The OFFICIAL... Neosd MicroSD Card thread...

[Rot]

Rot, what's the point of this thread if not discussing timers? I mean, if trough Raz's analysis, we are gonna end with a clear picture about SD cards and factors influencing loading times, I guess it should be welcome..or not? (at least in this very thread, I 100% agree being anal about this i.e. in the general thread, would be just meh-sturbation)

Meh Massi...

Over the past 3 weeks... I've sat in chat and watched the forum play silly buggers with the Neosd team...

Neosd and Neodev had to deal with:

1. Scammers...
2. People taking the piss... complaining about mainly cosmetic stuff...
3. People not understanding we are on a 3 decade old system and there will be variation across the board due to... age, components used and set-ups etc etc...
4. People complaining for the sake of complaining...

The reason I'm involved is this state of affairs is... it's annoying me somewhat...

xROTx

PS. I've had 2 hours sleep... tired... just sayin' there maybe limited Rot response "type"...

[Yodd]

Yodd, yup the previous flashed game indeed plays a role in the flash time but that is not part of this issue. Its why its important to flash the game twice and only time the 2nd run. Its a weird situation.

It has only come to light becase neosd mentions a time of 2:42 to flash kof2003 and in my effort to hit the same time on the neoSD, I can only hit a time of 3:25 (various mvs tried). Thats using an sdcard neosd sent me which he verified flashed in 2:42 before he sent.

We are just trying to gather info to find the cause and hopefully if possible a fix that can be applied via firmware.

Yeah, with the current firmware I hit right around 2:40 flashing KoF2003.


Has NEOSD looked at your neosd to see if there is some issue with it?


Cause if pretty much everyone else is getting faster flash times than you and you have tried different MVS boards...I think the conclusion is kinda obvious.

[neosd]

Yeah, with the current firmware I hit right around 2:40 flashing KoF2003.


Has NEOSD looked at your neosd to see if there is some issue with it?


Cause if pretty much everyone else is getting faster flash times than you and you have tried different MVS boards...I think the conclusion is kinda obvious.

He has a preseries NEOSD we gave him free of charge, his cart is different, minor changes, but there are changes

[Razoola]

Yup, its different. Its been ruled out however that this does not play a part given others with the final version have the same issue and your partner said his pre version (same spec as what you gave me to review) does not either. The tests in this thread also confirm this and you have known this for some time so I'm not sure why you point this out again like it means something.

Yodd, neosd already knows I have no problem returning the pre version but given neodev has neosd in his hands that also suffers the slow erase times like me its not as important to have the neosd I have on hand to help find the cause. When I agreed to review the neoSD it was also agreed a final neoSD version would be sent but this has not happened yet because neosd has been very busy. I guess once thats sent to me it will solve the flash times for me at least (if its tested first), at the end of the day though it important to see if a fix is possible for everyone effected via firmware.

[Yodd]

Maybe it's the phase of the moon. Or solar flares.


Yeah, solar flares.

[Razoola]

Maybe it's the phase of the moon. Or solar flares.

Yeah, solar flares.

We do get the Northern lights often in our part of the world

[Sceptre_JLRB]

Maybe it's the phase of the moon. Or solar flares.


Yeah, solar flares.

...Or cosmic rays! ...Or even meigas' things!!!

[massimiliano]

Meh Massi...

Over the past 3 weeks... I've sat in chat and watched the forum play silly buggers with the Neosd team...

Neosd and Neodev had to deal with:

1. Scammers...
2. People taking the piss... complaining about mainly cosmetic stuff...
3. People not understanding we are on a 3 decade old system and there will be variation across the board due to... age, components used and set-ups etc etc...
4. People complaining for the sake of complaining...

The reason I'm involved is this state of affairs is... it's annoying me somewhat...

xROTx

PS. I've had 2 hours sleep... tired... just sayin' there maybe limited Rot response "type"...

Fair enough, I had no visibility about it, I understand, thanks ...I guess that's why I like so much having you behind this... so I do not have to think too much!

[Rot]

Fair enough, I had no visibility about it, I understand, thanks ...I guess that's why I like so much having you behind this... so I do not have to think too much!

It has been suggested that I have bathed the Neosd in cotton wool... and to a point it's a valid statement...

However, I take the forum as a whole... for example... a LOT of our memberbase have the Neosd device now... there have been no complaints about it not doing what it's meant to do...

IF there was an issue with it not working... sure... I'd feel it's a valid complaint...

But concerning this whole load time situ... There's sometimes a 20% variance... IMO... so what, as long as it works...

The Neosd team have done very well concerning firmware updates to try and ensure the device works on the massive catalogue of NG.com hardware and individual setups... and that to me is where it really is...

People have argued in the past with myself as to where I draw the lines between what I feel is a valid concern and what they feel needs to be addressed... and I'm not perfect so I review my position from time to time...

However, I feel this load time thingy is a non-issue and the Neosd team are applying themselves very well here...

xROTx

PS. I draw imaginary lines in the sand... sometimes I shift those lines if you can persuade me...

I don't change my mind very often... but I do listen...

[Razoola]

However, I feel this load time thingy is a non-issue and the Neosd team are applying themselves very well here...

I'm not sure that can apply to neosd's last post here on the matter, given he knows that aspect had been ruled out.

[Neodogg]

Why isn't all this data in an easy to read spreadsheet with graphs?

[Rot]

Why isn't all this data in an easy to read spreadsheet with graphs?

You have my permission Poochie to create this excel spreadsheet and pie chart if you like...

I'm lazy...

xROTx

[Niko]

PNY_SD.jpg

0. Neo Geo AES
1. PNY Elite 32GB (P-SDU32U185EL-GE)
2. NeoSD Firmware v1.05

KOF2k3: 3:06
MagLord: 0:21

[Gyrian]

Why isn't all this data in an easy to read spreadsheet with graphs?

It is now. I transcribed every documented test in this thread (32) with as much detail as I could discern.

Google Doc Link

I'm not great with this tool, so the graph kind of sucks. The big data table is fully filterable, though. Kind of interesting to see them all like this in one place. No big insight beyond what we've been discussing all along, although it is apparent that variances are much wider on AES. SD Card brand & model don't appear to be large contributors, saving a handful of seconds between 'fast' and 'slow' models.

[DaytimeDreamer]

It is now. I transcribed every documented test in this thread (32) with as much detail as I could discern.

Google Doc Link

I'm not great with this tool, so the graph kind of sucks. The big data table is fully filterable, though. Kind of interesting to see them all like this in one place. No big insight beyond what we've been discussing all along, although it is apparent that variances are much wider on AES. SD Card brand & model don't appear to be large contributors, saving a handful of seconds between 'fast' and 'slow' models.

Great stuff. So longer flash times for KOF2003 are only to be found in the AES version of NeoSD. Wonder if it's something to do with the AES hardware itself. Again, great work with putting this together.

[Gyrian]

Great stuff. So longer flash times for KOF2003 are only to be found in the AES version of NeoSD. Wonder if it's something to do with the AES hardware itself. Again, great work with putting this together.

Thanks! Yes, the AES version is a touch slower than MVS. Neodev explained it here earlier:

Yes, AES is a bit slower flashing, that's normal. It happens because on MVS, in order to write the flashes, the MCU has direct access to the data bus and can stream data directly to the C flashes, but on AES, it goes through the sprite serializer chip and it introduces a small delay as there is a small protocol there to select which rom each value goes to.

[Razoola]

Unfortuntely although you have put a good effort into the table which is commendable it is inaccurate as a tool for finding out loading speeds based on microSD card brand and type. It is known there can be a 20% difference in speed based on the physical NeoSD which has nothing to do with the microsd card used, the type of system (AES/MVS), tempretures and voltages supplied to the unit.

To be clear, with the information as it is now at the moment it can be said the expected loading time for kof2003 is 3:00 plus or minus 10% variance, and that the fastest sdcards cannot be assumed to give the fastest loading speeds. A mircoSD card for one person may result in a loading time of 2:42. That same microSD card used in another NeoSD in the same NeoGeo can result in loading time of 3:25 and no faster.

The exact cause of this is not known but the neogeo, voltage and tempreture have been ruled out as the main cause. Presently, because a firmware bug has not been ruled out (to my knowledge) this difference in loading times is being put down to variances in flashchip preformance on each NeoSD unit. This means that while the NeoSD works and flashes correctly, you are either blessed with a NeoSD that can flash games quicker than the average, or you are unlucky and have a NeoSD than is slower than the average. While a faster microSD card can gain you some seconds it cannot get around the limiting time needed to erase and programe the flashchips.

[Gyrian]

I concluded as much too, Raz.
As it stands now, it's a decent sample describing what to expect within the normal range when you buy one of these. One bit that's interesting from the dataset is the frequency of the "20% speed difference" factor, which if roughly defined by KOF2003 times of around 3:20, has this showing up in 4 out of 10 NeoSD AES. The data also show much smaller impact for the MVS version, if we define 20% units as those posting times of around 2:55 (much smaller time spread). The fastest AES units are roughly comparable to the slowest MVS units.

[Razoola]

Yes, Whats annoying is if you have a neoSD flashing in the 3:20 range, there is presently no known steps you can do to improove the situation.

There probably is not enough information to say the following is right with certintity but I would guess (and remember this is a guess based on whats known so far only) that there is one of three speed ranges of each NeoSD is able to flash kof2003 providing a spec class10 SD card is being used. Roughly around 2:40-2:55, 2:55-3:10 or 3:20-3:35, Times within these ranges are then related to things like microSD card speed, voltages and tempreture.

[greatfunky]

The fastest AES units are roughly comparable to the slowest MVS units.

Not in my case , i got the same time on kof 2003 , 2:50 on both neosd ,which is not bad for a mvs version.
I really can't decide which is the faster , i probably got a good AES version.

[GadgetUK]

If you take the 2:55 vs 3:35 (that is, slowest times for the fastest and slowest examples)

It breaks down to

175 seconds vs 215, which comes out at approx 19% difference.

The reason we are seeing this an no one has queried it with other flash carts is you don't notice it on other flashcarts due to the often very short programming times related to tiny ROM images. For example, 19% difference between 2 x Everdrive or SD2SNES etc could result in a negligble difference. If you time say 5 seconds to program a cart on a SNES or Megadrive, would you honestly notice between 0.5 sec to 1 sec difference to anyone elses cart - almost definitely not, and in fact I've never seen anyone report that as a problem because no one else has noticed it up until now, and that's because the games on the neo geo are obviously much larger and this technology just hasn't really been used for games this large before.

An alternative is if you get a cart you think is too slow for you, just stick it on eBay and buy another - you will get 100% asking price (maybe more) due to PayPal on eBay! The reality is the manufacturer does not tell you in advance which flash ROMs will perform at the better end of that 20% tolerance and which perform at the slower end of that 20% - so what are the Neo SD team expected to do? I would suggest when they update their website with specs etc, they just clearly state "average programming time for KoF2003 is ~3 minutes, but it could be up to 20% more or less and performance can be affected by micro SD speed".

The other alternative is for Neo SD to start using even faster lower tolerance chips and increase the cost by a hefty percentage - wild guess, additional 100 euros or more, and maybe then it might have 10% tolerance still.

I am sure if there's any way to improve performance in future, they will let us know. Their whole aim from the start here was to make this as stable and as fast as possible, and as said before we are all better off than where we started because the 'faster firmware update' was around 30% faster to start with. Beyond that, what else would you expect, a recall to swap out the ROM chips with faster ones (that would involve massive costs and no one in their right mind would do it)???? For one thing, think of the cost of the chips alone - lets say 10% of them perform 20% faster than others, does that mean they have to start binning or selling 90% of the chips they buy in, and the logistics behind testing those chips in advance would be an absolute nightmare.

I did suggest that if people start getting stressed about this that the next firmware update should slow all carts down to equal footing - that sounds counter productive, but it seems to me like people are getting stressed about that 19% and it might stop the concerns here.

Sorry if that sounded like a rant, it wasn't supposed to be. It just feels to me like this thread has turned into another thread to p*** off the Neo SD team.

[greatfunky]

If you take the 2:55 vs 3:35 (that is, slowest times for the fastest and slowest examples)

It breaks down to

175 seconds vs 215, which comes out at approx 19% difference.

The reason we are seeing this an no one has queried it with other flash carts is you don't notice it on other flashcarts due to the often very short programming times related to tiny ROM images. For example, 19% difference between 2 x Everdrive or SD2SNES etc could result in a negligble difference. If you time say 5 seconds to program a cart on a SNES or Megadrive, would you honestly notice between 0.5 sec to 1 sec difference to anyone elses cart - almost definitely not, and in fact I've never seen anyone report that as a problem because no one else has noticed it up until now, and that's because the games on the neo geo are obviously much larger and this technology just hasn't really been used for games this large before.

An alternative is if you get a cart you think is too slow for you, just stick it on eBay and buy another - you will get 100% asking price (maybe more) due to PayPal on eBay! The reality is the manufacturer does not tell you in advance which flash ROMs will perform at the better end of that 20% tolerance and which perform at the slower end of that 20% - so what are the Neo SD team expected to do? I would suggest when they update their website with specs etc, they just clearly state "average programming time for KoF2003 is ~3 minutes, but it could be up to 20% more or less and performance can be affected by micro SD speed".

The other alternative is for Neo SD to start using even faster lower tolerance chips and increase the cost by a hefty percentage - wild guess, additional 100 euros or more, and maybe then it might have 10% tolerance still.

I am sure if there's any way to improve performance in future, they will let us know. Their whole aim from the start here was to make this as stable and as fast as possible, and as said before we are all better off than where we started because the 'faster firmware update' was around 30% faster to start with. Beyond that, what else would you expect, a recall to swap out the ROM chips with faster ones (that would involve massive costs and no one in their right mind would do it)???? For one thing, think of the cost of the chips alone - lets say 10% of them perform 20% faster than others, does that mean they have to start binning or selling 90% of the chips they buy in, and the logistics behind testing those chips in advance would be an absolute nightmare.

I did suggest that if people start getting stressed about this that the next firmware update should slow all carts down to equal footing - that sounds counter productive, but it seems to me like people are getting stressed about that 19% and it might stop the concerns here.

Sorry if that sounded like a rant, it wasn't supposed to be. It just feels to me like this thread has turned into another thread to p*** off the Neo SD team.

100% agree with you Gadget , I find that story of speed does not lead anywhere now ....

[Rot]

OK OK... NEXT person to start complaining about load times that is not completely necessary.... gets 24 hours in the War room...

NO EXCEPTIONS!

xROTx

PS. ....YOU KNOW WHO I AM REFERRING TOO!

[Razoola]

I am sure if there's any way to improve performance in future, they will let us know. Their whole aim from the start here was to make this as stable and as fast as possible.

I would suggest when they update their website with specs etc, they just clearly state "average programming time for KoF2003 is ~3 minutes, but it could be up to 20% more or less and performance can be affected by micro SD speed".

Sorry if that sounded like a rant, it wasn't supposed to be. It just feels to me like this thread has turned into another thread to p*** off the Neo SD team.

I think you have got it a little out of porportion there. No one is trying to rant or piss of neosd (or anyone else), this thread is about posting information gathered in relation to loading. You may very well be right about these kinds of differences affecting other devices but because in general the games are smaller the loading times are not as noticeable. Likewise if you look in relation to maglord loadtimes you can see these issues affecting kof2003 load times do not follow though for that game at all.

A time of 2:42 loading kof2003 was reported by neosd himself in the main thread and a microsd card known to hit that speed for neosd could only hit 3:27 on another NeoSD. Currently neosd himself holds the record for the fastest loading kof2003, this puts the percentage differentce a little over 20% but what a percent here or there.

I agree that if there is a way to resolve this situation in the future neodev will surely add it to a firmware. Bear in mind however its taken about two months of highlighting this situation (privately also) to get to a situation where neosd acknowledged this situation exists. Before then he said it was impossible for anything on the neosd to cause this difference in speed and it was all microsd card related. Even you yourself stated I likely had a fake sony microsd card (given I had 3:25 plus loading speeds). The issue only came to light because I wanted to be able to take the fastest possible loading neosd to the NeoGeo meet in Helsinki too show it off and verify the fast loading times reported.

I'm with you 100% that it would be really really good if neosd updated the main NeoSD page in the specs section stating that the 'expected loading time for kof2003 is 3:00 minutes +/- 10% even with a fast sdcard'. I have suggested it before already directly to him a month or so ago but he is yet to do it.

[Happosai]

how long is the lifetime of the AES chips ? How many times can we write/delete/rewrite games on the cardmod ?