The FrankenTandy 1000 SL: fastest XT in the world?

[CarlosTex]

Hi there.

I present to you some screens of my Tandy 1000 SL running at a staggering 14MHz:

Games that are heavy for an XT run very well:

TOPBENCH scores:

I'm still left to test higher OSC frequencies. I think the V30HL CPU will support higher than its rated 16MHz operation but that is yet to be proven. I still want to try a 48MHz OSCI at least. I have to check if i have a 50MHz OSCI around, it's just 0.6MHz overclock.


Could this be the fastest XT class system without using a 286, 386 or 486 upgrades?

 

[soviet9922]

can you post how you overclocked it ?, i have an sl and upgraded to a nec v30.

 

[CarlosTex]

can you post how you overclocked it ?, i have an sl and upgraded to a nec v30.

It involves cutting a trace that goes from the 24MHZ OSCI to the buffer blue chip, and then soldering some wires to a OSCi socket and patch the output from the socket to the cut trace. This way the buffer blue chip receives the clock signal i desire.

If you want to go crazy like me you will need some soldering skills and of course know what you're doing. Looking at the SL tech manual helps too. If you are not comfortable doing these things i advise against it.

 

[Mr. Horse]

14MHz XT class system? I believe I seen a V30 at 16mhz, but I could be wrong. Still 14mhz is a good OC and probably up there with the best.

 

[soviet9922]

I have no problem doing it, can you elaborate on the details of what trace have to be cut and how you wired up this ?.
You used a can oscillator or a crystal ?, also how you keep the rest of the machine at its original speed ?.

 

[Rick Ethridge]

It beats the hell out of my V30/87 VGA computer. I would love to stock more ram on her!

 

[Scali]

It sure is fast, and shows what 8086/v30 are capable of, compared to the more common 286 16-bit machines.
I only wonder if an 8086/v30 can be called XT-class.
At the very least it has 16-bit memory instead of the 8-bit memory found in a real XT. Are there also 8086/v30 machines that have 16-bit ISA slots?

 

[CarlosTex]

It sure is fast, and shows what 8086/v30 are capable of, compared to the more common 286 16-bit machines.
I only wonder if an 8086/v30 can be called XT-class.
At the very least it has 16-bit memory instead of the 8-bit memory found in a real XT.

Well IIRC the Tandy 1000 SL has always been considered XT class. It still can only address 1MB memory, 8bit slots, no RTC, no High Density floppy drives...

The Tandy 1000 TL could be considered a XT-286 class machine, like the IBM 5162. It does have a 286 CPU an RTC and integrated XT-IDE (old) interface. It only has 8bit ISA slots though. I think it can also address more than 1MB although i'm not sure, never dived too much on the TL info.

Are there also 8086/v30 machines that have 16-bit ISA slots?

Of all 8086/V30 motherboards i have ever seen, OEM or not they all had 8bit slots, or risers with 8bit slots.



I'm still going to try to push it a little more. I ordered 50 and 54MHz wchich will give 16.6 and 18MHz CPU clockspeeds respectively. I think 18MHz will be too much but i'm hoping that it works at 16.6. If it does i'm predicting that Checkit Dhrystones score will rise up to about 2100, which puts it close into fast 10MHz 286 territory, which is impressive to say the least.

It is very subjective to compare, because i've seen 10MHz 286's with appaling results and others that are 30% faster.


Does nyone around here have both IBM 5170 AT revisions to compare? I believe my Tandy will beat the crap out of the original 6MHz IBM AT, but i'm not sure how it compares against the 8MHz 5170

 

[CarlosTex]

I have no problem doing it, can you elaborate on the details of what trace have to be cut and how you wired up this ?.
You used a can oscillator or a crystal ?, also how you keep the rest of the machine at its original speed ?.

Read this thread:

http://www.vcfed.org/forum/showthread.php?56336-Tandy-1000-SL-overclocking

There's a picture on page 2 that shows the trace going under a ceramic capacitor that ends connects to R24. The datasheet also confirms this. That is the trace that leads to the OSCIN pin on the buffer blue IC.

 

[vwestlife]

It sure is fast, and shows what 8086/v30 are capable of, compared to the more common 286 16-bit machines.
I only wonder if an 8086/v30 can be called XT-class.
At the very least it has 16-bit memory instead of the 8-bit memory found in a real XT. Are there also 8086/v30 machines that have 16-bit ISA slots?

Even all of the 80286-based Tandy 1000s were still XT-class machines: they had an XT keyboard interface, 8-bit ISA slots, no high IRQs, no high DMA, and the "SmartDrive" IDE-XT hard drive interface.

The only Tandy 1000 that is AT-class and has 16-bit ISA slots is the 386SX-based 1000RSX.

 

[Scali]

Well IIRC the Tandy 1000 SL has always been considered XT class. It still can only address 1MB memory, 8bit slots, no RTC, no High Density floppy drives...

But there are XT clones that have an RTC and high density floppy drives, so my question is: where does one draw the line, really?
Is it the CPU? Virtually all XT clones use an 8088 (or the V20 replacement, which can also be used in a real PC or XT).

The Tandy 1000 TL could be considered a XT-286 class machine, like the IBM 5162.

The 5162 is an AT-class system though. It is functionally equivalent to the AT, has all the same hardware features, same type of CPU, same BIOS API etc.
So I see no difference between an 'XT-286' and 'AT' system.
They only called it the 'XT' because it was in an old XT case, basically. And I wonder why that was. Did IBM have a huge overstock of these old cases + PSUs to warrant that model?

With the 8086 one can at least debate that although the bus is 16-bit instead of 8-bit, it is functionally indistinguishable from an 8088.
The V30 however adds the 186 instructions to the mix, which effectively allows it to run certain software aimed at 286 (basically anything that uses 286 instructions without protected mode).

So I'm not really sure if there is any one single definition of 'PC class', 'XT class', 'AT class' etc.
I'd say the Tandy 1000 SL is somewhere in the gray area between XT and AT.

 

[CarlosTex]

BTW, the history of my overclocks is with original V30 CPU:

8MHz -- scored about 1023 Dhrystones


Changed all the original 150ns DRAM to 80ns :

9.54MHz - scores 1195 Dhrystones -- not stable

9MHz - scores 1120 -- stable

Changed the CPU to a V30HL:

10MHz

12 MHz

14MHz

The scaling with just ~2MHz increases is nothing short of impressive if you ask me. V30HL's are true speed demons.

I feel really excited to try a 50MHz OSCI to get 16.6Mhz. By the scaling this should get ~2100 Dhrystones.

This way i'll get a super fast Tandy 1000, and by activating mode slow i will have the original 8Mhz speed. Well, just 0.3MHz faster which won't be noticeable.

 

[vwestlife]

So I'm not really sure if there is any one single definition of 'PC class', 'XT class', 'AT class' etc.
I'd say the Tandy 1000 SL is somewhere in the gray area between XT and AT.

IMHO, if the CPU doesn't have protected mode -- the most defining characteristic of the IBM AT -- then it's not "AT-class". It's just a fast XT.

 

[Scali]

IMHO, if the CPU doesn't have protected mode -- the most defining characteristic of the IBM AT -- then it's not "AT-class". It's just a fast XT.

Why would you say that protected mode is the most defining characteristic of the IBM AT?
I would say it is the least defining characteristic, since 286 protected mode was very limited and somewhat broken. Hardly any software ever made use of it.
I think pretty much everything else that the AT introduced was more 'definining' than that. Think of 16-bit ISA slots, the AT form-factor for PSU and case, IDE (aka AT-BUS) harddisk interface, second DMA controller, second interrupt controller, CMOS/RTC.

 

[vwestlife]

Why would you say that protected mode is the most defining characteristic of the IBM AT?

Because the biggest promise of the AT -- even if it was never satisfactorily fulfilled -- was the ability to "break the 640K barrier" and do multitasking.

If it didn't have that, it'd just be a faster, more expensive XT in a bigger case.

 

[Scali]

Because the biggest promise of the AT -- even if it was never satisfactorily fulfilled -- was the ability to "break the 640K barrier" and do multitasking.

Multitasking can also be done on an 8088, and there was some software available for that.
And of course EMS already broke the 640k barrier as well.
And oh, the irony that the AT mainboard only takes 512k max, as opposed to the 640k of the later XTs

If it didn't have that, it'd just be a faster, more expensive XT in a bigger case.

That is how it ended up spending its life.
I see the AT/286 as a failed experiment, with the 386 getting protected mode right, and arriving in time for people to break the 1 MB barrier and starting to think about things like multitasking and multimedia.

 

[Trixter]

Does nyone around here have both IBM 5170 AT revisions to compare? I believe my Tandy will beat the crap out of the original 6MHz IBM AT, but i'm not sure how it compares against the 8MHz 5170

TOPBENCH, which you show in one of your screenshots, has a compare function built into the program, it's in the menus. And both of the ATs and the 5162 are in the database.

 

[CarlosTex]

I don't know the system limits yet, but i tried a 66MHz OSCI i found and the beeper protested furiously. So there's not gonna be a 22MHz Tandy.

TOPBENCH reports that my system is:
2.09x faster than the original IBM AT,
1.77x faster than the 8MHz AT,
1.28x faster than the IBM 5162,
1.53x faster than a Tandy 1000 TL/TX,
1.10x faster than the SEGA TeraDrive,
1.35x faster than IBM PS/1 Model 2011,
1.05x faster than a Compaq DeskPro 286.

I'm hoping to get the 16.6MHz speed, so i will still beat some 286 systems that are above mine.

 

[Trixter]

Make sure you submit your results, as I think that's the fastest XT I've ever seen verified!

 

[soviet9922]

Read this thread:

http://www.vcfed.org/forum/showthread.php?56336-Tandy-1000-SL-overclocking

There's a picture on page 2 that shows the trace going under a ceramic capacitor that ends connects to R24. The datasheet also confirms this. That is the trace that leads to the OSCIN pin on the buffer blue IC.

yes i'm reading the thread see the picture but is hard to tell what you ended up using to reach the 14 mhz.

I see you when back and fort and then removed the patch and then added it again and replaced the ram.
Don't whan't to take your time Carlos but can you elaborate on what is the final modification required to reach this speed ?.
Also what oscillator you used externally a can one and what speed how you wired it up ?