Test your CPU speed (and Sound DMA speed)

[ljbk]

Hi !


Following talks here:
http://www.atari-forum.com/viewtopic.php?f=51&t=24853
i propose to open this topis to allow anyone to test his or her own real hardware with this little utility: HWTST002.PRG.

It tries to measure the CPU clock rate using the MFP clock as reference via 2 methods.
For machines above STE, only one value will be presented at it might be wrong.
The TIME_LP method will fail if any kind of CPU cache is in use or if the main timing loop does not take the expected 24 cycles.
The MFP_VBL and MFPCALC values are related and correspond to a unique method. The MFP_VBL value will only be correct in PAL mode for a machine at 8 MHz. But the MFPCALC value will always be accurate as a correction of the MFP_VBL value will occurr depending on the VBL CYCLES.

If your machine has a Sound DMA, like the STE one, its speed will also be measured.

This version 002 should work in Mono(71 Hz), NTSC(60 Hz) and PAL(50 Hz).

The idea is to obtain a set of results that will correspond to every version of STs that appeared on the market: STs, STFs, STFMs, STEs, Mega ST, Mega STE, STacy ...

For those who wish to participate, please state the model, the TOS version and if possible the serial number (on the back of your machine for STs/STEs).

Mine is like that:

1040STF / 1MB / TOS 1.0 UK / SN: A1 65 4008220
Motherboard: http://www.atari-forum.com/wiki/images/ ... rboard.jpg
Video crystal: 32.02840 MHz => CPU should be 8.0071 MHz

Results:

Mono:
112220 / 8006438 / 8006434 / NO DMA
NTSC / 60 Hz:
133604 / 8006742 / 8006434 / NO DMA
PAL / 50 Hz:
160244 / 8006935 / 8006324 / 8006434 / NO DMA


npomarede machine:
520 STF TOS 1.0 FR (SN A17714012804)

VBL CYCLES: 160244 (160256 / 8 MHz)
MFP_VBL CLOCK: 8010583 or 8010713 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 800997 or 8010101
TIME_LP CLOCK: 8010145 or 8010170
SOUND DMA: TEST NOT POSSIBLE !


Thanks,
Paulo.

PS:
Version 003 now has all numbers valid even in Mono or NTSC.

[wietze]

MEGA ST 1:
serial: A1 95 4000274 MB
TOS 1.02 GB

VBL Cycles: 160244 (160256 (PAL) / 8 MHz)
MFP_VBL CLOCK: 8010322 (IF CPU @ 8 Mhz)
MFPCALC CLOCK: 8009710
TIME_LP CLOCK: 8009801
SOUND DMA: TEST NOT POSSIBLE !

1040 STF:
serial: A1 771 4006085
TOS 1.00 GB

VBL Cycles: 160244 (160256 (PAL) / 8Mhz)
MFP_VBL CLOCK: 8010453 (IF CPU @ 8Mhz)
MFPCALC CLOCK: 8009841
TIME_LP CLOCK: 8009973
SOUND DMA: TEST NOT POSSIBLE !

[Dal]

I think Darklord's STacy will run away with this one...

[ljbk]

I think Darklord's STacy will run away with this one...

Yes, it i should !

Anyway, the prg is prepared for clocks much bigger than 8 MHz.
It was tested correctly both with STeem 3.2 and Hatari at 16 MHz and 32 MHz.

Thanks for reporting your results.
Any STE, Mega STE or Falcon (with or without CT6x) around ?

Paulo.

[DarkLord]

Sorry I've not done this yet, real life and all that. I'm at work right now, but I'm
off tonight.

I'll try it with my various machines when I get back up this evening.

[DarkLord]

Here are some quick results, hope they make sense!

Mega ST4 (TOS 1.04, 4 megs RAM, 16mhz AdSpeed accelerator)
VBL CYCLES: 291428 (160256 (PAL) / 8mhz)
MFP_VBL CLOCK: 9656217 (IF CPU @ 8mhz)
MFP_CALC CLOCK: 17473021
TIME_LP CLOCK: 17474879
SOUND DMA: TEST NOT POSSIBLE!

STACY(TOS 3.06, 4 megs RAM, 40mhz 68030)
VBL CYCLES: 1920068 (160256 (PAL) / 8mhz)
MFP_VBL CLOCK: 8010583 (IF CPU @ 8mhz)
MFP_CALC CLOCK: 95976773
TIME_LP CLOCK: 95976177
SOUND DMA: TEST NOT POSSIBLE!

Falcon (TOS 4.04, 14 megs ST RAM, 256 megs FASTRAM, 95mhz, motherboard mods
VBL CYCLES: TEST NOT DONE !
MFP_VBL CLOCK: TEST NOT DONE !
MFP_CALC CLOCK: TEST NOT DONE !
TIME_LP CLOCK: 9880059
SOUND DMA: TEST NOT POSSIBLE!

On the Falcon, I tried it with the motherboard mods turned off - didn't make a difference,
as far as making all the tests work.

[ljbk]

Here are some quick results, hope they make sense!

Mega ST4 (TOS 1.04, 4 megs RAM, 16mhz AdSpeed accelerator)
VBL CYCLES: 291428 (160256 (PAL) / 8mhz)
MFP_VBL CLOCK: 9656217 (IF CPU @ 8mhz)
MFP_CALC CLOCK: 17473021
TIME_LP CLOCK: 17474879
SOUND DMA: TEST NOT POSSIBLE!

STACY(TOS 3.06, 4 megs RAM, 40mhz 68030)
VBL CYCLES: 1920068 (160256 (PAL) / 8mhz)
MFP_VBL CLOCK: 8010583 (IF CPU @ 8mhz)
MFP_CALC CLOCK: 95976773
TIME_LP CLOCK: 95976177
SOUND DMA: TEST NOT POSSIBLE!

Falcon (TOS 4.04, 14 megs ST RAM, 256 megs FASTRAM, 95mhz, motherboard mods
VBL CYCLES: TEST NOT DONE !
MFP_VBL CLOCK: TEST NOT DONE !
MFP_CALC CLOCK: TEST NOT DONE !
TIME_LP CLOCK: 9880059
SOUND DMA: TEST NOT POSSIBLE!

On the Falcon, I tried it with the motherboard mods turned off - didn't make a difference,
as far as making all the tests work.

Thanks for the tests.

Mega ST4 seems to be a 17.5 MHz one !
STACY seems to have a 8010600 Hz boosted by a 68030 that tricks the 24 cycles loop as expected !
As for the Falcon, i expected the 68030 to fool the 24 cycles loop but i am a bit surprised that you get "SOUND DMA: TEST NOT POSSIBLE!". May be there is a bug with the test ...

Paulo.

PS: Yes indeed, there was a stupid bug with a badly located branch label ...
Sorry about that !
Version 2A solves this for all machines above STE that will now test the Sound DMA !

[DarkLord]

Odd about the Mega ST - could the onboard cache on the AdSpeed make a difference?

I'll run it again.

Tied up right now, but I downloaded the new version and will try it and report back
later tonight.

Thanks!

[ljbk]

Odd about the Mega ST - could the onboard cache on the AdSpeed make a difference?

I'll run it again.

Tied up right now, but I downloaded the new version and will try it and report back
later tonight.

Thanks!

Hi !

Did you run the test on the Mega ST in NTSC / 60 Hz ?

The first number (VBL Cycles) and the last one (TIME_LP CLOCK) that appears will be influenced by the pure internal CPU speed as only internal registers and a condicional branch are used and code fetch from memory.
If the planned 24 cycles test loop is done in less than (24 / 8 MHz) of a second then the number shown will go up as much.
The 3rd number is directly linked to the first one so it will be influenced as well by a fast CPU.
The 2nd number is a relative speed number. This is why for STacy, you get the video clock / 4, that should be the 68000 clock. Your Shifter crystal should be a 32.0424 MHz one.
Again for STacy, the first test if it was run in PAL shows that your CPU is 1920068/160256 = aprox 12 times faster than the 68000 at 8 MHz !!! So its assumed clock would be around 8 * 12 = 96 MHz ! That is consistent with the reported numbers.
Regarding the Falcon and TTs, i assume that the video modes are quite different from the base ST so i skip the tests related to the VBL and so it displays: TEST NOT DONE !. The number left, (TIME_LP CLOCK) will be directly dependent on the real internal CPU speed, including the code fetch from memory. May be that is why you get only those 10 MHz. It is also possible that the MFP has something different in the Falcon, but i don't think so.

Finally for the MEGA ST, we get 291428 VBL cycles. So this either means a faster CPU, a faster CPU and a shorter VBL, or a longer VBL and that is not likely. If the test was done in PAL, then your CPU speed was 291428/160256 = around 1.82 times faster than the santard 68000 at 8 MHz. If it was done in NTSC, then it would be 291428 / 133604 = 2.18 times faster. If it was done in Mono, then it would be 291428 / 112224 = 2.60 times faster. But if the CPU works at 16 MHz, then the VBL can be 291428 / (2*160256) = 0.909 times shorter compared to PAL => around 55 Hz !
For the test involving only the MFP, you get a relative speed of 9.656 MHz. That is consistentent with the 1.82 value for PAL as 9.656 x 1.82 gives the 17.5 MHz value obtained by the TIME_LP CLOCK.
Do you have a special video mode at 55 Hz for this machine ?


Anyway, you can now run version 003 that has been improved for NTSC and Mono cases and that show the real number of VBL cycles considered for the test depending on the video mode.


Paulo.

[DarkLord]

Yes, all my tests were done in NTSC mode (I'm in North America, USA).

As far as I know, the Mega ST4 doesn't have anything video running at 55hz.

Here are the new results, using version 3 of your software:

STACY(TOS 3.06, 4 megs RAM, 40mhz 68030)
VBL CYCLES: 1920068 (160256 (PAL) / 8mhz)
MFP_VBL CLOCK: 8010583 (IF CPU @ 8mhz)
MFP_CALC CLOCK: 95976773
TIME_LP CLOCK: 95976349
SOUND DMA: TEST NOT POSSIBLE!

Falcon (TOS 4.04, 14 megs ST RAM, 256 megs FASTRAM, 95mhz, motherboard mods
VBL CYCLES: TEST NOT DONE !
MFP_VBL CLOCK: TEST NOT DONE !
MFP_CALC CLOCK: TEST NOT DONE !
TIME_LP CLOCK: 9874751
SOUND DMA: 49172

Mega ST4 (TOS 1.04, 4 megs RAM, 16mhz AdSpeed accelerator)
VBL CYCLES: 291452 (133604 (PAL) / 8mhz)
MFP_VBL CLOCK: 8010451 (IF CPU @ 8mhz)
MFP_CALC CLOCK: 17474458
TIME_LP CLOCK: 17474879
SOUND DMA: TEST NOT POSSIBLE!

Mega ST4 (only difference, running at 8mhz)
VBL CYCLES: 133604 (133604 (PAL) / 8mhz)
MFP_VBL CLOCK: 8010295 (IF CPU @ 8mhz)
MFP_CALC CLOCK: 8010295
TIME_LP CLOCK: 8009850
SOUND DMA: TEST NOT POSSIBLE!

Hope this is better and not more confusing!

[ljbk]

Yes, all my tests were done in NTSC mode (I'm in North America, USA).

As far as I know, the Mega ST4 doesn't have anything video running at 55hz.

Here are the new results, using version 3 of your software:

STACY(TOS 3.06, 4 megs RAM, 40mhz 68030)
VBL CYCLES: 1920068 (160256 (PAL) / 8mhz)
MFP_VBL CLOCK: 8010583 (IF CPU @ 8mhz)
MFP_CALC CLOCK: 95976773
TIME_LP CLOCK: 95976349
SOUND DMA: TEST NOT POSSIBLE!

Falcon (TOS 4.04, 14 megs ST RAM, 256 megs FASTRAM, 95mhz, motherboard mods
VBL CYCLES: TEST NOT DONE !
MFP_VBL CLOCK: TEST NOT DONE !
MFP_CALC CLOCK: TEST NOT DONE !
TIME_LP CLOCK: 9874751
SOUND DMA: 49172

Mega ST4 (TOS 1.04, 4 megs RAM, 16mhz AdSpeed accelerator)
VBL CYCLES: 291452 (133604 (PAL) / 8mhz)
MFP_VBL CLOCK: 8010451 (IF CPU @ 8mhz)
MFP_CALC CLOCK: 17474458
TIME_LP CLOCK: 17474879
SOUND DMA: TEST NOT POSSIBLE!

Mega ST4 (only difference, running at 8mhz)
VBL CYCLES: 133604 (133604 (PAL) / 8mhz)
MFP_VBL CLOCK: 8010295 (IF CPU @ 8mhz)
MFP_CALC CLOCK: 8010295
TIME_LP CLOCK: 8009850
SOUND DMA: TEST NOT POSSIBLE!

Hope this is better and not more confusing!

Hi !

Thanks for your tests !
I forgot to remove the (PAL) stuff...

Tests results are crystal clear:

Falcon:
- Sound DMA speed is below STE DMA speed: that is correct;
- 9.87 MHz and 9.88 MHz before: CPU is faster but prefetch memory access is not as fast;

STacy:
- Video clock is 32.0424 MHz (you can check the crystal) => 68000 clock is 8.0106 MHz;
- the 030 board boosts the speed almost 12 times, for the used code, so for the program it is like a 68000 at 96 MHz;

MegaST @ 8 MHz:
- CPU clock is also around 8.010 MHz like STacy so the video crystal should also be a 32.0424 MHz one;

MegaST with Adspeed:
- The number of VBL cycles goes from 133604 to 291452 => 2.18 times faster: this can happen because the clock is faster and/or because the number of cycles one instruction takes is lower => anyway the effect is increasing speed by 2.18 times;
- So the aparent speed compared to the 68000 base is then 17.47 MHz even if the detected video crystal, or is you want the VBL speed, is aprox. 4 times 8.010451 MHz;

For Hextracker the only important number is MFP_VBL CLOCK and only for PAL modes because then it is important to know how many VBLs there are per second. The clock ranges until now go from 8.007100 MHz to 8.021247 MHz. That 0.2 % difference can mean, at high replay frequency, 1 or 2 MFP interrupts of difference inside the VBL that need calcs or not or they can mean some cracks in the sound.

I manage to increase the precision of the calculations in the now available version 004.

Thanks again,
Paulo.

PS: I noticed you are in the US, but you could also run a test in PAL like you did for STacy.

[ljbk]

Any STE or 8 MHz /16 MHz Mega STE around ?

[DarkLord]

I know there are lots of STe and Mega STe users on this board, someone gonna
step up and help?

Don't make me drag my Mega STe out of the closet.

[wietze]

Any STE or 8 MHz /16 MHz Mega STE around ?

I have an Atari STE (520, but with 4mb ram); which I will benchmark whenI get home today.

[Jungsi]

1040 STE - 4MB:
VBL CYCLES: 160258 (160256 @ 8 MHz)
MFP_VBL CLOCK: 8020732 (IF CPU @ 8 MHZ)
MFPCALC CLOCK: 8020732
TIME_LP CLOCK: 8020811
SOUND DMA: 50065

[ljbk]

Thanks for posting your results.

DMA Sound speed is ok.
VBL clock cycles is really close to what is expected.
CPU clock is around 8.0208 MHz which means that video clock is 32.0832 MHz.
That is close to the 8.021248 MHz emulators are using but not the same.
Here are the values obtained using Hatari with version 4:

8 MHz/PAL:
==========
VBL CYCLES: 160256 (160256 @ 8 MHz)
MFP_VBL CLOCK: 8021225 (IF CPU @ 8 MHz)
MFPCACL CLOCK: 8021225
TIME_LP CLOCK: 8021254
SOUND DMA: 50068

16 MHz/PAL:
===========
VBL CYCLES: 320514 (160256 @ 8 MHz)
MFP_VBL CLOCK: 8021347 (IF CPU @ 8 MHz)
MFPCACL CLOCK: 16042694
TIME_LP CLOCK: 16042491
SOUND DMA: 50066

32 MHz/PAL:
===========
VBL CYCLES: 641026 (160256 @ 8 MHz)
MFP_VBL CLOCK: 8021347 (IF CPU @ 8 MHz)
MFPCACL CLOCK: 32085388
TIME_LP CLOCK: 32084992
SOUND DMA: 50063

Can any one else benchmark any STE or Mega STE ?
The more results are posted, the more the community will know about the different crystals used by the Atari STs.

Paulo.

[devo]

520 STe + 4mb RAM
serial nr: A1 9B 3013779
TOS 1.06 (from SysInfo), HiRes/mono, PAL

VBL CYCLES: 112228 (112224 @ 8MHz)
MFP_VBL CLOCK: 8020835 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 8021079
TIME_LP CLOCK: 8021032
SOUND DMA: 50066

Note: my STe has the faulty DMA chip.

[wietze]

520 STe + 4mb ram:
serial: A1 11A 3048045

TOS: 1.06 GB
VBL CYCLES: 160244 (160256(PAL) / 8MHz)
MFP_VBL CLOCK: 8020632 (IF CPU @ 8MH)
MFPCALC CLOCK: 8020020
TIME_LP CLOCK: 8020246
SOUND DMA: 50065

[bullis1]

I downloaded the test prog a couple days ago but I've been away for the last few days and won't be able to test on MSTE until Tuesday probably. Hold tight! It's an NTSC machine btw. If you don't hear from me feel free to pester me via PM

[ljbk]

Hi !


Thanks for the result posting.

My request for more results is only informative for the forum community.

The upcoming Hextracker 0.849 already has this feature included and running.
It accepts MFP_VBL CLOCK from 8000000 to 8100000. If this RELATIVE value is outside the range then 8021247 is assumed.
This will ONLY have impact for ST(F) or STE machines running in PAL. Above STE machines, or machines running in Mono or NTSC will continue to have their 50 Hz frame controled by a timer that only depends on the MFP and not on the CPU.
Here is one example:
STE mode with DMA speed of 50066 Hz
Simplistic calc: 50066 Hz / 50 Hz = 1001.32
Mono: timer freq = 2457600 / 49100 = 50.052953 Hz => number of updates per frame = 1000.26
NTSC: timer freq = 2457600 / 49100 = 50.052953 Hz => number of updates per frame = 1000.26
PAL:
- case 1 CPU freq = 8021247 => frame freq = 8021247 / 160256 = 50.052709 => updates per frame = 1000.27
- case 2 CPU freq = 8021000 => frame freq = 8021000 / 160256 = 50.051168 => updates per frame = 1000.30
- case 3 CPU freq = 8010600 => frame freq = 8010600 / 160256 = 49.986272 => updates per frame = 1001.59
- case 4 CPU freq = 8007100 => frame freq = 8007100 / 160256 = 49.964432 => updates per frame = 1002.03

There are also impacts, of the same order, on the sample read speeds for every note:
C2 = Amiga PAL(7093790) / 856 = 8287.13785 Hz as read speed
With 50.00 Hz frame the read speed is: 8287.13785 / 50.00 = 165.742757 bytes per frame
=> simplistic calc: 8287 Hz / 50 Hz = 165.74 bytes per frame
=> timer freq: 50.052953 => 165.567411 bytes per frame
=> case 1 CPU freq = 8021247 => 165.568217 bytes per frame
=> case 2 CPU freq = 8021000 => 165.573315 bytes per frame
=> case 3 CPU freq = 8010600 => 165.788276 bytes per frame
=> case 4 CPU freq = 8007100 => 165.860744 bytes per frame

With this hardware test version 4:
- VBL CYCLES precision is around +- 2 cycles @ 8 MHz for 68000 CPUs on "normal" machines;
- MFP_VBL CLOCK precision is around +- 250 Hz;
- MFPCALC CLOCK depends on the two above;
- TIME_LP CLOCK precision is around +- 40 Hz for 68000 CPUs on "normal" machines;
- DMA speed precision is +- 4 Hz;


Paulo.

[Nikolas]

soon i will test my stfm

[MasterOfGizmo]

MIST board in Monochrome mode, TG68K CPU:

VBL CYCLES: 132556 (112224 @ 8 MHz)
MFP_VBL CLOCK: 1012216 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 11956860
TIME_LP CLOCK: 9612648
SOUNDS DMA: TEST NOT POSSIBLE !

[Cyprian]

My results:

STE pass1:

VBL CYCLES: 160256 (160256 @ 8 MHz)
MFP_VBL CLOCK: 8020978 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 8020978
TIME_LP CLOCK: 8021032
SOUNDS DMA: 50068


STE pass2:

VBL CYCLES: 160256 (160256 @ 8 MHz)
MFP_VBL CLOCK: 8020978 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 8020978
TIME_LP CLOCK: 8021057
SOUNDS DMA: 50067



Mega STE pass1:

VBL CYCLES: 160256 (160256 @ 8 MHz)
MFP_VBL CLOCK: 8020978 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 8020978
TIME_LP CLOCK: 8021155
SOUNDS DMA: 50065


Mega STE pass2:

VBL CYCLES: 160256 (160256 @ 8 MHz)
MFP_VBL CLOCK: 8020978 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 8020978
TIME_LP CLOCK: 8021180
SOUNDS DMA: 50067

[bullis1]

Looks like I forgot to test! Thanks to Troed for jogging my memory. I apologize for the information overload but I had to test some special cases. If you want more info (like warm boots/reset button boots, etc.) let me know. Here goes:

1st Pass, 16Mhz cache ON
VBL CYCLES: 291478 (133604 @ 8 MHz)
MFP_VBL CLOCK: 8054082 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 17571235
TIME_LP CLOCK: 17570578
SOUNDS DMA: 50067

2nd Pass, 16Mhz cache ON after full reboot
VBL CYCLES: 291476 (133604 @ 8 MHz)
MFP_VBL CLOCK: 8054082 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 17571112
TIME_LP CLOCK: 17570578
SOUNDS DMA: 50067

3rd Pass, 16Mhz cache OFF
VBL CYCLES: 133616 (133604 @ 8 MHz)
MFP_VBL CLOCK: 8053933 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 8054547
TIME_LP CLOCK: 8054235
SOUNDS DMA: 50065

4th Pass, 8Mhz cache OFF
VBL CYCLES: 133606 (133604 @ 8 MHz)
MFP_VBL CLOCK: 8208785 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 8208785
TIME_LP CLOCK: 8054038
SOUNDS DMA: 50066

5th Pass, 8Mhz cache OFF, screen set to 50hz
VBL CYCLES: 133606 (133604 @ 8 MHz)
MFP_VBL CLOCK: 8053933 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 8053933
TIME_LP CLOCK: 8054038
SOUNDS DMA: 50067

6th Pass, 8Mhz cache OFF after full reboot, screen set to 50hz
VBL CYCLES: 133606 (133604 @ 8 MHz)
MFP_VBL CLOCK: 8053784 (IF CPU @ 8 MHz)
MFPCALC CLOCK: 8053784
TIME_LP CLOCK: 8054038
SOUNDS DMA: 50067

[AtariZoll]

Mega STE PAL TOS 2.06 UK. Org. 2.05 German. Machine was sold in Germany

VBL CYCLES 160268 (160256 @ 8 mhZ)
MFP_VBL CLOCK 8020763 (IF CPU @ 8 MHz)
MFPCALC CLOCK 8021252
TIME_LP CLOCK 8020344
SOUND DMA 50065

At 16 MHz:
VBL_CYCLES: 349604 (160256 @ 8 MHz)
MFP_VBL CLOCK 8021025 ...
MFPCALC CLOCK 17498137
TIME_LP CLOCK 17497710
SOUND DMA 50067

Well, I don't want to be negative, but there are some strange values with Mega STE . Especially in tests made by Bullis1 . - May be that SW he uses sets not correctly the cache ?
In tests on my machine I see strange MFPCALC CLOCK at 16MHz - it can't be more than 2x as by 8MHz . Or I don't get this test.
OK, you said that it will be wrong with cache ...Is it just because long time period when outside of counting - even first test shows more than 2x cycles at 16MHz, but then calculated value is same ?
I made proggy for testing pure CPU clock on ST, STE, using TImer C. Test takes only 30mS (6 periods at 200Hz), and I getting very close 2x result with 16MHz on MSTE in compare with at 8Mhz. Code is short, so it executes completely in cache. No bra, dbf , what maybe takes 2T states less when running from cache , so it can not affect speed.

As I see, all ST, STE machines have separated crystal for MFP, and always with same freq. What is nice, and it may serve for accurate measuring. CPU and video + many other chips getting clock deliverer from main clock, close to 32 MHz, but with slight differences, to avoid interference by color encoding - therefore diff of some 0.4% between PAL and NTSC (at least in case of STE) . DMA audio clock is delivered too from main - therefore strange sample rates. 3rd crystal in machines is for color encoder, if it is stuffed. And fourth is at IKBD chip. PAL video is always with 313 lines, etc (to not repeat here known values).

Done tests with v3 .