I have a hard time thinking Apple would go with anything other than the 1600s. The 2600s are wacky expensive for similar core count/clock speed.
Soooo......
Let the throttling threads commence.
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.
Haswell-E doesn't have much over Ivy
- Thread starter VirtualRain
- Start date
- Sort by reaction score
I have a hard time thinking Apple would go with anything other than the 1600s. The 2600s are wacky expensive for similar core count/clock speed.
Soooo......
Let the throttling threads commence.
s;maxwell;haswell;g ??
Maxwell is the current/immediate future architecture from Nvidia.
Haswell is the current/immediate future architecture from Intel.
Where "current/immediate future" means that some lower end chips using the architecture are shipping, but not yet the full range.
Maxwell is the current/immediate future architecture from Nvidia.
Haswell is the current/immediate future architecture from Intel.
Where "current/immediate future" means that some lower end chips using the architecture are shipping, but not yet the full range.
With regards to the power envelope, there were at least a couple of threads late last year with many folks claiming there wasn't sufficient power in the PSU to power a 130W CPU, a pair of workstation GPUs and everything else. I recall trying to convince people that it would be possible with selective binning. We all know that even with selective binning and under clocking there's not a lot of power or thermal headroom in the design. But maybe you're right and it's a total non issue. (Although apperently not only Haswell is running hot but GCN 1.1)
My agenda stems from originally waiting for a Haswell Mac Mini which appears to not be forthcoming, and while the most likely reason is that it's simply a low priority product for Apple (perhaps not much unlike the Mac Pro), it made me start looking into why Apple might have skipped over Haswell... And what I found was reasonable evidence for Apple to skip it and wait for Broadwell (especially beard on its original schedule).
I'm not convinced the Mac Pro is not getting an update, I'm only trying to look into why it potentially might not.
If it does get updated, great.
Maybe I'm looking for smoke where there is no fire, but I think it's an interesting topic for discussion. But others might disagree or think I'm smoking drugs.
This attitude hasn't been coming through your other posts, you seem to take every opportunity to dismiss the idea of a Haswell upgrade.
As I said in my "there are two issues confused as one" post, IMHO Haswell is an important upgrade for Apple to show some commitment to the MP6,1 platform - but not a must-have upgrade for most people who already have an Ivy Bridge MP6,1.
A Senior Moment Is Inconsistent with the Niods (a/k/a paranoids)
Thanks for catching the error. I drank from the wrong well.
----------
I don't disagree with you - it's an interesting topic. I've found it to be a very good thought and learning motivator. Just don't inhale too deeply and share.
Thanks for catching the error. I drank from the wrong well.
----------
... .
Maybe I'm looking for smoke where there is no fire, but I think it's an interesting topic for discussion. But others might disagree or think I'm smoking drugs.
I don't disagree with you - it's an interesting topic. I've found it to be a very good thought and learning motivator. Just don't inhale too deeply and share.
Last edited:
Agreed... that has been intentional to create some discussion and debate and my apologies to anyone who may have felt I stubbornly dismissed their views without much consideration - all the counter posts have been enlightening and helpful if not convincing.
I think we'll see an update, but not as soon as we might think.
In the next rev they will need to sort out the USB, come out with a new video card and implement DDR4 ECC.
So, not in 2014. Q2/Summer 2015 feels a little more likely. If they skipped Haswell altogether I would not be shocked. The nMP has some upgrade issues in terms of thermal. They can't just throw a 140w CPU and an AMD R9 based video card in there and just go with it.
The Mac Pro is still a niche Apple product.
The CPU alone is not really that compelling.
I have a nMP 6c, a 4930k and a 3770k and I'm not thinking about replacing any of the computers this year.
Most of us know that we could build a computer for half the price of the nMP that would be faster. I did with my 4930k which runs all day at 4GHz sans ECC ram. But we choose the nMP anyhow. Maybe the form factor, maybe the style, maybe OSX. Apple is in no hurry to keep us on the bleeding edge. We'll buy what they make when they make it.
In the next rev they will need to sort out the USB, come out with a new video card and implement DDR4 ECC.
So, not in 2014. Q2/Summer 2015 feels a little more likely. If they skipped Haswell altogether I would not be shocked. The nMP has some upgrade issues in terms of thermal. They can't just throw a 140w CPU and an AMD R9 based video card in there and just go with it.
The Mac Pro is still a niche Apple product.
The CPU alone is not really that compelling.
I have a nMP 6c, a 4930k and a 3770k and I'm not thinking about replacing any of the computers this year.
Most of us know that we could build a computer for half the price of the nMP that would be faster. I did with my 4930k which runs all day at 4GHz sans ECC ram. But we choose the nMP anyhow. Maybe the form factor, maybe the style, maybe OSX. Apple is in no hurry to keep us on the bleeding edge. We'll buy what they make when they make it.
Last edited:
Over the last year this misconception popped up in the other Macrumor forums on the other Mac devices. You are just effectively recycling it here. For the Haswell iteration Intel has moved some powerconditioning/management functions onto the CPU package.
"New Features
.....
Fully integrated voltage regulator (FIVR) ... "
http://en.wikipedia.org/wiki/Haswell_(microarchitecture)#New_features
Yet another example of not x86 non-core features that you seem intent to sweep under the rug.
The 2013 MBP, MBA all got Haswell CPU packages with incrementally higher TDP and went on to deliver higher battery life and lower overall system TDP. Their PCH chipset is lower. This power management shift is a ballon squeeze. It is already there in the current Mac Pro. Moving it to another location on the board (inside the CPU package) does nothing to increase overall system TDP.
Broadwell isn't going to change the power management location ballon squeeze design ( because targeted to compatible motherboards). I think I read something that Intel was going to drop (or adjust) this design move at a future iteration. Even then, overall system TDP impact is not an issue.
Xeon E5 v4 is very likely going to buy jack squat over Xeon E5 v3 in terms of the TDP envelope.
Actually Haswell's much improved dynamic power management fits much better with Apple's limited cap 450W power supply issue. The "race to sleep" is part of how they manage to do pretty good with the limitation. E5 v4 will be incrementally better, but no where near revolutionary (worth skipping a release) better.
Why would Apple spend all that money redesigning the Mac Pro if they weren't going to keep it updated and competitive against other workstations? We know that Dell is going to have systems available with Haswell-EP on September 9th, the same day Intel plans to hold a presser revealing them. OEMs have had them for awhile, likely including Apple.
If they were just going to leave the Mac Pro on a 2-3 year update cycle, they wouldn't have spent so much money redesigning/remarketing the thing. The creative industry (meaning video and audio production) is still heavily reliant upon OS X, and upon computers that can run more than a 4c/8 thread processor.
I still say that, barring any complications with the new CPUs/GPUs, we'll at least hear of an update in the pipeline around Oct/Nov/Dec, if not see one by the end of the year. They wouldn't have gone through all the hassle of what is the nMP if they didn't intend to keep the product line supported and updated. Just like I feel whatever October event is happening around Yosemite will come with an updated/redesigned Mac Mini as well.
Also, those of you that think Apple wouldn't have planned for a 10w TDP increase somewhere in the system, or couldn't somehow work around it, and that they are already maxing out the thermal properties of the current system....wow. Considering the voltage regulator is now built into the CPU core, adding 5w to it, that leaves 5w of an actual increase inside the system. The new GPUs will be clocked as low as they need to target the TDP they need. I really don't see or understand the problem here...
If they were just going to leave the Mac Pro on a 2-3 year update cycle, they wouldn't have spent so much money redesigning/remarketing the thing. The creative industry (meaning video and audio production) is still heavily reliant upon OS X, and upon computers that can run more than a 4c/8 thread processor.
I still say that, barring any complications with the new CPUs/GPUs, we'll at least hear of an update in the pipeline around Oct/Nov/Dec, if not see one by the end of the year. They wouldn't have gone through all the hassle of what is the nMP if they didn't intend to keep the product line supported and updated. Just like I feel whatever October event is happening around Yosemite will come with an updated/redesigned Mac Mini as well.
Also, those of you that think Apple wouldn't have planned for a 10w TDP increase somewhere in the system, or couldn't somehow work around it, and that they are already maxing out the thermal properties of the current system....wow. Considering the voltage regulator is now built into the CPU core, adding 5w to it, that leaves 5w of an actual increase inside the system. The new GPUs will be clocked as low as they need to target the TDP they need. I really don't see or understand the problem here...
sort out the USB ? The new C610 series chipset puts USB 3.0 right out of the box. Apple already has a USB 3.0 discrete controller working if somehow they didn't like that one.
If somehow alluding to USB 3.1 then you don't remember how long Apple delayed jumping on the USB 3.0 bandwagon? Apple was a quite slow adopter (like years behind). Apple took the very conservative route of waiting until USB 3.1 made it into the Intel PCH/chipset. USB 3.1 isn't coming to the PCH any time very soon in the mainstream. In the chipset workstation space, ..... well USB 3.0 is just arriving this year. Again no time soon.
USB 3.1 is in conflict with Apple's quick transition to all PCI-e SSDs. 3.1 has additional PCIe lane consumption pressure than 3.0 does ( x2 from x1 ).
Apple really doesn't need to implement DDR4. Apple is going to buy DDR4 devices (CPU that is capable and DIMMs that implement it). All Apple has to do is put the traces on the board from the DDR4 DIMM sockets to the CPU package socket so there is tolerable line noise. It is work, but not rocket science work.
The only way that specific level of work slides into Summer 2015 is if there is has been nobody working on it for most of 2014. It isn't difficultly.... it is just plain simple nobody working on it. That is less than a years worth of work for a competent design team. Even allowing for a long beta/verification process it is less than a year.
I would not be completely shocked if Apple had pulled practically all R&D resources away from the Mac Pro. It isn't like they haven't done it before. It is a bonehead move but then hanging your EU customers out to dry for almost a year is a bonehead move too.
This assertion makes very little sense. The E5 Xeon v3 and AMD GCN 1.1 components are actually substantively better at dynamic power control then the components in the Mac Pro now. If dealing with a limited fixed amount of power envelope is an issue those components are better. Skipping them is goofy. Apple isn't going to put in an overclocked R9 and/or E5 v3. So the max, 'barn burner' AMD R9 limits are just largely FUD.
The current AMD GCN 1.0 cards are an OpenCL dead end. Apparently there are a small set of features missing to fully support OpenCL 2.0. The Mac Pro is in part a 'bet the farm" on OpenCL play. Languising for another year (or so) on something that inhibits the long term OpenCL adoption is beyond goofy.
Again doesn't make any sense because more than just the CPU package is available. This is a tock so the chipset will increment. AMD and Nvidia have already announced in process of releasing new GPUs. Only is the alternative universe of this thread is the CPU only changes happening right now.
Apple isn't trying to sell any updated Mac Pro to anyone who has just purchased on in the last 9 months. They simply are not. This thread really shouldn't be about excuses as to why someone should replace a less than 14 month old Mac Pro. What you have now is probably way faster than what you had before you upgraded. That is good enough.
If have a "must buy newest shiny" problem then arm flapping about how Hawell (and new GPUs) don't offering anything is just plain outright smoke. You are in denial.
Perhaps Apple is trying to goose demand into every other year. Makes it easier for them to create demand bubbles and hype around launches. At least for those who stick around. It is a probably a pretty effective way to kill off the Mac Pro customer base over the long term though. There are only so many "have to buy the newest shiny" customers out there. Folks who need tools with regular updates are going to go elsewhere.
But there is no technologic basis there. That is just Apple playing some Scrooge McDuck game probably thought up by someone with an M.B.A. degree. There is no engineering there at all.
The good news (or bad news for some?) is with Broadwell delayed there really isn't anywhere else to move the Mac Pro resources right now. As far as we know. There was a new Macbook line rumored, but that's likely delayed with Broadwell.
I don't see the point in skipping Haswell since its going to be the same socket as Broadwell. And, if there are any, Apple is going to have figure all the DDR4/new chipset features out eventually regardless. So why let competitors market Haswell stuff left and right, while you're stuck on Ivy Bridge? Seems really dumb.
Right. What I meant is that new Macbook Pros aren't going to appear this year with Broadwell, so there isn't any other team to move the Mac Pro engineers to. There aren't any other projects going on for the second half of 2014.
Again, as with many new features in Haswell, the FIVR is all about mobile computing.
In Anand's review they noted about a 20% improvement in system power draw at idle, which amounts to great battery life under light workloads. However, under full load, power consumption was 13% higher than Ivy.
Of course the other benefit of the FIVR is decreased component count and main board real-estate, which is a nice benefit if you're putting it into a MacBook Air, but less so in a Mac Pro.
In the last 24hrs I've also read reports of Haswell desktop CPUs running 15-deg hotter than Ivy Bridge equivalents and also having more aggressive thermal throttling than Ivy Bridge all of which may not bode well for use in a system like the Mac Mini or Mac Pro where cooling system performance is designed for low noise vs. maximum cooling. I guess at the very least, a Haswell 7,1 nMP would likely have higher fan speeds under load than the 6,1.
As for the E5 v4, that will be a die shrink that will offer a lot of extra TDP headroom that can be used (or not) in a variety of different ways.
Last edited:
VirtualRain's right. I believe it wouldn't get updated as well (not soon but maybe in 2 years). They have already showed a big commitment(MP5.1->6.1) by shrinking its size, keeping it way quieter, using Ivy Bridge processors (If I'm not mistaken, Apple skipped Sandy Bridge for mac pros) and 2 Dual GPUs. And also like you say, it's not a must-have upgrade for most people who already have an Ivy Bridge, Apple would think the same too.
Yosemite is adding support for the 290X/W9100.
Seems like a no brainer there is a new Mac Pro coming around Yosemite's release time, and that there is one in development. The addition of drivers for a W9100 class GPU should be a giveaway.
They aren't writing those drivers for the Mini
Though it is a pity that the components have to fit in the existing TDP instead of choosing components and then supplying them with needed power and cooling.
I think the only problem with W9100, as with Haswell, is that they both seem to increase power draw and heat. This is not unexpected since neither technology has benefitted from any die-shrink improvements.
As noted above, Haswell desktop CPUs have been widely reported to be running 15-deg hotter at load than Ivy and it seems Hawaii is too... Attached is the result from Tom's test of the W9100 that shows it's also running 15-deg. hotter than the W9000.
When you combine two of these GPUs and a Haswell processor together at load, that's a lot of extra heat to dissipate which can only mean one or more of...
1. Higher fan speeds
2. Underclocked GPUs
3. More likely to hit thermal limiting
When you couple this with the negligable performance improvements that Haswell offers, and the modest (but largely untapped) performance improvements that the W9100 offers, it doesn't seem to make for a sensible upgrade.
A Haswell system with dual under-clocked W9100s is not going to benchmark particularly well and it probably won't fare as well in Anand's torture test either. It could be a dud.
W9100 temps from Tom's Review...
Attachments
I think part of the problem i see people making is that Apple is not waiting for Broadwell for the Mac Pro, they're waiting for Broadwell-E/EP and those are not coming any time soon, and they weren't waiting for Haswell-E either, they're going to use Haswell-EP if they do, and considering how much time will probably be between Haswell-EP and Broadwell-EP they better use Haswell-EP.
Granted it won't be a great upgrade for the current mac pro owners unless they need more cores since the top of the line xeon will supposedly have 14 cores.
Now a Haswell-E iMac with 8 cores... now that would be really great.
Granted it won't be a great upgrade for the current mac pro owners unless they need more cores since the top of the line xeon will supposedly have 14 cores.
Now a Haswell-E iMac with 8 cores... now that would be really great.
It was never going to be a must have upgrade over Ivy Bridge-EP, people generally buy workstations to last years. Apple needs to maintain interest in the Mac Pro though. 8-cores just dropped by $800 and the E5-2797 V3 gains 2 cores over its predecessor - making an 8-core Mac Pro $5000 and an 8-core PC workstation $2,500 for similar performance.
Maybe Apple only care about people who would buy anyway or who are only buying because Apple's other products aren't efficient, but good luck growing sales when you show disregard for your user's time.
Apple have nothing to lose by moving to Haswell-EP. This isn't like last time, where sales had died and they didn't know what to do with it. If they don't upgrade by February that says all I need to know about Apple's attitude towards "pros". If they couldn't design a brand new system and manufacturing process and account for guaranteed updates from Intel in that and have to skip one because it wasn't worth it to them and they don't care about users getting the latest hardware, then they aren't worth bothering with. It should be nothing for them to move forward each tick-tock, just as it is for every other workstation vendor.
As for GPUs running hotter, Apple will just clock them at whatever they need. If there was extra heat and the design, or modifcations to it can't handle that then the whole product redesign is laughable.
Last edited:
We know the Mac Pro is power constrained, but do we know if it's hit thermal constraints?
The W9100 part only has a TDP of one watt higher than the W9000, from what I'm reading. It's peak output may be higher, but have we determined the Mac Pro couldn't handle that? Most of the underclocking has usually been traced back to the power supply output, not the heat.
At this point, the addition of W9100 drivers into the Apple builds, specifically starting with Yosemite, doesn't seem accidental. I don't think any other Mac right now is using AMD parts?
I've been holding onto my oMP looking for this update but from what I'm reading here it's not really worth the wait especially if it doesn't happen until sometime in 2015.
I was planning on getting a 6C/D300 combo but my only fear is that the D300 will be another GT120 situation in 2 years but unlike that card I won't be able to upgrade. Sadly, adding the D500's + 6C is just out of my price range.
I was planning on getting a 6C/D300 combo but my only fear is that the D300 will be another GT120 situation in 2 years but unlike that card I won't be able to upgrade. Sadly, adding the D500's + 6C is just out of my price range.
I assume that you're still playing the devil's advocate, and making weak arguments to encourage the discussion
.1. Higher fan speeds
Problem's been solved since forever (at least on my Dell laptops and HP servers).
Have a control panel to select:
Have a control panel to select:
- High Performance - system performance will only be reduced if necessary to stay within thermal or power envelopes.
- Balanced - system performance will be reduced if necessary to avoid running the fan at full speed.
- Quiet - system performance will be reduced to avoid noticeable fan noise.
2. Underclocked GPUs
See above, controlling CPU and GPU clock speeds is part of "reduced performance".
3. More likely to hit thermal limiting
Most of the reports (published and anecdotal) that I've seen have said that the MP6,1 is very quiet under any load. Even the "power virus" reports that saw throttling due to power supply limits didn't talk about loud fans.
I don't think anyone outside of the windowless labs in Cupertino knows the true thermal limits of the design.
I don't think anyone outside of the windowless labs in Cupertino knows the true thermal limits of the design.
...and a weak argument that you left out:
4. Power supply overload
We're really only talking about a few 10s of watts - who knows how much power those clever blokes in Cupertino can really get out of that power supply form factor. (Some of my HP servers have dual 1200 watt hot-swappable supplies, and they're really quite small. (It is recommended to run them on separate 240volt circuits - but my racks are supplied with 380volt 3phase, so not an issue...))
Last edited: