Originally Posted by dorianmuthig

Don't flash a custom ROM, flash a vendor ROM for your area (via goldcard method, will run you the equivalent of ~10-15 bucks) and you should be set.

Thank you, will try it out.

This memory works on the HD7? My Sandisk memory (32GB) isn't compatible with my phone and it do automatically an hard reset

I will try this method.

This is a Patriot PSF32GMCSDHC43P 32GB Class 4 MicroSD card. From what I read, it is unsuitable due to the inconsistent 4k write speeds and inconsistent random 512k and sequential write speeds. Is this correct?

-----------------------------------------------------------------------
CrystalDiskMark 3.0.1 x64 (C) 2007-2010 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 19.444 MB/s
Sequential Write : 5.510 MB/s
Random Read 512KB : 18.455 MB/s
Random Write 512KB : 0.538 MB/s
Random Read 4KB (QD=1) : 3.127 MB/s [ 763.5 IOPS]
Random Write 4KB (QD=1) : 0.980 MB/s [ 239.3 IOPS]
Random Read 4KB (QD=32) : 3.574 MB/s [ 872.6 IOPS]
Random Write 4KB (QD=32) : 0.171 MB/s [ 41.8 IOPS]

Test : 4000 MB [L: 0.0% (0.0/28.9 GB)] (x1)
Date : 2012/07/03 19:23:26

Originally Posted by dorianmuthig

If a card works depends on parameters based on which it was manufactured. The model number is no longer relevant as SanDisk/Toshiba fabs have all moved on to newer processes, that are either better than all previous ones or completely incompatible.
Here's what doesn't work with first generation Windows Phones like the HTC HD7 or the Samsung Focus:

• 8KB sector NAND (this is used to save wafer space, because it reduces addressing circuitry)
• 256 sector (or higher) block NAND (blocks are the internal NAND address channels inside the flash chip control processor for simultaneous access and determine sequential R/W or NCQ write speeds)
• SDXC cards
• Class 10 and higher cards (High Speed Bus I/F)
• UHS cards (UHS-I Bus I/F)

The only thing you can do, is buy and test a card (class 2,4 or 6), and return it if it doesn't work. You can exclude a lot of cards prematurely by running a benchmark like CrystalDiskMark 3.

On all second generation and newer phones, the internal flash is a soldered on SD/MMC/SDIO NAND chip and cannot be replaced.

does this upgrade memory cause to prevent update to tango?

Originally Posted by Riley600

This is a Patriot PSF32GMCSDHC43P 32GB Class 4 MicroSD card. From what I read, it is unsuitable due to the inconsistent 4k write speeds and inconsistent random 512k and sequential write speeds. Is this correct?

-----------------------------------------------------------------------
CrystalDiskMark 3.0.1 x64 (C) 2007-2010 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 19.444 MB/s
Sequential Write : 5.510 MB/s
Random Read 512KB : 18.455 MB/s
Random Write 512KB : 0.538 MB/s
Random Read 4KB (QD=1) : 3.127 MB/s [ 763.5 IOPS]
Random Write 4KB (QD=1) : 0.980 MB/s [ 239.3 IOPS]
Random Read 4KB (QD=32) : 3.574 MB/s [ 872.6 IOPS]
Random Write 4KB (QD=32) : 0.171 MB/s [ 41.8 IOPS]

Test : 4000 MB [L: 0.0% (0.0/28.9 GB)] (x1)
Date : 2012/07/03 19:23:26

This card won't work. (It still might, but it is below recommended spec, that 41 IOPS should be 200 on allocated sectors)

Originally Posted by saeedk64m

does this upgrade memory cause to prevent update to tango?

Not to my best knowledge, then again the Tango update isn't out for HTC phones, yet (only select Nokia phones). Updating will be using the same traffic scheduling method as the other 7.5 updates did.

Originally Posted by dorianmuthig

This card won't work. (It still might, but it is below recommended spec, that 41 IOPS should be 200 on allocated sectors)



Not to my best knowledge, then again the Tango update isn't out for HTC phones, yet (only select Nokia phones). Updating will be using the same traffic scheduling method as the other 7.5 updates did.

Thanks for the input. I'm a little confused because in this post you say that QD=32 doesn't matter as much as QD=1. Based on my results for the Patriot card, they all exceed what you've posted there for the 8gb J35A card (except the 512k random writes which are about half as fast). Or is it that I'm missing something/not interpreting your post correctly.

Thanks!

Originally Posted by Riley600

Thanks for the input. I'm a little confused because in this post you say that QD=32 doesn't matter as much as QD=1. Based on my results for the Patriot card, they all exceed what you've posted there for the 8gb J35A card (except the 512k random writes which are about half as fast). Or is it that I'm missing something/not interpreting your post correctly.

Thanks!

Yes, they don't matter as much, because you will likely never reach a queue depth higher than 5. However since there is an obviously large difference between the QD32 and the results without command queuing (the former should usually be faster) I would be very careful with that card as this is an indicator for sequential write optimization logic in the SDIO processor inside the card, there may even be something seriously wrong with it.
You're welcome to try it in your phone, however, the card should satisfy the absolute minimum performance requirements for 1st generation phones, performance wise. Since the last HTC vendor update, IO is cached as well, so you shouldn't run into too many issues, but the only definitive way to find out if a card works is to actually put it in a phone. Performance should be better than before (opening apps and syncing media should be faster), if it isn't you may run into problems.

And as far as the original 8GB card goes, that card is actually inadequate, performance wise, as a lot of vendor installed cards were at the time, which is an issue Microsoft has been made aware during development of Mango.

Thanks for the info, that makes sense. From what I read, the Sandisk B35 cards can be hit and miss as well now, right?

Originally Posted by dorianmuthig

Yes, they don't matter as much, because you will likely never reach a queue depth higher than 5. However since there is an obviously large difference between the QD32 and the results without command queuing (the former should usually be faster) I would be very careful with that card as this is an indicator for sequential write optimization logic in the SDIO processor inside the card, there may even be something seriously wrong with it.
You're welcome to try it in your phone, however, the card should satisfy the absolute minimum performance requirements for 1st generation phones, performance wise. Since the last HTC vendor update, IO is cached as well, so you shouldn't run into too many issues, but the only definitive way to find out if a card works is to actually put it in a phone. Performance should be better than before (opening apps and syncing media should be faster), if it isn't you may run into problems.

And as far as the original 8GB card goes, that card is actually inadequate, performance wise, as a lot of vendor installed cards were at the time, which is an issue Microsoft has been made aware during development of Mango.

Originally Posted by Riley600

Thanks for the info, that makes sense. From what I read, the Sandisk B35 cards can be hit and miss as well now, right?

Pretty much, but if you get a recent card, the miss part should be very low.