deployed systems and peripherals. The 3rd generation USB 3.0 controllers are coming soon.

What's with your confrontational tone? Of course SATA3 and USB3 expansion cards and peripherals can be easily had... I'm talking about what I need to see in a platform before I consider upgrading. I'm not forcing you to think the same way. An SB-E Mac Pro with a USB3 and SATA3 expansion card may feel like a good future-proof buy for some folks, that's not going to get me to fork over $3-5K for a new system. And it's not an artificial distinction, it's a signal to me... when I see those technologies in the chipset, it's more likely to trigger a purchase by me.

EDIT: I'm wondering what triggers you to buy a new system? Do you buy every refresh of the Mac Pro? If not, what do you want to see in an update that makes you want to lay out the cash to upgrade?

 

[jnpy!$4g3cwk]

With PCI-e v3.0 you can because the above amounts to 2 x8 slots and 1-2 x8 slots. In the latter case, if it is 1 store card then with v3.0 in 16-8-8-4(-4) setup this fits exactly. It is not lanes but bandwidth that people need. If somehow the E5 did slip back to v2.0 in deployment even 4 slots with no embedded switches would be an improvement.

According to the Toms Hardware article, Intel is not going to ship PCIe 3.0 in the early chip releases. No explanation for why, but, this is not official either. Maybe they will get 3.0 working in the next couple of months.

In short, v3.0 reduces the need for x16 physical slots.

Disagree completely. AMD is counting on v3.0 to reduce the immediate need for x32 slots, which otherwise they would want to support the next gen GPUs. See, e.g.,

http://www.legitreviews.com/news/11105/

Furthermore there aren't any IB cards now at any speed so I extremely doubt 40Gb ones will show for new Mac Pro during its one year tour as "top of the line". I suspect it will be a while before any 40GbE cards show either since have to hook to a switch which few will be willing to pay for.

I'm not a big fan of IB, but, lots of people are using it instead of FibreChannel. Just because Apple doesn't support it now doesn't mean they won't need to in the future.

Bottom line is, with PCIe 2.x, 40 lanes is constraining. I agree that with PCIe 3.0, the pressure to exceed 40 lanes would be significantly diminished.

That task that would more likely heavily drive a 16x v3.0 set-up is something like a high end GPGPU card where constantly shuttling new data and new results back and forth at lower latencies. Again, none of those currently supported and widely deployed.

But, could be in the next two years.

Chuckle... if you paid $10,000+ to access some disk over IB I seriously doubt you are going to screw around with a large number of local SATA drives. Pick one or the other. Both is getting into the blowing money just because you can range in the workstation context for 99% of workloads.

One user wants a standalone local RAID system, another set of users is using an IB or xGbE SAN in a studio environment, perhaps with a special node or two with local scratch disk built out of SSD. It would be nice to have the flexibility to build all three types of systems.

Now, you may disagree, but it seems clear that SATA3, PCIe3, and USB3 are all just around the corner... probably Ivy Bridge. To buy an SB-E system that's largely based on SATA2, PCIe2, and USB2 right now, seems foolish to me.

If you are waiting anyway, then, I agree about SATA3 and PCIe3, but, some folks seem to be betting that TB will be a better option than USB3, since you will have a lot of the flexibility of FireWire and higher speed than USB3.

But Mac Pro already offers alternative solutions, some of which are even faster than TB. You could buy for example a 6Gb/s eSATA card and get 6Gb/s per port. eSATA enclosures go for pennies compared to TB ones.

For some people, TB might be a big deal but with current product offerings and pricing, it really seems like a fancy version of mDP.

I don't see this as either/or. I think of TB as the FW replacement. It might be either/or TB/USB3. Not clear that USB3 can replace FW.

USB 3.0 is "just around the corner"? You should lay off the drugs. USB 3.0 is here. I can walk into common office supply stores and find USB 3.0 devices Go search for "USB 3.0" at www.walmart.com. The grossly artificial distinction you are making is whether it is in the core chipset or implemented in a discrete component. That has very little to do with having significantly arrived in deployed systems and peripherals. The 3rd generation USB 3.0 controllers are coming soon.

USB 3.0 appears to have the shortcomings of USB 2.0, only faster. You might get throughput similar to FW800, and, the latency and jitter will still be there.

 

[Hellhammer]

Disagree completely. AMD is counting on v3.0 to reduce the immediate need for x32 slots, which otherwise they would want to support the next gen GPUs. See, e.g.,

http://www.legitreviews.com/news/11105/

There is nothing that indicates an immediate need for more PCIe bandwidth due to GPUs. See e.g. this. The performance loss from x4 or x8 is minimal.

 

[xgman]

USB 3.0 appears to have the shortcomings of USB 2.0, only faster. You might get throughput similar to FW800, and, the latency and jitter will still be there.

I've processed a lot of video over my caldigit usb 3.0 port from a usb 3.0 external drive and have no latency and no jitter.

 

[Inconsequential]

Just heard a rumor today from someone who was at IBC 2011 in Amsterdam. They said that a source close to Apple told them the Mac Pro is OVER.

Now this is just a rumor, but if it's true I hope to hell they have something else up their sleeve to replace it.

Rubbish and just unfounded rumour.

Are Apple really *really* going to limit people to a iMac?

Not a chance.

There would be a rebellion.

It's coming, I recon November given that is when the chips are released.

 

[derbothaus]

Things not so dire after all:
http://www.xbitlabs.com/articles/cpu/display/lga-2011-idf.html

PCI 3.0 Support WILL happen by launch.
6x SATAlll Port Max (10 total)
36 PCIe lanes

It's a tock, no doubt.

 

[Hellhammer]

Things not so dire after all:
http://www.xbitlabs.com/articles/cpu/display/lga-2011-idf.html

PCI 3.0 Support WILL happen by launch.
6x SATAlll Port Max (10 total)
36 PCIe lanes

It's a tock, no doubt.

It's kind of weird that they talk about 36 PCIe 2.0 lanes but somehow the same lanes should also support 3.0 too Makes no sense IMO

 

[deconstruct60]

Things not so dire after all:
http://www.xbitlabs.com/articles/cpu/display/lga-2011-idf.html

PCI 3.0 Support WILL happen by launch.

Just because no one has getting the official final pass doesn't mean there aren't cards that aren't deep into the test phase. Until the PCI-e group starts announcing passing devices, no one is going to "pre declare" that they have passed.

6x SATAlll Port Max (10 total)

The intel board uses just two. Wouldn't be surprising if the Workstation reference board with the slightly different parts does also. Apple is likely to follow closely in the trail of the reference design (minus all these tweaker overclocking features.)

36 PCIe lanes

That actually would be bad if extended to the E5's and was a late adjustment. If Intel had to chop down the E5's 1600 and 2600 by 4 lanes that has probably screwed up lots of workstation designs that maximized throughput.

 

[deconstruct60]

It's kind of weird that they talk about 36 PCIe 2.0 lanes but somehow the same lanes should also support 3.0 too Makes no sense IMO

v3.0 lanes are v2.0 capable lanes. It is backward compatible. Just like putting v1.0 devices in v2.0 slots.

They may have had to choke down the PCI-e switch because it really couldn't handle 40 active v3.0 lanes. That is a huge amount of bandwidth.
That's about 40GB/s. Given the old X58 era PCI-e controllers had 36 lanes then moving just that many to v3.0 would be doubling the throughput performance. Adding 4 more increases the level of difficulty.

Given the "slow" I/O controller has v2.0 lanes now some designers can just punt some v2.0 only slot lanes to that controller if have enough left over (after adding stuff like USB 3.0 , etc. with discrete chips. ) .

 

[Hellhammer]

v3.0 lanes are v2.0 capable lanes. It is backward compatible. Just like putting v1.0 devices in v2.0 slots.

I am aware of that. 2.0 slot can't, however, operate at 3.0 speeds. You don't say it as "2.0 slot with support for 3.0 speeds" if it's 3.0 slot. You say that it's 3.0 slot which is backwards compatible with previous standards.

Right now, it seems like we won't know this for sure until the retail units have been reviewed. TH has a SB-E CPU, yet they claim the PCIe is 2.0. Too much dissimilar info.

 

[deconstruct60]

I am aware of that. 2.0 slot can't, however, operate at 3.0 speeds. You don't say it as "2.0 slot with support for 3.0 speeds" if it's 3.0 slot. You say that it's 3.0 slot which is backwards compatible with previous standards.

This issue is that they are embargoed from self declaring they have v3.0 slots. Until the officially stamped v3.0 tests are passed and PCI org says "OK" they aren't v3.0 slots.

The standards org aren't going to get any member or test fees if vendors just self declare they passed. So if they are playing within the rules they can't say they have v3.0 slots now. They are trying to tip toe around it by saying these possibly will be v3.0 later on. They are likely designed to be v3.0 slots. They probably are v3.0 slots. The only thing that is missing is the "official stamp". This is like the time between finishing college and marching across the stage to pick up the diploma. You are finished. Just missing the pompt and circumstance.

Right now, it seems like we won't know this for sure until the retail units have been reviewed.

Sure there are likely some lower tier vendors who just think they are going to pass because they haven't been working with that many vendors. Intel is at a nexus point where it would be extreme strange if they haven't seen a good number of PCI-e v3.0 in their labs or been looped into diagnostic sessions with folks with cards in their labs.

TH has a SB-E CPU, yet they claim the PCIe is 2.0. Too much dissimilar info.

Is that a final board? In order to minimize cross comparison glitches, Tom's hardware could have a board that hardwired for v2.0. Simply to maximize compatibility so that these blogs/talking-heads can do their "we used the same disk/ graphics card / etc in each motherboard" comparisons. The last thing Intel (or MB vendor) would want is for a high profile media site to moan and groan about glitches. Neither would it be particularly prudent to depend upon TH's site to pick a "good" v3.0 board since there are no cross vendor , certified test to pass.

 

[Hellhammer]

This issue is that they are embargoed from self declaring they have v3.0 slots. Until the officially stamped v3.0 tests are passed and PCI org says "OK" they aren't v3.0 slots.

Has Ivy Bridge passed the testing then? Because Intel is quite openly stating that there is PCIe 3.0 support.

Is that a final board? In order to minimize cross comparison glitches, Tom's hardware could have a board that hardwired for v2.0. Simply to maximize compatibility so that these blogs/talking-heads can do their "we used the same disk/ graphics card / etc in each motherboard" comparisons. The last thing Intel (or MB vendor) would want is for a high profile media site to moan and groan about glitches. Neither would it be particularly prudent to depend upon TH's site to pick a "good" v3.0 board since there are no cross vendor , certified test to pass.

Nothing is final at this point. The board could make it to retail as it is but at the same time, it could also be just a prototype. The problem is that nobody knows for sure.