I never had any doubts in Ivy Bridge supporting PCI-Express 3.0. I am somewhat surprised that Sandy Bridge-E will not in its current iteration.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Mac Pro "Sandy Bridge-E" Performance Analysis
- Thread starter Amethyst
- Start date
- Sort by reaction score
It likely will at the time of release.
Currently, the current implementations haven't yet been approved by PCISIG.org, which is what deconstruct60 has been trying to explain (must get this organization's approval before they can claim "PCIe 3.0 Compliant" as there requirements that have to be proven before product approval is granted (licensing fees involved to get this approval as well = PCISIG's incentive to grant it). As there hasn't been a lot of PCIe 3.0 compliant devices to test with, it's taking longer to complete this process.
But it's in everyone's (PCISIG's + members) interest that this process is completed by the launch date due to the financial implications (helps generate sales for various vendors, and generates income for PCISIG from the member fees).
Now assuming the Nov. 15 date is correct, then they've ~ 2 months to get this completed (it may come down to the wire as this is a new standard). Worst case, approval will occur soon after release, and the product and specifications pages will be updated to reflect they are compliant.
BTW, Intel has a member on PCISIG's board (Ramin Nashati).
No, Intel is saying next year they can be an Ivy bridge and it will probably support 3.0. It is not a real product being marketed as something ready to sell. You have probably clipped off the Intel slide that has the standard boilerplate about these being "forward looking statements that may/may not turn into real products". There is a difference between technology roadmap presentations and people actively marketing stuff to sell. For the latter folks will say just about anything to get folks to buy more product, so need to put those folks on a short leash least they go wild.
From the xbit story
"... And these were not just some dummy boards, but the actual operational products. ... "
That should be why folks are being conservative in their statements. It is a real product, practically ready to sell but still hasn't passed final official v3.0 tests .... because nobody has because the tests haven't been voted on yet.
My impression is the the story is about information collected from the trade show floor. That is not somewhere that generally every sales pitch starts off with standard disclaimers. More like "our stuff is hot, you should hype it." is the standard prefix.
6-8 months ago people had Intel slides that tagged Sandy Brige E with PCI-e 3.0 because it far from shipping. Right now though there is a peculiar log-jam because v3.0 was put on a predetermined timeline back in June '10 to not have final set of tests certified until Fall '11 (technically it is still not Fall yet). So for a while while marketing real products they have to engage in the tap-dancing. Usually, the standard passes and then everyone drags their feet implementing it. PCI-e v3.0 has run into numerous delays along the way and is just now finally wrapping things up. I think they were being rather conservative ( final standard in Nov '10 and final test certification almost 12 months later gives them lots of time to try out several incremental solutions so have something so have very solid release. )
There are also two things that are going on. One this "PCI-e v3.0 capable" stuff is getting out of control. In one of these threads someone posted some MSI versus some other vendor where they are posting testing of pre-release testing ivy bridge test processors on some other vendor's board that shows how they fail PCI-e v3.0. That's ridiculous. They aren't even PCI-e org official org tests that can be used for public qualifications. It is also indicative of the "he said , said she " back and forth bickering that having a standard's org testing and validation is suppose to avoid. Unless the folks running the PCI-e org are completely asleep at the wheel, there should have been a broad communication sent out to put the market dogs and sales wonks back on their chains for PCI-e v3.0 validation proclamations.
The other factor I suspect at play is that some conservative system vendors may be looking for air cover to ship v2.0 only certified board. The economy is bad and they have gone into deep risk adverse mode. Since there would be short window between v3.0 tests going official and launch they have committed to not doing it. If enough talk goes out that v3.0 is just generally optional then they can get away with that. There will be a slew of conservative boards. [ think $8-10K server boxes where folks do week or month long evaluations before buying. Those folks usually have at least one Nervous Nelly on the buying committee. System vendors will actually sell more to those folk but kneecapping the system. ]
So yeah. There will probably be a mix and waiting to evaluate till actually shipping will give better information.
The other possibility that it really is 40 lanes on the chip, but have to wire 4x to the X79 to support the additional bandwidth for what were the "extra" 6.0Gbps SATA lines. That's exactly what the chipset was projected to prune off for the SAS support: (from older register story http://www.theregister.co.uk/2011/05/30/intel_sandy_bridge_xeon_platforms/print.html )
So that is really a "pragmatically" 36 because the other 4 are pre-committed to the chipset. Counting them would almost be like counting the DMI links. they aren't really available to slots (or board embedded features ) at all. So it is better to reset the expectation down to 36.
At least that would represent someone not screwing up the basic design parameters for the CPU PCI-e switch in the beginning. I given Intel has already screwed up SATA III already it would not be surprising if the "extended" SATA/SAS unit on the chipset was also screwed up. It probably took copy and paste elements from the screwed up Cougar point chips too early in the design process. Scaling it back to just SATA and only 4 SATA III lanes likely would be fixable.
If just turning off SAS support and routing a more limited amount of SATA III traffic through what was oriented to SAS, but keeping the x4 leach off the processor at v2.0 speeds (for worse case and/or ultra conservative boards) got them back to four stable 6Gbps SATA lanes then they probably did. (likely means Platsburg -B is dead or going to be seriously delayed. I'bd bet dead. And that -D and -T will have to be down marketed if released as SATA only. )
Again, I don't think it really matters for MacPro because Apple likely was lined up for Patsburg-A which doesn't have the active links for that anyway. If that just means two 6Gbps ports (beacuse the 4 lane SCU unit is switch off because doesn't have leach links) they'll roll with that. [ and it more closely matches to the 2 6Gbps Intel board in xbit story as I said previously]. There are better ways for Apple to spend those extra 4x lanes on the single package MacPro. So I still think the E5's will come with 40 because 4 are not committed by default in all designs because the base chipset can vary more.
From variety of core configuration on "SNB-E" mac pro. (4,6,8,12,16)
Do you guys think we will give a lower start price?
Ps.
1,1 core configuration 2x2 starting at $2199 (2xdual2.0)
2,1 core configuration 2x2 , 2x4 starting at $2199 (2xdual2.0)
3,1 core configuration 1x4 , 2x4 starting at $2299 (quad 2.8)
4,1 core configuration 1x4 , 2x4 starting at $2499 (quad 2.66)
5,1 core configuration 1x4 , 1x6 , 2x4 , 2x6 starting at $2499 (quad 2.8)
Do you guys think we will give a lower start price?
Ps.
1,1 core configuration 2x2 starting at $2199 (2xdual2.0)
2,1 core configuration 2x2 , 2x4 starting at $2199 (2xdual2.0)
3,1 core configuration 1x4 , 2x4 starting at $2299 (quad 2.8)
4,1 core configuration 1x4 , 2x4 starting at $2499 (quad 2.66)
5,1 core configuration 1x4 , 1x6 , 2x4 , 2x6 starting at $2499 (quad 2.8)
Well, 3.0 apparently wasn't working in a particular engineering sample, although it was unclear what that will mean for the released product. It is easier the other way -- if the roadmap shows a feature, and then it is present in the engineering sample, you would assume that things are on track. But, what does the negative prove? Be careful I guess. I'm hoping they get it working for the released product - I think PCIe 3.0, the improved memory bandwidth, and the extra PCIe lanes, will all make for more balanced systems with Sandy Bridge than the current generation (e.g. Westmere-based) of systems.
No. LGA 2011 UP Xeons are the same price as the LGA 1366 UP Xeons, why would the price change from Apple?
The pricing is not out though, so I wouldn't be dead-set yet. Although if the prices are changing, then the direction is most likely up. And I wouldn't be surprised if Apple raised the prices even if the component prices remained the same.
The pricing was leaked as being the same a while ago. Apple are Apple and all, but $2,499 obviously works for them. Can't see them going to $2,599+ without some significant changes which don't seem to be expected.
A rumor is always a rumor. Without official statement by Intel, or even something more than just "our sources say", I would take everything with a grain of salt. I know the source (VR-Zone) is fairly reliable when it comes to stuff like this but it still doesn't meant it's 100% accurate.
The price of 8-core went up by 200$, even though there was no significant change (133MHz faster CPU which costs $14 more each). Other models received the same upgrades (faster GPU, WiFi etc) without a price increase.
I'm not saying the price will go up, but then again, I wouldn't be surprised if it did.
I agree. A rumor is a rumor. And, sometimes rumors from some companies are part of test marketing -- float a price out there and see how many people cry foul (price might be too high), or, are happy about the price (probably need to raise the price a little). These products are not like bread and rice -- demand is very elastic.
I wouldn't object to a small Mac Pro price increase if they made a graphics/ I/O monster out of the dual-CPU model. OTOH, if they crimp graphics and I/O, it wouldn't be worth it.
Last edited:
VR-Zone wasn't the only source. Dig deeper and there were prices from retail sources and OEMs; and they were from multiple regions confirming same pricing as the current rates.
Of course Apple can increase the price as they have before. I just don't think you can say Apple will do so based on processor pricing. Not when nearly all the evidence points to the pricing being maintained and with the huge margins they already have. I would expect a $2,499 Mac Pro whether the 1620 is $294 or $349.
My price ideas is all about the different between the model.
If apple choose all SNB-E configuration to made a mac pro.
We will have a variety of mac pro configuration.
4 Core, 6 Core, Dual 4 Core, Dual 6 Core and Dual 8 Core.
What a slot that apple can put the product? if they have 5 different configuration.
Let's debate!!
ps. I still believed. It better to kill and SP model. and start Mac Pro at $2799+
If apple choose all SNB-E configuration to made a mac pro.
We will have a variety of mac pro configuration.
4 Core, 6 Core, Dual 4 Core, Dual 6 Core and Dual 8 Core.
What a slot that apple can put the product? if they have 5 different configuration.
Let's debate!!
ps. I still believed. It better to kill and SP model. and start Mac Pro at $2799+
Care to link then?
You were the first one to indicate that the Apple's pricing depends on CPU prices (or at least that was the only point you raised in that post):
Umbongo said:
I wasn't trying to suggest that Apple will change the pricing. I agree that $2499 is what we will most likely see, and it's already quite a lot (compared to e.g. 2006 MP).
Apple system price is NOT based on processor PRICE.
The examples
Model -> iMac 2010 -> iMac 2011 Processor Price
21" $1199 -> i3 540 $133 -> i5 2400s $195
21" $1499 -> i3 550 $138 -> i5 2500s $216
27" $1699 -> i3 550 $138 -> i5 2500s $216
27" $1999 -> i5 760 $205 -> i5 2400 $184
Model -> MBP 2010 -> MBP 2011 Processor Price
15" $2199 -> i5 620M $322 -> i7 2720QM $378
The examples
Model -> iMac 2010 -> iMac 2011 Processor Price
21" $1199 -> i3 540 $133 -> i5 2400s $195
21" $1499 -> i3 550 $138 -> i5 2500s $216
27" $1699 -> i3 550 $138 -> i5 2500s $216
27" $1999 -> i5 760 $205 -> i5 2400 $184
Model -> MBP 2010 -> MBP 2011 Processor Price
15" $2199 -> i5 620M $322 -> i7 2720QM $378
Last edited:
Don't really feel like digging through forums and RSS feeds, if I come back to them I will post them.
I took Amethyst's post to be related to price, it doesn't make much sense to me otherwise.
We really do not have an explanation right now. I have also read comments on the lack of 3.0 hardware in order to get the certification or that said lack of hardware prevents getting final verification of the circuitry itself, requiring a respin. It works just fine under 2.0 modes though.
Ivy Bridge appears to still be far enough out and AMD should be offering PCI-Express 3.0 on its Bulldozer (AM3+) successor platform. AMD is also moving its server and derivative enthusiast desktop platform to a CPU + FCH system as well.
PCI-Express 3.0 does feel quite guaranteed for 2012.
I'm not saying it's a guarantee, but there is a strong financial incentive to have the compliance/certification issues sorted by the time the CPU's actually ship (much more likely for IB due to the additional time available, but it needs to be there for SB as well, or some users will hold off purchasing systems built from them).
Oh it is my feeling that 2012 is when PCI-Express 3.0 will show up. I did not make it a complete guarantee though.
The Xeon SP E5-1600 series prices showed up today. Starting at $294 for 4 cores, 8 threads. It is not very surprising though compared to the current entry offerings from Bloomfield.
I'm interested to see how much Apple will charge for the 1650. One would expect the same $400 upgrade price for that tier, but I wonder if they will push it to $500 as I can imagine a lot going for it as an upgrade and I doubt that extra $100 would be too off putting compared $1,200-$1,300 for a 1660.
No thanks.
I don't want a 2.4Ghz DP 8 core vs a 3.2Ghz 6-core.
SP CPUs are as posted before:
4 core @ 3.4
6 core @ 3.2 and 3.3.
DP CPUs will probably be (Given the DP always uses 95W TDP CPUs):
4(8) cores @ 2.4 and maybe 3.3 (but that has a 130W TDP
)6(12) cores @ 2.5Ghz
8(16) cores @ 2 or 2.2Ghz
In all honesty I don't think they will even bother with the 4(8) core, it will be outperformed by the 6 core chips as the W3680 3.33Ghz 6 core outperforms the 2.4Ghz 8 core 2010 MP.
I have absolutely no intention of buying a Mac Pro for 8/12 cores at 2.4Ghz/2.5Ghz.
I'll be having the 6 core at 3.2 thank you very much.
So I recon the Mac Pro will be (as long as they stick to 95W in the DP machines):
4 @ 3.4
6 @ 3.2
6 @ 3.3
12 @ 2.5
16 @ 2
16 @ 2.2
Last edited:
I don't want a 2.4Ghz DP 8 core vs a 3.2Ghz 6-core.
SP CPUs are as posted before:
4 core @ 3.4
6 core @ 3.2 and 3.3.
DP CPUs will probably be (Given the DP always uses 95W TDP CPUs):
4(8) cores @ 2.4 and maybe 3.3 (but that has a 130W TDP
6(12) cores @ 2.5Ghz
8(16) cores @ 2 or 2.2Ghz
In all honesty I don't think they will even bother with the 4(8) core, it will be outperformed by the 6 core chips as the W3680 3.33Ghz 6 core outperforms the 2.4Ghz 8 core 2010 MP.
I have absolutely no intention of buying a Mac Pro for 8/12 cores at 2.4Ghz/2.5Ghz.
No, they shouldn't kill the SP Mac Pro. With virtual 12-core machines now the SP Mac Pros will be quite a beast. Especially running at 3.3 Ghz, and turbo-ing to 3.6 or so for jobs that are not as multi-threaded.
However, I'd still like to see a DP 4 Core option. Some of us have work that is not as limited by the processor and rather RAM limited. More cores can always help, but it reaches diminishing returns. It gets to the point where you might rather spend the marginal $1-2K and get 96-128GB of RAM rather than have a SP with a 3.3 or 3.6 clock speed but be limited (by DIMM slots or simply cash) to less RAM.
Also, might it be possible that these E 26xx even though they are running with a lower base clock rate, make up more of the gap in clock rate with 16xx using turbo-boost 2.0?
However, the 4 core 2.4 isn't even multithreaded.... So yeah, if they can't get 4 core 3.3Ghz 130W in the Mac Pro, I don't see the point of the 4 core option. Might as well go back to the DP 4 core Westmeres if you really want it.
No thanks.
I don't want a 2.4Ghz DP 8 core vs a 3.2Ghz 6-core.
SP CPUs are as posted before:
4 core @ 3.4
6 core @ 3.2 and 3.3.
DP CPUs will probably be (Given the DP always uses 95W TDP CPUs):
4(8) cores @ 2.4 and maybe 3.3 (but that has a 130W TDP
6(12) cores @ 2.5Ghz
8(16) cores @ 2 or 2.2Ghz
In all honesty I don't think they will even bother with the 4(8) core, it will be outperformed by the 6 core chips as the W3680 3.33Ghz 6 core outperforms the 2.4Ghz 8 core 2010 MP.
I have absolutely no intention of buying a Mac Pro for 8/12 cores at 2.4Ghz/2.5Ghz.
I'll be having the 6 core at 3.2 thank you very much.
So I recon the Mac Pro will be (as long as they stick to 95W in the DP machines):
4 @ 3.4
6 @ 3.2
6 @ 3.3
12 @ 2.5
16 @ 2
16 @ 2.2
If this will be the case regarding 12/16 core Pros, I don't see them outperforming the current 12c 2.93?!
Don't get fooled by the frequency. 16-core offers 33% more cores and the clock for clock improvement should also be around 15-20%.
A very rough estimation would be:
12*2.93GHz = 35.16GHz
16*2.2GHz*1.15 = 40.48GHz (~15% faster)
Or one option is that Apple uses CPUs with TDP of 115W, allowing them to go up to 2.6GHz. That would yield 47.84GHz with a similar calculation, 36% more than the current 12-core.
Don't take these figures too seriously though, performance isn't usually as simple as f*n.
Urggggggghhhhh ! Can't wait to see an all-new SNB-E Mac pro!!!!!!
----------
----------
With this conf. Top of the perch mac pro may be slow than it's entry level in some situation.