Future LGA 2011 chipsets

[wiyosaya]

I'm not up on the future of LGA 2011, however, it is my understanding that the X79 chipset does not have either of native USB 3.0 support or SATA 6GB/s support.

Are there any planned LGA 2011 chipsets that will support both, and if so, what is the timeline for it(s) release?

Thanks.

 

[jaquith]

The X79 does indeed offer (2) SATA3 and (4) SATA2 ports, all of which can run up to a (6) drive array. As far as USB 3.0 it is not native to the chipset. Further, the current SB-E is 32-lanes of PCIe 3.0 (40 lanes total), and secondary 8 lanes of PCIe 2.0.

The current SB-E CPUs are PCIe 3.0.

Personally, I really couldn't care less about native USB 3.0 support, and I am not aware of any slated replacement to the LGA 2011 X79 replacement; it will support the Ivy Bridge-E CPU later this year.

 

[wiyosaya]

Thank you for your reply. For the capabilities of the X79 chipset, it would have been better if I had checked here first.

While the lack of native USB 3.0 support is not a deal breaker for me since MB manufacturers are including that capability, I would prefer to have it.

In addition, I was unable to find any info about Ivy Bridge-E being slated for release in the LGA 2011 form factor before I posted my question, however, searching for "Ivy Bridge-E" and "LGA 2011" does turn up results with a possible CPU release in Q4 of this year. I am not sure I will wait until then, however, the fact that the TDP on Ivy Bridge is less, perhaps substantially, than that of Sandy Bridge may be reason enough for me to wait to upgrade.

Right now, I am in an information gathering mode, and your reply helped my efforts.

Thanks again.

 

[wiyosaya]

Best answer selected by wiyosaya.

 

[jaquith]

There are Pros and Cons to 32nm vs 22nm lithography; my concern is potentially faster degrading with the 22nm - but it's just an 'unsubstantiated' concern. The lower TDP helps with heat but if your vCore is more limited (degradation) it also limits your OCs.

Again, the only advantages to a native USB 3.0 are in Boot and BIOS, that's it. So 99.99% of the users out there it means zip, and 99.8% of the USB drives cannot saturate USB 2.0 (60 MB/s). Using some oddball boot drive, hmm... get eSATA SATA3 (e.g. ASMedia). Otherwise, it's a I want it thing with no use for it now or later.

The initial C1 SB-E had a Vt-d bug (updated BIOS corrects it), but so far the 'fixed' SB-E C2 seems to degrade faster with 1.5Xv vCores. It's recommend to avoid >1.45v vCore (don't forget LLC adds up to +0.05v), and exceeding 1.30v VCCSA or VTT seems to do the same degradation.

Unless you really need 6-core (8-core on die, 2 disabled due to low yields) <or> 32 lanes of PCIe to the GPUs then IMO either the SB or IB LGA 1155. The IB has been delayed and I won't be too surprised to see IB-E in 2013. The IB will be the first to offer USB 3.0 native IF using the newer LGA 1155 Panther Point chipets (Z77, Z75 and H77).

Q - What's the concern and need for Native USB 3.0??

 

[wiyosaya]

Thanks for your informative post.

As I said, USB 3.0 is not a deal breaker for me. However, I have a disk imaging program and a USB 3.0 hard drive case from SIIG with a SATA 300 hard-drive. Using the imaging program with this disk/case combo on the same system from a USB 2.0 port and a USB 3.0 port reveals that the USB 3.0 port is about 4 - 5 times faster than the USB 2.0 port.

As for the LGA 2011 setup, I am aware of the discussion on its benefit to gamers. I am not a gamer, and I may or may not use the system that I am planing for games. However, I am planning a purchase of a program, SolidWorks, that benchmarks have shown runs significantly better, about 30% faster, on the lga 2011 relative to lga 1155. In addition, the i7-3820 has shown to be on par in most situations with the i7-3930 or the i7-3960. In some situations, benchmarks show that in the i7-3820 is faster than either of the two black editions while the areas that the black editions best the 3820 are limited.

Also, there are reviews out there that discuss the cost of an i7-3820 system relative to the cost of an i7-2600K or an i7-2700K and show that the costs are similar. While I am not necessarily looking for upgradability of this particular system, it sounds like lga 1155 is on its way out. For my uses and based on all information available to me, the i7-3820 system sounds like the "value" buy.

Finally, as to the degradation issues of Ivy Bridge, it is my understanding that it will be the first time Intel has incorporated their 3D transistor technology. The 3D architecture may affect the way that heat is dissipated in the chip, perhaps making it more uniform; I guess we will have to wait until IVB is released to find out.

Thanks again for your informative posts.

 

[jaquith]

I really don't recommend using USB as a HDD SATA port, there are too many problems even with Native and most decent X79/Z68 MOBO's offer eSATA with SATA3 interface which I already mentioned.

LGA 1155 is not on it's way out, the Sandy Bridge (i7/i5/i3-2XXX) are on their way out, but as mentioned delayed to later this year.

I suspect now that AMD is effectively 'out' that Intel will be much slower to release CPUs and Chipsets, and the old Tick Tock a thing of the past. Tick Tock has cause too many problems in its effort to kill AMD.

Regarding, IB the information out there is limited I did see this article about the i7-3770K @ 7GHz (it's probably toast and hopefully not a hoax) -> http://www.brightsideofnews.com/news/2012/2/24/intel-ivy-bridge-core-i7-3770k-is-capable-of-10025-overclock.aspx which does seem very intriguing. However, as best as I can read it the dummy forgot to delete the ID above the CPU tab, and 2182052 leads you to an unpublished CPU-z validation. The fastest i7-3960X I've seen is still South of 6GHz but at >1.60v.

 

[wiyosaya]

It would be nice if both AMD and Intel stopped trying to beat each other over the heads with their parts and were more focused on quality. Then we, perhaps, would get away from the seemingly ever recurrent hardware problems that each has encountered. For once, IMHO, a focus on quality, rather than flogging each other, would be nice. Each has certainly had their share of hardware problems.

I suspect, though, that Intel's IVB delay (though this is purely my own speculation and therefore, likely coming from the vapors), is as a result of the commercialization of Intel's 3D tech. It is one thing to announce success in new technologies from a research standpoint, and another thing entirely to successfully commercialize such technologies.

That said, I see the delay as a good thing since it implies, to me at least, that Intel may be focusing on releasing a quality product. As with software where releasing a quality product rather than rushing it to market is known to be the most cost-effective approach, I imagine that it is the same for hardware, and probably much more cost effective in the HW than in the SW case.

Also, the OC aspect is, while interesting, something I rarely engage in. I do, however have a Phenom II 965 that I am running on air at 3.8 GHz. I may push whatever LGA 2011 proc I end up getting, however, I undoubtedly will only go with an air-cooled solution rather than water or cryogenics.

While your point is taken on the eSATA approach, the USB HD is a solution that must work with a number of different systems all of which do not have eSATA capability. That said, it is a little used one. I presented it more as a personal experience with the differences in USB 2.0 and 3.0.

 

[jaquith]

I have little doubt that any new litho can for a variety of reasons cause a delay, but my more cynical nature makes me think Intel wants to get more money out of their SB (LGA 1155) before releasing the IB CPUs. We, the 'consumers' are the real driving force and we only understand in our primitive nature 'eat or be eaten.' However, it looks like we will be 'eating' one (1) flavor. I agree it's bad that AMD didn't have success with the Bulldozer, it's really bad for innovation.

Only a microscopic percentage of folks need anything more than the cheapest CPU produced by either Intel or AMD which means there's many millions of fools out there.

As far as cooling, it depends on which CPU. If LGA 2011 and a 6-core then I recommend the Corsair H100 for a variety of reasons: 1. Cooling, 2. Access to the DIMMs, 3. Looks (but to a lesser degree). Disassembling a large HSF to get to a stick is a PITA and they're a dust magnet.

Rehashing the 'Portable Drive' topic, you implied portable, then clearly it's not a 'boot' drive and USB 3.0 can work, and it does not need to be 'Native.'

 

[wiyosaya]

Opps. I didn't mean to imply the USB drive would be a boot drive. I will be putting the boot drive on a SATA III connection.

I doubt I will go for a 6-core as the value of such appears diminished in comparison to the 3820 at half the cost of the 3930. I am a value buyer, and I won't depart from that as performance to value is, IMHO, diminished when moving from the 4 to 6 core 2011 variants. One might say that the 6 core 2011 variants target a percentage of the percentage of users to whom the features of a 2011 system are valuable.

As far as cooling goes, while liquid has its advantages for some audiences particularly the OC audience, if I overclock it will not be to a point where it cannot be done on air. This build will be a system that I can depend on lasting for on the order of 4-years, and as such, I intend to be "conservative" in how far I push it. Still, 3820s have been "pushed" on air to 4.7 GHz, but I doubt that I will push it even that far. IIRC, to push a 3820 to 4.7 GHz, one needs to adjust voltage, and adjusting voltage is something I avoid due to longevity issues.

As far as air cooling goes, I am fan of Thermalright's heat sinks. Typically, Thermalright's heat sinks can be oriented either parallel to the back of the case, or perpendicular to the back of the case. This typically allows for easy access to memory since it is easy to position the HS so that memory access is easy. Your comments raise a good point, and I will likely send an e-mail to Thermalright prior to buying to ensure that orientation and memory interference will not be a concern.

 

[jaquith]

There are so many factors to selecting the 'right CPU' and still not knowing what all your uses 'are' it's impossible to recommend: SB, SB-E, IB, IB-E, (SB) Xeon, (SB) Xeon(s), etc. There are plenty of case studies to compare cost to time analysis.

However, every bench I recall with SolidWorks showed a 30%+ gain with the 6-core i7-3960X/3930K SB-E and the 4-core i7-3820 was slightly ahead of the LGA 1155 SB i7-2600K. The ONLY way for the e.g. i7-3820 to be anywhere on or near to 'PAR' is to OC the i7-3820 >4.7GHz.

The Thermalright runs at least 5C~8C hotter than the H100 at Stock and under Load. Further, the X79's VRM's get toasty and again the H100 don't obstruct case air flow. The setup I choose takes into account all this and was well planned out ahead of time. I by all means am not concerned about or even think of the H100 as anything more than a glorified HSF. It's well purposed and that's it.

MOBOs: In theory I like the Sabertooth X79 because: 1. 5 year warranty, 2. Active cooling for both the X79 Chipset and VRM. Proven, if needed, solid SB-E OC'ing MOBO.

 

[wiyosaya]

jaquith,

I appreciate your willingness to help, and your comments about performance are noted. I won't be building for a while, so I am not looking for recommendations at the present time. You answered my original question. Perhaps by the time I am ready to build, I will be looking for further recommendations. If so, I will be sure to seek the forum's help at that time.

Thanks for your help.

 

[wiyosaya]

If you are curious, though, please note this http://www.techspot.com/review/492-intel-core-i7-3820/page4.html

Perhaps the difference is in what Tom's tested - photoworks rendering vs the other site above. Rendering is only one small aspect of SW. Things like FEA, which SW does, is something that should inherently benefit from the larger memory bandwidth. In my opinion, FEA is much more important than rendering when considering SW performance. Sometimes, FEA problems can be extremely large, i.e., they use large amounts of memory, and I suspect that due to this, the 2011 platform will outperform other current platforms for large FEA models.

I mean no disrespect - as it does sound like you know what you are talking about. However, I think we have some differences in our target applications, and what we find important in our builds.

That said, if I make a mistake and get a build I am not satisfied with, I will find only one person to "blame" - myself.

 

[jaquith]

You said SolidWorks, and these are the benchmarks that I seen, as follows:

Note the Overclocks

ref - http://www.tomshardware.com/reviews/core-i7-3930k-3820-test-benchmark,3090-7.html

ref - http://www.tomshardware.com/reviews/core-i7-3960x-x79-sandy-bridge-e,3071-9.html

 

[wiyosaya]

Yes, I saw those, too.

3820, OC is 24% faster, by those numbers, than 2600K, and I agree that the non-oc test indicates that the 3820 is only marginally faster than the 2600.

Yet, the test that Tom's performed is only for Photoview, and Photoview is only a small part of solidworks. What photoview does is take a solidworks model and generate a photo-realistic scene from that model. It is only a minor part of solidworks that is used when you want to make a photo-realistic presentation for a model and does not indicate solidworks performance in general.

The Specviewperf solidworks test from the link I posted tests a larger subset of solidworks that, IMHO, is a better indicator of the performance in common solidworks functions. Non-OC, the Specviewperf SW test at the link I posted indicates the 3820 is 36-percent faster than the 2600. Trust me on this. The specviewperf sw test is a better measure of SolidWorks performance - at least as I see it.

 

[jaquith]

I claim to be no expert with running SolidWorks, my expertise is in enterprise SQL/PHP and even a single 6-core SB-E isn't even close enough to being usable for anything more than small record batches. You render what you render and ditto with my form of 'rendering'.

So if an i7-2600K or i7-3820 can do the work in the time-frame required then that's good enough for me. The folks I know that 3D render require farms or MP Xeon for small jobs with 3~4 GPUs.

 

[nikorr]

This topic has been closed by Nikorr