Krokodile Cartridge

[Kroko]

I asked Atmel support for a typical endurance of their flash chip and here is the answer:

Dear Armin:

 

   The typically endurance for the AT29C040A is 100K cycles. However, we

only guarantee 10K cycles.

 

Victor

[Stingray]

I've been in contact with Kroko and am planning to write an OS X native app.  I'm just waiting for my Krok Cart to ship at this point.

 

Outstanding, I look forward to your software.

 

-S

[PAH]

My experience with flash devices is that as they wear out from too many erase/write cycles, they get slower (i.e. they take longer to erase/write) rather than just stop working.

 

Once I left a test running at work while I was on vacation. When I left it was logging about 500 messages a second to a file. When I got back from vacation a week later, it was still running, but had slowed to a crawl - about 1 message every 5 seconds. It took a while before I realized I had been logging to a flash drive, and had essentially worn the flash out.

 

There were no errors in the flash drive; it just took about 5 times as long to erase and write a block as it did when the flash was brand new.

 

From what I've read, after about 10,000 erase write cycles, they take about 10% longer to erase/write. After 100,000 erase/write cycles, they take about 500% longer (i.e. 5 times as long). So it takes a long time to start noticing it getting slower, but after you start noticing it, it will keep getting slower at a faster pace.

 

Keep in mind, each erase block wears based on how many times it has been erased/written. Erasing and writing one block has no effect on the wear of the other blocks. That is why flash file systems use wear leveling to even wear the flash. I doubt the Krokodile cart can do wear leveling, so I image every time you load different stuff into it, the boot menu has to be updated in a constant block. If so, that block would wear out the fastest. But you should still get your 100,000+ programming cycles - it will just get slower.

[Thomas Jentzsch]

Couldn't that also be the result of a built-in defect management?

[vdub_bobby]

Once I left a test running at work while I was on vacation.  When I left it was logging about 500 messages a second to a file.  When I got back from vacation a week later, it was still running, but had slowed to a crawl - about 1 message every 5 seconds.  It took a while before I realized I had been logging to a flash drive, and had essentially worn the flash out.  

 

There were no errors in the flash drive; it just took about 5 times as long to erase and write a block as it did when the flash was brand new.

From 500 write/1 sec to 1 write/5 sec is 2500 times as long as brand new. Of course, at 500 /sec, you had almost written 100,000 times in the first day!

[PAH]

Couldn't that also be the result of a built-in defect management?

 

I don't think large block erase flash devices have runtime defect management. Sector (512 byte) erase flash designed for things like SD cards may, but the devices I was using had 64K byte erase blocks. Even after it getting that slow, the flash still had the original capacity, and I'm sure they don't build in 100% extra for defect management.

[PAH]

Once I left a test running at work while I was on vacation.  When I left it was logging about 500 messages a second to a file.  When I got back from vacation a week later, it was still running, but had slowed to a crawl - about 1 message every 5 seconds.  It took a while before I realized I had been logging to a flash drive, and had essentially worn the flash out.  

 

There were no errors in the flash drive; it just took about 5 times as long to erase and write a block as it did when the flash was brand new.

From 500 write/1 sec to 1 write/5 sec is 2500 times as long as brand new. Of course, at 500 /sec, you had almost written 100,000 times in the first day!

 

However, these were data writes into a file within a flash file system. Not all writes would cause an erase cycle. Many of them would just write to an already erased block. For example, if you erase a block (64K in this case), then write 512 bytes, then write another 512 bytes,... after 128 writes of 512 bytes, you finally fill up the block. That counts as 1 erase/write cycle. Then you erase and start writing to a different block. When you're done with that that block also has had 1 erase/write cycle. The flash file system has a wear leveling algorithm to make sure that all blocks get erased/written very close to the same number of times. I had a 16MB flash with 256 64K blocks. This would take 256*128 or 32768 writes of 512 bytes each to erase/write then entire flash once.

 

When you take the wear leveling algorithm and flash file system overhead into consideration, it starts doing cleanup erase-writes more frequently as the drive fills up.

 

In any case the first 100,000 writes to my file would have only erased/written the entire flash device a handfull of times, but as my log file got huge and filled the drive up, there would have been a lot of overhead erase/writes going on and the device would start getting hammered much more rapidly.

[Albert]

Just a heads up for those who pre-ordered the Krokodile Cart and are patiently waiting for it to arrive.. The first batch of 30 carts will be shipped Monday morning! The latest delay was due to Andrew Davie finding a bug in the 3E Bankswitching mode, which he is using to develop his new game engine. That has been fixed, and I have reprogrammed the firmware in the first batch of carts that Armin Vogl sent me. With that out of the way, all the other elements are also complete (label, manual, CD) and we can get the show on the road.

 

Here are some pictures I took last night,

 

 

 

..Al

[Great Hierophant]

I hope PCs continue to come with Serial Ports, as they seem to be getting scarcer. Most current high end motherboards have only one. I have even seen a few with none, replacing them with audio connectors for on-board audio controllers.

[keilbaca]

They also make USB to Serial adapters, so that will not be an issue for those that don't have a serial port.

[Thomas Jentzsch]

I still hope the next version will be wireless (Bluetooth!).

[Albert]

I still hope the next version will be wireless (Bluetooth!).

 

A bluetooth cart would be pretty cool, although if you wanted to program the cartridge outside of the 2600, you'd still need to get power to the cart somehow. Unless you included a battery of some sorts (a lithium ion battery might work well). It would be awesome if you could program the cart while it's sitting in the 2600, via Bluetooth. Select a new game (or multicart image) from your computer, click "Program", then reset your 2600 and the game's there!

 

..Al

[Thomas Jentzsch]

A bluetooth cart would be pretty cool, although if you wanted to program the cartridge outside of the 2600, you'd still need to get power to the cart somehow. Unless you included a battery of some sorts (a lithium ion battery might work well).

But why would you ever want to take the cart out again then?

 

Maybe something recharable would work too.

[AtariAger]

The first batch of 30 carts will be shipped Monday morning!  

..Al

 

Hey Al...

 

Is there a viewable list of the first 30 people who pre-ordered the Krok cart or will the recipients be notified by email... ?

 

Thanks much...

 

-Tom

[adams_ale]

How many Kroks were pre-ordered?

[Albert]

Is there a viewable list of the first 30 people who pre-ordered the Krok cart or will the recipients be notified by email... ?

 

I just sent email to everyone who pre-ordered the Krokodile Cart with a more detailed status update.

 

..Al

[Haydn Jones]

i am planing of ditching pc's for good and getting a mac mini. when the mac driver is finished, will you consider bundleing a usb to serial lead. (could be good for laptop owners as well)

 

H

[Zach]

I'm enjoying the Krokodile cart, and I hope everybody who wants one can get one. Still, I have a concern about the hardware:

 

I don't know much about electricity, but I'm in the habit of turning off the PC before connecting anything out of a vague understanding that it is safer for the hardware. The KC is the first hardware that I connect and disconnect frequency. How risky is it to do so while the PC is on?

[mos6507]

Even for the developer, who may download quite often (Thomas!) we have invented this incremental download feature, which will only change parts that really need to be changed. This is also a good way to make it survive longer  

My biggest concern would be, that if fails too late and I cannot find any replacement chips anymore.

 

 

Then stock up on replacement chips now while they are still readily available.

[Kroko]

I'm enjoying the Krokodile cart, and I hope everybody who wants one can get one. Still, I have a concern about the hardware:

 

I don't know much about electricity, but I'm in the habit of turning off the PC before connecting anything out of a vague understanding that it is safer for the hardware. The KC is the first hardware that I connect and disconnect frequency. How risky is it to do so while the PC is on?

 

There is no risk at all. You can connect and disconnect it as often as you want - computer on or off.

 

Armin

[firebug]

for those of us who missed out on the first batch, when will the next batch be ready for pre-order?

[adams_ale]

from the e-mail that Al sent to everyone, the second batch is anticipated to be shipping out next week.

[firebug]

so the second batch is already spoken for? are there going to be any more made?

[adams_ale]

Yes. Krokodile Carts will be available in the AA store after all pre-orders have been filled. There is also the Cuttle Cart.....

[Albert]

Yes.  Krokodile Carts will be available in the AA store after all pre-orders have been filled.  There is also the Cuttle Cart.....

 

Yes, this is exactly right. Right now we are focused on getting the second batch of pre-orders shipped, and after that we will work on producing a run of carts that can be sold without delay. Once those are done, the Krokodile Cart will return to the store.

 

..Al