Yes this works, some motherboards still have a BIOS option for this.

I found that throttle would work on an Atom based netbook 😀

It all comes down to someone writing a utility for specific CPUs.

P4 K7 and K8? What about Core i3, 5, 7, FX, etc? 😀

Here is a score I noted for a Barton Core Sempron 3000+, which is technically an Athlon XP:

no-L1 cache, speed 12.0x 166= 2000 MHz, SpeedSys score: 33,62 points.

This is between a 486 DX2 and DX4. But what is the point when PCI Sound Blaster emulation is failing under this setting.

A VIA C3 Nehemiah on a i440BX can be slowed from approx. 386SX-20 speed to 486DX-40 speed, by toggling cache and switching FSB speed (50..133MHz).
The Socket 7 Cyrix 6x86L for example, can be slowed to any classic 486 speed, by toggling cache and switching multiplier (1.0x-2.0x-3.0x @ 60MHz FSB).

VIA C3, interesting...

Yeah, it is.
Here are the raw scores:

1SpeedSys Scores for VIA C3 "Nehemiah" C5XL core (2003) and C5P core (2006), Gerwin 2014.
2VIA C5XL no-L1/L2  7.5x  50=  375MHz:      3,77    (cas 3)
3VIA C5XL no-L1/L2  7.5x  50=  375MHz:      3,98    (cas 2)
4VIA C5XL no-L1/L2  4.0x  66=  266MHz:      4,34    (cas 2, no-Branch Prediction makes no difference)
5VIA C5XL no-L1     9.0x 133= 1200MHz:     11,10  
6VIA C5P  no-IC     9.0x 133= 1200MHz:     14,35    
7VIA C5XL no-IC     9.0x 133= 1200MHz:     14,46    (Only "I-cache" disabled, Doom framerate somewhat low, as it should be)
8VIA C5XL no-BP/L2  4.0x  66=  266MHz:    127,30    (PCP-Bench= 19,4 FPS.)
9VIA C5P  no-BP/L2  6.0x  50=  300MHz:    141,50    (not stable. same for 7.0x.)
10VIA C5XL no-BP/L2  7.0x  50=  375MHz:    177,24    (Minimim setting at 50 FSB. PCP-Bench= 24,5 FPS.)
11VIA C5XL no-B.P.   9.0x 133= 1200MHz:    569,54  
12VIA C5XL           9.0x 133= 1200MHz:    793,93    (Quake SVGA=71,5; PCP-Bench=106,4; cas 2)
13VIA C5P            9.0x 133= 1200MHz:    806,02    (Quake SVGA=71,0; PCP-Bench=106,1; cas 2)

Hardware used: GA-6BXC mainboard, Upgradeware Slot-T or MSI MS-6905 master slotket. Voodoo 3 AGP.
Software used: MS-DOS, SetMul for Multiplier and SMBus for FSB. Speedsys, PC Player benchmark at default setting and Quake 1 at 640x480.

leileilol wrote:

P4 K7 and K8? What about Core i3, 5, 7, FX, etc? 😀

That would be neat as well. With their high performance DDR3 IMCs these processors probably wouldn't take nearly as much of a hit as, say, Core 2, which depended on its very large, low-latency L2 to compensate for the lack of an IMC.

P4 would really be interesting. Netburst was extremely sensitive to cache size especially during the Willamette/Northwood days. The 128K Northwood Celeron in particular was a disaster in anything that didn't involve media encoding. The P4's caches were also a fair bit faster than the PIII's at the same frequency. I'd guess that a cache-disabled P4 would be even slower than a cache-disabled PIII--perhaps down to the speed of a 386SX? Has anyone tried this?

gerwin wrote:

Here is a score I noted for a Barton Core Sempron 3000+, which is technically an Athlon XP: […]
Show full quote

Here is a score I noted for a Barton Core Sempron 3000+, which is technically an Athlon XP:

no-L1 cache, speed 12.0x 166= 2000 MHz, SpeedSys score: 33,62 points.

This is between a 486 DX2 and DX4. But what is the point when PCI Sound Blaster emulation is failing under this setting.

A VIA C3 Nehemiah on a i440BX can be slowed from approx. 386SX-20 speed to 486DX-40 speed, by toggling cache and switching FSB speed (50..133MHz).
The Socket 7 Cyrix 6x86L for example, can be slowed to any classic 486 speed, by toggling cache and switching multiplier (1.0x-2.0x-3.0x @ 60MHz FSB).

That's very interesting. Was the L2 also disabled on the Barton system?

Standard Def Steve wrote:

That's very interesting. Was the L2 also disabled on the Barton system?

My old notes do not mention L2 cache at all. They do mention that the BIOS "Internal cache disable" setting remained 10x faster compared to the CACHEOFF.COM score which I stated earlier.
The Sempron system has been disassembled since.

As I said, PCI SB-Live! Sound Blaster emulation will not work properly when L1 cache is disabled. Lacking an ISA slot, the slowed system is basically mute.

Had try to disable caches on northwood. On 12x100 perfomance is about slow 486. 12 is lowest multiplier on p4. So CPU speed on 1200mhz is too fast for such games as test drive 3, wing commander 1. But if you will be able to set CPU on 66mhz bus, these games will run fine. Big half of pll ic on p4 mobos will do 66mhz via hardware strapping. If you don't want to do this, good idea to use dos software that plays with throttling to slow down p4 cpu's

I tried this today with my dual core Xeon...clocked it down a little and it is equivalent of a 386 DX 33

But....... No Sound Blaster sound 😀

I remember using Throttle on a netbook. It then ran Wing Commander almost perfectly.

But serious: Just use DOSBox for such games.

Standard Def Steve wrote:

[...]
Netburst in particular was quite cache-hungry. Would a P4 with both caches disabled be even slower than a 386DX? How about Athlon XP and Athlon 64?

A while ago I tested a P4 Prescott @ 3.2 GHz with both L1 and L2 caches disabled, running on an i875P chipset - it gave around 30+ FPS with 3DBench (could've been faster after tweaking). This corresponds to 486DX- DX2-66 speeds...

EDIT: Sound worked too, because the board also had an ISA slot.