New Macbook Pro SLOWER than Previous?

[rmoroto]

Hey everyone,

Just want to see if I am the only one getting this result. I am running geekbench on my old machine and brand new machine and they are coming out to be slower on the multi core and the single score increase is so very underwhelming. (Macbook Pro 13,3 is the product name for the 15inch).

The new one has 2.9 i7, 16gb, Raedon Pro 460 and 1tb.

 

[dk808]

For some reason geekbench shows faster scores but in real world performance the 2016 is faster. I would suggest watching Jonathan Morrisons latest video to see the comparison

 

[leman]

There are already tons on threads about this. Yes, the Skylake CPUs are barely faster in microbenchmarks. Its just the state of the CPU technology.

 

[ljjycss]

You are expecting too much from Intel.

 

[Freyqq]

I think the difference lies in the loss of 128gb of L4 cache from the iris pro. That increased memory bandwidth for CPU as well.

 

[nstz]

Here's a fun comparison:

http://barefeats.com/hilap2016.html

 

[Spoony Bard]

Geekbench results are not useful--period. Do check out that barefeats link for some relevant test results.

 

[leman]

Here's a fun comparison:

http://barefeats.com/hilap2016.html

Geekbench results are not useful--period. Do check out that barefeats link for some relevant test results.

The problems with those is that they are focusing on GPU performance, which the 2016 model obviously has in abundance. So far, I haven't seen any real-world tests that would concentrate on CPU performance.

 

[Ries]

Skylake brings efficiency, but you won't notice that, because Apple chose to give you smaller battery to earn more money.

 

[magbarn]

I think the difference lies in the loss of 128gb of L4 cache from the iris pro. That increased memory bandwidth for CPU as well.

Yup Skylake was optimized for high DDR4 speeds and the pairing of slow DDR3 + lack of L4 cache is crippling the CPU from performing from it's best potential. If you don't believe this, there's plenty of benchmarks on the web on how much difference fast DDR4 makes for the skylake architecture.

 

[leman]

Yup Skylake was optimized for high DDR4 speeds and the pairing of slow DDR3 + lack of L4 cache is crippling the CPU from performing from it's best potential. If you don't believe this, there's plenty of benchmarks on the web on how much difference fast DDR4 makes for the skylake architecture.

Do you have any links? I found a number of articles, but they are comparing slower 1600/1866 DDR3 to 2133 DDR4, where MBP uses 2133 LPDDR3. So its not really clear if the difference is the RAM clocks or the RAM type. Then again, we don't really know the parameters of that LPDDR3. So far, I haven't seen any detailed tests of the RAM.

P.S. The Anandtech article on this topic finds virtually no difference between 1866 DDR3 and 2133 DDR4.

 

[rmoroto]

Just did another test - converting 175 50.4mp Raw to 15 bit TIFFs. The 2013 finished the task in 22 minutes the 2016 in 24 minutes.

 

[aevan]

Skylake brings efficiency, but you won't notice that, because Apple chose to give you smaller battery to earn more money.

Yes, because, basically - they are evil. It has nothing to do with making the laptop weigh less.

 

[magbarn]

Just did another test - converting 175 50.4mp Raw to 15 bit TIFFs. The 2013 finished the task in 22 minutes the 2016 in 24 minutes.

Not bad for the 2016 MBA 15, It's got almost Pro performance in an Air Chassis ROFL!
What good is a lightning fast SSD when the CPU/GPU is too slow to take advantage of it. I waited 4 years on my rMBP for this?! (I upgraded my laptops annually from 2007-2013) What a POC.

 

[therealseebs]

Do you have any links? I found a number of articles, but they are comparing slower 1600/1866 DDR3 to 2133 DDR4, where MBP uses 2133 LPDDR3. So its not really clear if the difference is the RAM clocks or the RAM type. Then again, we don't really know the parameters of that LPDDR3. So far, I haven't seen any detailed tests of the RAM.

P.S. The Anandtech article on this topic finds virtually no difference between 1866 DDR3 and 2133 DDR4.

2133 DDR3 should be comparable to 2133 DDR4. It's just that this is the very high end for DDR3, and the low end for DDR4. So for comparable cost of memory, DDR4 would tend to win. Differences aren't especially large, but they are noticeable.

 

[Charlesje]

@leman Part two of the barefeat.com tests show some cpu tests.

The cinebench score of 676 for the 2016 2,9 ghz seems strange however, when compared to the ars technica review cinebench score of 696 for the 2,7 macbook pro...

It would be interesting to see more 2,9 or 2,7 macbook pro results (real performance, eg cinebench, prime...). No new owners on this thread?

 

[therealseebs]

I have one, but I haven't run any benchmarks on it.

 

[magbarn]

Do you have any links? I found a number of articles, but they are comparing slower 1600/1866 DDR3 to 2133 DDR4, where MBP uses 2133 LPDDR3. So its not really clear if the difference is the RAM clocks or the RAM type. Then again, we don't really know the parameters of that LPDDR3. So far, I haven't seen any detailed tests of the RAM.

P.S. The Anandtech article on this topic finds virtually no difference between 1866 DDR3 and 2133 DDR4.

This is a OpenCL test on the CPU. BTW DDR4 3000 laptop ram is available for purchase on PC's.

Another memory intensive benchmark

At the very least, if Apple was adamant on LPDDR3, they could've at least used the edram equipped skylake to soften the blow

Skylake responds much more to faster memory than Haswell,Ivy, or Sandy ever did.

Apple's obsession with thinness/lightness in their pro machine is seriously hobbling the performance of their latest laptop. I doubt Kaby Lake would've brought much more performance with such a tight thermal/power envelope and slow memory subsystem.

 

[leman]

@magbarn: its definitely weird to see that DDR4 is faster than DDR3 at the same bandwidth. Maybe something else going on? More efficient prefetch? One of my problems with these charts is that they both show workflows that exhibit linear RAM access and therefore improve in linear relation to the bandwidth. How does it look for more random access patterns? An in-memory database or something like that would be a good test.

 

[grennis]

You're just holding it wrong

 

[therealseebs]

@magbarn: its definitely weird to see that DDR4 is faster than DDR3 at the same bandwidth.

It's not.

It's faster than DDR3 at the same clock frequency. It also has slightly higher latency, and a lot more bandwidth, at any given clock speed.

 

[leman]

It's faster than DDR3 at the same clock frequency. It also has slightly higher latency, and a lot more bandwidth, at any given clock speed.

I was under the impression that the bandwidth formula was the same?

 

[davidf18]

Skylake brings efficiency, but you won't notice that, because Apple chose to give you smaller battery to earn more money.

The smaller battery means a lighter, smaller 15" laptop which for me at least, is very helpful. Since the power adapter is the same as the 2015 model, there unit will charge the battery to full faster. Also Skylake does have additional hardware decoding video codecs, so viewing time on battery could be much better.

 

[therealseebs]

I was under the impression that the bandwidth formula was the same?

Apparently not.

http://www.corsair.com/en-us/blog/2014/september/ddr3_vs_ddr4_synthetic

So, purely-theoretical bandwidth, they should be the same at a given clock rate. Actual rate of memory access with a decent controller? Way faster. Of particular note, DDR4-3000 dual-channel is noticably faster than DDR3-1600 quad channel. In theory, the DDR3-1600 should be roughly equivalent to a 3200MHz dual-channel. In practice, it ends up being noticably slower.

 

[Freyqq]

Yup Skylake was optimized for high DDR4 speeds and the pairing of slow DDR3 + lack of L4 cache is crippling the CPU from performing from it's best potential. If you don't believe this, there's plenty of benchmarks on the web on how much difference fast DDR4 makes for the skylake architecture.

Indeed. I find it interesting that my late 2013 rmbp 15" is still holding up so well - turned out to be a really good time to buy.