Archive for the ‘Apple’ Category

Apple reported 10 millions registered users (of its 800 millions iTunes accounts) for its Music service following the first month (June 30 to end of July). We can suppose there were additional 5 millions registrations in August, then 3 millions in September, and 2 millions the first half of September. This could then match with Apple’s announce mid October stating their were 15 millions still opened Music accounts (and 6.5 millions paid accounts among these).

Apple offered an option to disable automatic renewal of Music subscription, then many users diddn’t wait the 3 months trial ending to cancel (or not) their subscription, they just let it stop by itself in case they wouldn’t renew. Did Apple take into account the users that  left the automatic renewal on for its 6.5 paid accounts ?

On one hand their may be 100% of early “subscribers” (the about 5 millions ones that entered trial bewteen June 30 and mid July) that let their subscription stop by mid October, that could explain the 15 millions active accounts (from 20 millions total trials) reported at that time.

On another hand there should have been at least the same amount that would have renewed their subscription to reach the stated 6.5 paid accounts (most of them would have choosed to renew a few days to one week before trial ending, that is early to mid October). Apple offered an option to initially extend and pay for many months, however few users may have opted for this. Then these same users couldn’t both have renewed and let stop their subscription.

Apple then may have taken into account users that let the automatic renewal option enabled. There shouldn’t be more than 2 millions users that did renew their subscription for some months, that is only 10% of users that entered the trial (and 0.25% of Apple’s iTunes users).

Considering the deceiving quality of the service (iTunes Store is much better and many songs or artists weren’t present in Music service), Apple should lower the price, replace the service by access to the full ITMS catalog, and provide songs in lossless format (a new HD audio format is expected by Apple early 2016).

Also many would have used the trial to copy most of songs (using Soundflower), then Apple shouldn’t provide a Trial (as users still can try iTunes Store by listening to long tracks samples).


A few days after having replaced an Intel 320 160GB SSD with a Crucial M500 480 GB into my mac mini (mid 2010), the fan was going to run at maximum speed (that is 5500 rpm) time to time, wihtout any reason (the computer was cold, also reported by sensors). Using free MacFans control the HD termal sensor indeed seemed to work fine (SSD temperature ranges from 27°C to 40 °C), or it may be that the software also uses Smart informations (that is temperature reported from SSD internal sensor).

The SMC reset (unplung the mac mini for 15s, then plug it, wait 5s and finally power it up) didn’t helped. It seemed that playing games under Windows 7 (through Bootcamp) leader to more and more fan overuns. It was the same under Mavericks, then not Yosemite related.

I also noticed that the PSU (power unit) temperature was somewhat high (58 to 65°C) – didn’t remember that before -, however some reported even higher values for other mac models.

I finally ended using MacFan Controls (added to launch items), set to manage fan speed based on CPU temperature sensor, with 65°C as start for speed increase and 85°C for max allowed temperature (that is max fan speed). I may however change these values to 70/90°C as the mac is colder than before with these settings when playing games (I also uses the Windows version of MacFan controls).

OSX Yosemite slowdown fix

Posted: January 5, 2015 in Apple
Tags: , , , ,

Yosemite’s UI slowdown seems partly due to too many interrputs coming from ACPI kernel driver.

sudo powermetrics -s interrupts

These news were reported by VMWare and Parallels :

It seems running OSX in debug mode (does not have any hit on performance) allows retrieving some speed (despite not at a Mavericks level) :

sudo nvram boot-args=”debug=0x10″

Also OSX Lion leaded to slow UI performance (same 10 times slower than previous Snow Leopard), that was fixed  by Mountain Lion. We wonder then whether we will have to wait for OSX 10.11 to retrieve acceptable performance or Apple is preparing a fix for next Yosemite update.

Update : there weren’t many interrupts per second (1 to 20) on mac mini (mid-2010) and adding the boot args didn’t helped (only 10 percent performance more in Quartz and OpenGL, no improvement in UI speed results). To remove the boot options, use this command :

sudo nvram boot-args=””

After removing these options some of the slight improvements were still present, however I ended with installing Mavericks back on another drive (as its UI is 10 times faster and OpenGL is 50 percent faster – I also now play games under Windows 7 through Bootcamp).

Yosemite / mac mini

Posted: October 22, 2014 in Apple
Tags: , , , , , ,

OSX Yosemite performance on a mid-2010 mac mini (NVidia 320M) reminds first 10.0 version : text rendering is 3 times slower (slow redraw when resizing windows), OpenGL is 2 times slower, and user interface test is 10 times slower ! Also opening windows (files listing) is slower, while the new UI theme shouldn’t require more power (it should be faster as it is simplistic – Googlish / Windowish -, without all previous refinements that ultimately landed into Mavericks, not mentioning still non fixed firewire disks sleep issue and trim issue).

Then Apple may use some new OpenGL 4 only APIs for Quartz, while older macs only support OpenGL 3.3.

The new mac mini have half the CPU cores than previous ones, do not allow adding RAM, and only provide an Iris graphics chipset (Iris Pro would have been at least required). Also new imac’s Nvidia 750M graphics card is old.

We now then have to either buy a retina imac ($2500) or mac pro ($2999), or wait for upcoming Broadwell based hackintosh (that will offer same performance for half power requirements, and then will allow smaller enclosure). Also Intel will release an updated Iris Pro.

We then want Scott Forstall (OS features and design), Jon Rubinstein (hardware leading edge with custom processors – should come with A8x) and Bertrand Serlet (Snow Leopard) back, as new team really isn’t talented (except iPad/iPhone hardware engineering).

Apple may unveil ARM based mac mini running an ARM version of Yosemite this month, as it didn’t release Haswell refresh based mac mini nor would wait next year’s Broadwell processors. The new mac mini may include (many) new PowerVR GX6650 GPU, that is 82x faster than first PowerVR chip in iPhone 3G. It includes 192 FP32 ALU cores with up to 384 FLOPs per cycle, that means up to 250 GFLOPs (384 x 650 000 000) at 650MHz (with 32 bits precision). It is able to process 12 pixels per clock and 12 texels per clock, that is about 8 Gpixels/s and 8 Gtexels/s.

Intel Iris Pro 5200 chipset reachs 832 GFLOPS, 10.4 Gpixels/s and 20.8 Gtexels/s. That is a mac mini fitted with 3 PowerVR GX6650 chips may run as fast, for less than 10W power consumption (about 2-3W per PowerVR GX6650 chip at that frequency), while the Iris Pro requires 47W.

Apple A7 performance reached 75 Gflops at 1.3 Ghz (that is 4W power consumption). The new A8 (also 2 cores hower at 1.4 Ghz) is 25 percent faster and 50 percent more power efficient. As an A7 processor at the same frequency would require 4.3W, the A8 power consumption may be about 2.2W (and that is including a PowerVR GX6450), then an A8 fitted with a PowerVR GX6650 may require 4 to 5W, and reach 94 Gflops.

The new mac mini (or iMac) may finally include 4 A8 chips (that is 8 CPU cores) with a PowerVR GX6650 each, and then reach 376 Gflops for CPU and 1000 Gflops for GPU. It would then be close to a common x86 Intel Corei5 (fitted with HD4400 GPU chipset), that reaches 430 Gflops for CPU (however at 35W power consumption compared to about 18W for a quad A8) and far more less score (about half performance) for GPU. Also the extra GPU resources available on the quad A8 may be used through OpenCL and compensate for the lower base CPU performance.

The new mac enclosure may be a smaller version of new mac pro’s one. It may be priced $500.

Update : since Apple would have to provide some Intel emulation (as with Snow Leopard’s Rosetta for PPC) for existing applications, this would make the switch to ARM useless (Intel cores processors were at least two times faster than PPC processors, while ARM are slower). Apple indeed would have to use twice the A8 processors count than previously stated, that is 8 chips / 16 CPU cores, then exactly the same power usage as Intel’s Core i5.

OSX Mavericks 10.9.1 update

Posted: December 17, 2013 in Apple
Tags: , ,

The long awaited first update to OSX Mavericks is available. However it doesn’t fix the Firewire sleep issue with external drives.

Mac mini 2013/2014 coming

Posted: December 10, 2013 in Apple
Tags: , ,

New Mac Pro was released only few weeks before holidays, as with pro users these timings don’t count. Also the mac mini isn’t what leads Apple (while more and more are buying it on refurb) and wasn’t updated during October’s update.

Then the new mac mini may be released at any time, and one forum user reported some information that may hint at an upcoming release in the following weeks, bringing Haswell and Iris graphics :

I work in an Apple Reseller and like many of you i’m waiting for the mini 2013 refresh.

From monday the two major Apple suppliers in Italy are suddenly and completely out of stock of minis. I know that this happen from time to time, but the timing is no coincidence.

Trust me, a new mini is coming next week, or at least we have solid evidence to believe it.