Sidebar Sidebar
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

VirtualRain

VirtualRain

macrumors 603
Original poster
Aug 1, 2008
6,304
118
Vancouver, BC
Anand posted a review with a number of benchmarks today for Intel's new Haswell-E i7 enthusiast desktop CPUs. Based on what I can see, there's nothing notable in terms of performance improvements in real-world benchmarks although I'd be interested in hearing differing opinions...

The new 5930K should be indicative of a replacement for the nMP 6-core (1650v2) which is comparable to the IB 4930/60 used in their benchmarks.

As you will see, the Haswell 6-core 5930K offers nothing in terms of added performance over the Ivy-Bridge equivalents. Case in point... the new Haswells are right in-between the two Ivy variants...

67025.png


So I wouldn't hold your breath for Apple to refresh the nMP based solely on Haswell... there's just nothing there but added expense. The only thing I can see driving a refresh of the nMP in the next year would be next-gen GPU tech.
 
Last edited:
the new 5860x 8 core is 20% faster than the 6core 4860x...i dont think money is burning through the peoples pockets to go from ivy 6core to haswell 6 core. They are going for the 8 core.

Thunderbolt natively supported. usb 3.0 native.. ddr 4 with 3ghz ram is crazy.
 
Makes sense. The main benefit of Haswell was battery life, which is less of a concern on a desktop which you will always keep plugged in.
 
AidenShaw said:
AVX2

Lets see some benchmarks with AVX2 optimized code before we conclude that E5-v3 is a waste.
Click to expand...

Yep.

The benchmarks are... odd... Sure, clock for clock Haswell isn't too much faster, but the core counts have doubled for the reviewed chips. So sure, a Haswell 4 core and a Ivy 4 core are about the same, but someone who bought a 4 core Ivy is probably buying a 6 core or 8 core Haswell now, with 1.5x or double the performance. Derp derp.

A 12 core Ivy won't hold up to a 16 core Haswell, even if you want to say they're even clock for clock.
 
AidenShaw said:
AVX2

Lets see some benchmarks with AVX2 optimized code before we conclude that E5-v3 is a waste.
Click to expand...



Aiden,

Regarding the AVX2 seasoning, what kinds of functions/applications do you foresee AVX2 being a better ingredient than AVX(1) for OSX users and how much so? Also, does Haswell have any other instruction set advances that you foresee being of significance over those embodied in Ivy Bridge.
 
Last edited:
RoastingPig said:
the new 5860x 8 core is 20% faster than the 6core 4860x...i dont think money is burning through the peoples pockets to go from ivy 6core to haswell 6 core. They are going for the 8 core.

Thunderbolt natively supported. usb 3.0 native.. ddr 4 with 3ghz ram is crazy.
Click to expand...

In the Handbrake benchmark I quoted above, 5960 8-core is only 16% faster than the 6-core Ivy... It's not surprising really because it's base clock is 3.0GHz compared to 3.6GHz on the Hex. The Octo has 33% more cores but 20% lower clocks, hence the small performance gain. It appears to have little to do with architecture. Only in synthetic benchmarks could any difference in architecture be measured and we're talking single digit percentages. In real world bench marks there was absolutely nothing to get excited about.

Native USB 3 would save Apple a few dollars, and might net you a bit of performance in extreme storage setups, but TB is the way to go there anyway with double or triple the throughput potential. And unfortunately, as you can see, DDR4 at 3GHz buys you nothing in terms of performance.

It all sounds great on a spec sheet but real world benchmarks show it's a wash.

As I said above, I wouldn't wait on a Haswell refresh... Clearly Apple skipped Haswell on the Mini, and I won't be surprised at all if they skip it here too - or wait at least until some compelling next-gen GPUs are worthy of refreshing the nMP.
 
VirtualRain said:
As I said above, I wouldn't wait on a Haswell refresh... Clearly Apple skipped Haswell on the Mini, and I won't be surprised at all if they skip it here too - or wait at least until some compelling next-gen GPUs are worthy of refreshing the nMP.
Click to expand...

I would still be extremely surprised if they skipped. AVX2, DDR4, and higher core counts (along with nice GPU upgrades waiting in the wings) make it a good time to upgrade.

If the iWatch is going to be announced this month, that might push the Mac Pro out of any keynote announcement, but I still wouldn't be surprised to see a quiet upgrade between now and January.
 
goMac said:
I would still be extremely surprised if they skipped. AVX2, DDR4, and higher core counts (along with nice GPU upgrades waiting in the wings) make it a good time to upgrade.

If the iWatch is going to be announced this month, that might push the Mac Pro out of any keynote announcement, but I still wouldn't be surprised to see a quiet upgrade between now and January.
Click to expand...

I guess time will tell.

What does AVX2 bring to the table?

EDIT: and what is the current state of next-gen GPUs from AMD?
 
Actually, if this is what's to be expected near term for GPUs out of AMD... Not looking great... (All about efficiency)...

AMD’s R9 280 is one of the best-value graphics cards out there, but it’s still running on old Tahitii GPU hardware from the last generation. The rumour mill is grinding away at the moment, and we're hearing suggestions that AMD are working on a replacement for that old chip, code-named Tonga.

The new Graphics Core Next silicon is meant to solve the problems the current crop of AMD mid-range cards have, namely that they’re both too power-hungry and too hot for mid-range GPUs. The new Tonga GPU is meant to be far more power-efficient, while still being able to deliver good gaming performance.

Some rumours have the Tonga GPU running on a new 20nm production process, but I’m taking that with a big pinch of salt. Nvidia look like they’re going to be releasing a high-end Maxwell GPU later this year on their existing 28nm process; I can’t see AMD getting the jump and going to 20nm on a mid-range refresh GPU.

Whatever the actual spec of the new GPU’s component transistors, it seems likely the Tonga GPUs will slot into AMD's lineup, in a like-for-like fashion, with the current R9 280 and R9 280X. Tonga is expected to arrive with the same 2,048 cores the 280X houses, and the same 2GB GDDR5 frame buffer.

If the new GPUs arrive with the same dollar price, with at least the same gaming performance, but lower temps and power-draw, then we could be looking at some serious mid-range graphics cards. With Maxwell currently doing impressive things in terms of both performance and power efficiency at the low-end, AMD need to get working on their own GPU’s efficiency if they want to stay in the game.

Source
Click to expand...
 
Throughout all the benchmarks I was more surprised by how well the 980X / Xeon W3680 was still holding up after all these years.

http://www.guru3d.com/articles-pages/core-i7-5960x-5930k-and-5820k-processor-review,1.html

Bar missing instruction sets, it's amazing how well it's still going.
On another tech forum I frequent people are buying up old dual socket x58 boards, and CPU's, and just overlocking them to build a bunch of workstations.

I'm interested to see what Apple does for the next Mac Pro update.
 
Last edited:
New Fire Pro Models for the new improved MP 2015..

VirtualRain said:
Actually, if this is what's to be expected near term for GPUs out of AMD... Not looking great... (All about efficiency)...
Click to expand...

Don't forget this article, guys. Its not really about the 280 platform any more.
I think the new Fire Pro models look really promising, don't you think Rain?

The only questions is which one of those models will be pimped up to Apples special requirements and go into the 2015 MP model.

http://www.anandtech.com/show/8371/amd-firepro-w7100-w5100-w4100-w2100
 
This is how it always is. Much cheaper 8-core, 14, 16 or 18 at top end are the bigger changes for CPU. The whole platform sees improvements, but will Apple bring them? I don't see why not. What does it say about the Mac Pro if they don't again? They changed and took control of so much with the new Mac Pro, to think that they didn't consider bringing new updates easily in to this seems a bit silly.
 
VirtualRain said:
I guess time will tell.

What does AVX2 bring to the table?

EDIT: and what is the current state of next-gen GPUs from AMD?
Click to expand...

How about
http://embedded.communities.intel.c...-breakthrough-performance-for-medical-imaging

...introduces a fused multiply-add (FMA3) that effectively doubles the peak floating point throughput ...

...expands most integer Intel AVX instructions from 128 bits to 256 bits, doubling fixed-point (integer) performance...

...with gather support, Intel AVX2 now enables vector elements to be loaded from non-contiguous memory locations to simplify code vectorization...
Click to expand...
IntelNAS1.gif
 

Attachments

  • avx2.png
    avx2.png
    167.9 KB · Views: 246
Last edited:
maplingstorie said:
I think the people here are actually aiming more towards DDR4 for the next update.
Click to expand...

Why? Besides lower power consumption, what are the benefits? It doesn't appear to offer any performance improvement. :confused:

AlexMaximus said:
Don't forget this article, guys. Its not really about the 280 platform any more.
I think the new Fire Pro models look really promising, don't you think Rain?

The only questions is which one of those models will be pimped up to Apples special requirements and go into the 2015 MP model.

http://www.anandtech.com/show/8371/amd-firepro-w7100-w5100-w4100-w2100
Click to expand...

Yeah, while AMD have indeed refreshed their FirePro line (9100, 8100, 7100) with Hawaii and Tonga, these are iterations that are all about efficiency improvements. Apple might jump on these just for improved thermals, but buyers shouldn't expect much in improved performance until AMD releases truly next-gen parts on 20nm process. Again, like Haswell, there's really nothing to get excited about here.

Mr. Retrofire said:
x264 in HandBrake is already AVX2 optimized, AFAIR. And this are only the processors for desktop applications.

----------


And how many threads does HandBrake use in this benchmark!?
Click to expand...

Not sure what you're saying? Of course they are only the desktop CPUs but they are indicative of what to expect from the EP v3 series.
 
How well will Haswells meet the higher/highest advertised turbo boost stages?

VirtualRain said:
Why? Besides lower power consumption, what are the benefits? It doesn't appear to offer any performance improvement. :confused:
... .
Click to expand...

Besides the benefit of lower power consumption in and of itself, lower power consumption may result in a cooler CPU. Both are factors in whether, when, how frequently and the extent to which a CPU turbo boosts. I seem to remember a post, which as I recall, you made about your not seeing your Ivy Bridge CPU reaching the highest advertised levels of turbo boost. Haswells' lower power consumption might be somewhat of a remedy to that complaint.
 
Tutor said:
Besides the benefit of lower power consumption in and of itself, lower power consumption may result in a cooler CPU. Both are factors in whether, when, how frequently and the extent to which a CPU turbo boosts. I seem to remember a post, which as I recall, you made about your not seeing your Ivy Bridge CPU reaching the highest advertised levels of turbo boost. Haswells' lower power consumption might be somewhat of a remedy to that complaint.
Click to expand...

Well, there's a few things I could say in response... I think it's a stretch to think the RAM will impact CPU temperatures... I think any gains in thermal performance will come directly from Haswell's own efficiency gains - not from the RAM (which only draw a few Watts). Secondly, on the matter of max turbo boost, the conclusion was that nothing short of disabling cores into the C3 sleep state (which Apple firmware apparently doesn't support) would provide those maximum multipliers.

I think it's unwise to assume or hope that DDR4 is going to offer any kind of measurable performance improvement in a desktop. It's only real benefit will be in battery life improvements for laptops.
 
VirtualRain said:
Well, there's a few things I could say in response... I think it's a stretch to think the RAM will impact CPU temperatures... I think any gains in thermal performance will come directly from Haswell's own efficiency gains - not from the RAM (which only draw a few Watts). Secondly, on the matter of max turbo boost, the conclusion was that nothing short of disabling cores into the C3 sleep state (which Apple firmware apparently doesn't support) would provide those maximum multipliers.

I think it's unwise to assume or hope that DDR4 is going to offer any kind of measurable performance improvement in a desktop. It's only real benefit will be in battery life improvements for laptops.
Click to expand...

I thought that you were referring to Haswell since Haswell brings DDR4 support to the Mac. My previous response was directed towards Haswell, not the ram that it supports.
 
Tutor said:
I thought that you were referring to Haswell since Haswell brings DDR4 support to the Mac. My previous response was directed towards Haswell, not the ram that it supports.
Click to expand...

Sorry, since you quoted my comment on DDR4, I thought you were talking about DDR4... :eek:
 
VirtualRain said:
Sorry, since you quoted my comment on DDR4, I thought you were talking about DDR4... :eek:
Click to expand...

No problem. I too apologize. I should have done more research on Haswell before responding. Here’s what Wikipedia has to say about Haswell vs. Ivy Bridge, in general:

“Performance[edit]
Compared to Ivy Bridge:
• Approximately 8% better vector processing performance.[9]
• Up to 6% faster single-threaded performance.
• 6% faster multi-threaded performance.
• Desktop variants of Haswell draw between 8% and 23% more power under load than Ivy Bridge.[9][10][11]
• A 6% increase in sequential CPU performance (eight execution ports per core versus six).[9]
• Up to 20% performance increase over the integrated HD4000 GPU (Haswell HD4600 vs Ivy Bridge's built-in Intel HD4000).[9]
• Total performance improvement on average is about 3%.[9]
• Around 15*°C hotter than Ivy Bridge, while clock frequencies of 4.6*GHz are achievable.[12][13][14][15][16][17]” [ http://en.wikipedia.org/wiki/Haswell_(microarchitecture) ].

So, for all of the reasons set forth in Wikipedia's comparison, I believe that you are most likely 100% correct in your overall assessment of what Haswell has to offer to Ivy Bridge MacPro users. If Apple decides that's not enough of an improvement to the MacPro line to upgrade it, then I would also expect that OSX will still maintain the 32-core limitation. However, for an added benefit to those who have clock tweakable Windows systems, I recall reading that Haswell Xeons, at least at the top end, may have several top bins un-locked [ http://vr-zone.com/articles/computex-will-show-desktop-alive-well/77282.html ; http://www.chiploco.com/intel-haswell-ep-e5-1600-v3-35072/ ]. That would allow those whose system bios allows clock tweaking Xeons to achieve significant performance increases by clock tweaking one or two Haswell Xeons. That would be a lot more than any Sandy or Ivy Bridge Xeon offers because all of their bins are locked tight, except for about a 1.0755% maximum upward tweak and as to your reaching that maximum - your mileage will vary.
 
Last edited:
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back