snufkin wrote on 2021-04-25, 19:43:

Ah, this is firmly in to diminishing returns territory for the possible space savings. Yours is probably as far as is sensible. […]
Show full quote

Doornkaat wrote on 2021-04-25, 12:14:

If you think it's worth your while I posted the schematics in this thread so feel free to alter the design however you please. (And also to post your design here. 😀 )

Ah, this is firmly in to diminishing returns territory for the possible space savings. Yours is probably as far as is sensible.

Although...
Straight connector (I think): https://www.digikey.co.uk/product-detail/en/t … 01-4-ND/1864025
Surface mount dip switch: https://uk.farnell.com/alcoswitch-te-connecti … -smd/dp/2792786

Now, me, I was sketching out a plan to go the other extreme. Couple of 16 position rotary switches, low voltage micro-processor, charge pump to get ~3V from Vcore, some shift registers for the 12 outputs. Could then just dial the Vcore and frequency up and down, and avoids any possible voltage level issue. Attach a thermometer to the uP and could maybe start to implement thermal throttling/frequency boost (do we know if the athlon only reads FID at boot?). I'm sure there're more ways to make it needlessly complicated (bluetooth with control app on phone?). All for something that probably gets set once... So I managed to stop doing that.

PLEASE can you go back to doing ‘that’ (bluetooth with control app on phone)!!!!!!! 😁

Xplo wrote on 2021-04-25, 19:54:

snufkin wrote on 2021-04-25, 19:43:

Now, me, I was sketching out a plan to go the other extreme. Couple of 16 position rotary switches, low voltage micro-processor, charge pump to get ~3V from Vcore, some shift registers for the 12 outputs. Could then just dial the Vcore and frequency up and down, and avoids any possible voltage level issue. Attach a thermometer to the uP and could maybe start to implement thermal throttling/frequency boost (do we know if the athlon only reads FID at boot?). I'm sure there're more ways to make it needlessly complicated (bluetooth with control app on phone?). All for something that probably gets set once... So I managed to stop doing that.

PLEASE can you go back to doing ‘that’ (bluetooth with control app on phone)!!!!!!! 😁

Ah, that gets beyond what I know by some way, was just speculating about what could be done once there's a uP on there. Getting to thermal throttling would be doable (assuming the cpu doesn't just read FID when it comes out of reset). Eventually.
Plus there's the whole 'needlessly complicated' bit for the purpose of just finding a CPU's maximum stable frequency for a given Vcore. It's not something that gets changed a lot, so doesn't really matter if it could be made fancier.

Actually, the major benefit of putting a uP in there is that it allows for the meme of "I heard you like CPUs, so I put a CPU in your CPU".

I’ve actually been quite busy myself as the weather turns warmer here. I have assembled and sold 15 worldwide and have heard no complaints.

I still have 4 fully assembled and ready to go if anyone wants one.

I’ll likely pickup the original passive design later this year.

It's admirable how people spend their free time in this wonderful project. Thanks all. 😀

I have changed the first post to link directly to the necessary files/links to make wiretap's and doorknaat's clone and use them 😀
I also have changed the thread's title slightly making it cleared what it is about.

Feel free to contact me if there's anything wrong with the first post or if you think there's something I should add 😁

Awesome 😀

It might be good to put a small picture of each in that first post so people can see what they look like depending on which link they click.

snufkin wrote on 2021-04-25, 19:43:

Ah, this is firmly in to diminishing returns territory for the possible space savings. Yours is probably as far as is sensible. […]
Show full quote

Ah, this is firmly in to diminishing returns territory for the possible space savings. Yours is probably as far as is sensible.

Although...
Straight connector (I think): https://www.digikey.co.uk/product-detail/en/t … 01-4-ND/1864025
Surface mount dip switch: https://uk.farnell.com/alcoswitch-te-connecti … -smd/dp/2792786

Hey, thanks for helping with looking for alternative components! 😀 I just had a look at them:
Without modifications the connector will probably not fit the CPU PCB:

Using an SMD switch would worsen the issue with the pins on the connector sticking out of the GFD PCB and having to be trimmed down. If you can find the connector in SMD as well I can redesign the PCB to be even smaller for you. But to be perfectly honest I think having all through hole components is an advantage because that way more people will be able to assemble their own device.

Doornkaat wrote on 2021-04-28, 07:26:

Without modifications the connector will probably not fit the CPU PCB:

I thought about that, but then if I was planning to snip pins off to reduce the number of through holes, and trim down the others to stop them sticking right through the board, then at that point sniping those retaining clips off wouldn't be much extra work. The standard connector is obviously far more suitable for just getting and using.

Using an SMD switch would worsen the issue with the pins on the connector sticking out of the GFD PCB and having to be trimmed down. If you can find the connector in SMD as well I can redesign the PCB to be even smaller for you. But to be perfectly honest I think having all through hole components is an advantage because that way more people will be able to assemble their own device.

I had a fairly good look around the AMP site when looking for those, but couldn't find anything surface mount, which was disappointing. Even if I had found one then it's not worth a redesign. If I really wanted to do that then I should make the effort myself.

Is surface mount a problem for soldering? I think it depends on what they are. That dip switch looks fine, with the legs sticking out a fair way. It's when pads are under components, like QFN packages, that life becomes more difficult. Or when things get small. I find down to 0603 is fine, 0402 is doable but starts to take more time and 0201 is a right pain and generally more trouble than they're worth.

I suppose there's an availability problem for some components depending on where you are. Here (UK) we used to have Maplin as somewhere to pop in to to grab a couple of resistors, and they'd be through-hole. So designing with through hole would make things easier. But since they closed down it's on-line ordering, at which point may as well go surface mount.

Back to the goldfinger... bit cheeky, but does anyone in the UK have one of those non-working Northwind2 versions they wouldn't mind posting to me? I had some ideas (posted earlier in this thread) about how to it might be possible to mod them to work, and would quite like to find out if I'm right or not. I can then post it back once I've proven things one way or the other.

If you want partial overlap you're going to get away with snipping off two legs on the connector and trimming down the rest as well as on the switch or mounting the switch not fully inserted/slightly elevated. Also in any case you'll have to solder the switch from the side you're mounting it on after soldering on the connector. It works, it's just a bit unusual. 😉
If you want full overlap you'll have to go SMD on both components because there are required pins on the connector that overlap with pins on the switch. I'm always happy to be proven wrong but believe me, I have tried with the parts in hand and I didn't find a possible position.
As soon as we get into modifying the components or stacking PCBs new possibilities open up but as you already said that takes us into the land of diminishing returns. 😅

SMD soldering takes a bit more skill, better tools and a steadier hand than simple through hole imho. With parts that big it shouldn't be any problem for somebody with a bit of SMD experience - as long as the solder points are accessible from the outside, like you said. Through hole is just more beginner friendly which is why I reckon it's the better solution for an intentionally simple design which people are supposed to be able to assemble themself. 👍

Hey just want to give an update after receiving the goldfinger device from gex85. thank you again for sending me this device.
After purchasing a 750mhz thunderbird cpu i tried to overclock it with the instructions but i didnt had much luck.
The CPU runs only at 900mhz with 1.75 voltage but is very unstable and hangs on the win2k boot process. 1000 and 950mhz is not possible , i dont get a screen signal with these clocks. not at 1.70V, 1.75V or 1.8V

Same with my 800mhz Pluto. but this CPU handles the 900mhz a bit better. But it crashs all the time running 3dmark 2001SE. so maybe there are special serial number cpu´s who work stable with this device? the 700 TB? but i thought there is no difference because i already have a 750TB and it runs normal with stock clock frequence..

my motherboard is the asus K7M rev 1.04 with a updated bios.

Melvyn

Did you take into account the cache divider (for the Pluto, not the Thunderbird)? That's actually VERY important.

Later in the Athlon Pluto/Orion production cycle, AMD was seeing very good yields, so it was not uncommon for them to sell an 800 - 900 MHz CPU die as a cheaper Athlon model, effectively downclocking it.
The only way to know for sure what's under the hood is to remove the metal clips and check the die itself.

I have two 750 MHz Athlons, one of them turned out to be an 800 MHz part, and I've since taken apart the other as well, and that one is awesome: it has a 900 MHz die. It works flawlessly at 900 MHz / 1.65V and it can surely go even higher - however I am still unable to control the cache divider on my Gigabyte GA-7IXE, so anything above 900 MHz is not stable. Since it was sold as a 750 MHz part, the built-in divider is fixed at 2/5. For 1 GHz you need a 1/3 divider, otherwise the cache won't handle it... The divider can be forced either by modifying the CPU PCB, or by modifying the motherboard BIOS to force a custom divider (some motherboards apparently had an easy to use option in the CMOS setup to control the divider, but those are very rare).

In fact, when it comes to stability at 1 GHz, I had better luck with the 800 MHz (sold as 750MHz) CPU (see the link above), because aparently the cache on that CPU is semi-stable at a whopping 400 MHz (way above its rating!).

Does your board offer any control over the cache divider?

As for the Thunderbird, I also tried overclocking a 700 MHz CPU, and the max (frequency/performance) that I was able to achieve was 110FSB x 8 = 880MHz. At 900 MHz it was very unstable. I think there's a better chance of overclocking Plutos than Thunderbirds, since Slot A Thunderbirds were early in the production cycle, so yields were probably not as good.

Hey thanks for informations. but i dont have a cache divider edit function in the bios. i have to check whats inside the cpu.. i thought its a bit easier with the goldfinger device 😁 i watched a youtube video from a guy who overlocked a 750tb to 1ghz without a problem. so i thought thats it 😁

https://www.youtube.com/watch?v=LgUkHjfyaeY&t=274s

There's also always a chance that your CPU just doesn't overclock very well.😐

gex85 wrote on 2021-04-24, 08:13:

I have sent out the GFDs to everyone who ordered one on Thursday. Sorry for the delay again, my spare time is quite limited currently and soldering, testing, packing and posting took quite a while.
I currently have parts for 11 more GFDs that I am willing to sell, but if you need it assembled please expect it to take up to three weeks before I can ship. Unassembled kits can be sent out much faster. PM me if you want one 😀

Just found this post searching for goldfinger devices. Since I am a new user I cant send any pms, wondering if you still have parts for sale?

i have a little update for you all. after purchasing a 900 mhz thunderbird, the overclocking went very well. i have now stable 1ghz with the goldfinger device. i had to increase the voltage from 1.75v to 1.85v.

Awesome, that's great to hear.

Mel427 wrote on 2021-08-26, 11:17:

i have a little update for you all. after purchasing a 900 mhz thunderbird, the overclocking went very well. i have now stable 1ghz with the goldfinger device. i had to increase the voltage from 1.75v to 1.85v.

The voltage is a bit too high, IMO. Make sure you have a very good cooler. 😀 They tend to get very hot at 1.8V and above.
Keep in mind that, as opposed to more modern CPUs, the max safe operating temperature for these early Athlons is 70 degrees.

thank you 😀 i will have an eye on this. but with 1.80v the system was very unstable. just with the 1.85v the system is completely stable with all benchmarks and games. no crashes. i have a dual cooler fan.. i have an flir thermal camera, i checked the temperature and its aorund 50 - 55degrees. so i think i will be safe 😀

Mel427 wrote on 2021-08-26, 18:20:

thank you 😀 i will have an eye on this. but with 1.80v the system was very unstable. just with the 1.85v the system is completely stable with all benchmarks and games. no crashes. i have a dual cooler fan.. i have an flir thermal camera, i checked the temperature and its aorund 50 - 55degrees. so i think i will be safe 😀

My 1GHz Slot A T-Bird also runs fine with the Stock clockrate at 1.64V VCore under Prime 95. You should also check the stability with Prime Blend Test for a few hours, the best stability test. I always use it for testing after undervolting/overclocking. The Blend Test also tests the memory.

thanks will try this, but i thought the stock clock rate for thunderbird cpu´s is 1.75v? http://www.bytes-and-more.de/tech/amd_athlon_tech.htm