Probably DOS Quake. Just wondering if the accelerated graphics cards are making the L2 cache less important.

Hello Feiopa and Gona.I would like to ask feiopa if he is running his tillamook with a 3,5 or 4 multiplyer and if it is possible to run it on a 83mhz fsb.4*83 equals 333mhz in our old non k6-2 core motherboards.
Gona can you give the forum some sysspeed and glquake benchmarks for 4x75hmz=300mhz and 4x83mhz=333mhz?

amijim, I am not currently running a Tillamook or a Pentium MMX in any system. The Tillamook didn't work in the 430TX board I tested it in many years ago. My 430TX system currently contains a AMD K6-III+ running at 6x83 = 500 MHz. It has 4x64 MB EDO SIMMs and a GF4MX, running W98SE, NT4, W2K, and XPpro. The PCI cards seem OK with the 41 MHz PCI bus. The challenge was getting it to work with fast timings and 256 MB of RAM with the 83 Mhz FSB. SDRAM could only handle 83 MHz if the RAM is limted to 64 MB. SDRAM could handle 256 MB, but only if using a 66 MHz FSB. For 83 MHz and 256 MB, I had to use 72-pin EDO SIMMs. I'll have to retest the Tillamook at some point.

amijim wrote:

Gona can you give the forum some sysspeed and glquake benchmarks for 4x75hmz=300mhz and 4x83mhz=333mhz?

I have tried with TX97-XE but none of them can start the system.

feipoa wrote:

Probably DOS Quake. Just wondering if the accelerated graphics cards are making the L2 cache less important.

You were true, the accelerated graphics is diminish the advantage that can give the L2 cache. Here they are DOS Quake 1.06 640x480 results:

Looks like with accelerated graphics, the use of L2 cache provides around 1.5% performance increase, while without accelerated graphics (software mode) the use of L2 cache provides about a 5.8% performance increase. Makes me wonder if L2 cache provided a better enhancement on older boards, e.g. 386, 486, and socket 5. 5.8% seems like such a small increase for the cost of cache back then.

Yes, these results show that without accelerated graphics the use of L2 cache provides 5.6 - 6.1% performance increase. Likely on slower CPU based systems the motherboard cache gives better performance increase as on faster systems the motherboard cache gives smaller accereration. Might this was the main reason that CPU manufacturers has moved the L2 cache from motherboards to the CPU package/PCB (off-die) and later inside to the CPU core (on-die).
With same CPU and cache parameters 430TX (TX97-XE) is faster than VIA VPX (M537-DMA33), but with Tillamook, TX97-XE cannot work with L2 (disables it) and PC Chips M537-DMA33 with working L2 cache is faster than ASUS TX97-XE (if the graphics is not accelerated).
I'm going to continue Tillamook benchmarking on the late Socket 7 chipsets by DOS Quake instead of GLQuake.

Thank you Gona.I think we can say that the myth is busted!

I think it would be interesting to see the benefits of L2 on different CPUs as well. I remember looking long and hard for a quality Tyan MVP3 board with 2 MB of L2 cache for a Cyrix MII-433 build. That amount of cache must have cost a lot for only 1% - 6% increase. At the same time, it makes cacheless Cyrix Media GXm boards not look all that bad

Got my DFI P5BV3+ working with this CPU!

Simply added bridges between ADS#/ADSC# and BRDY#/BRDYC#

Cache is on, CPU is working fine at 300MHz, see below for results.

266MHz (75x3.5):

300MHz (100x3):

Wires:

Bonus footage:

Picture of a Tillamook in BGA-case, which i reballed, just in case:

System seems also stable at 350MHz, getting 70fps in Quake timedemo. 😉

One could also mod the CPU and not the board, because the MMX has trough-hole pins. That's the next thing i'll do to keep my MoBo stock and test the CPU in other boards.

Wow! Very nice work! Did it boot beTiffani Amber Thiessenfore Tiffani Amber Thiessenthe mod (without cache working) or did it not post at all before the mod?

And could you make clear which pins to connect on the UPPER side of the cpu? As i can not see too clearly which pins you connected.

THANX! I want to try it also, beeing able to use a super socket 7 board with this cpu would be awesome.

It booted before, but only when i disabled the cache completely. Otherwhise, it got stuck after OPROM-scan.
Just look at the picture of the solder joints. Bottom/left is the key side of the cpu.
You just need to rotate the cpu in your head. But i'll do some pictures after i modded the cpu itself!

Nice! Any chance this mod will make the CPU/cache work in a 430TX board?

meljor wrote:

Did it boot beTiffani Amber Thiessenfore Tiffani Amber Thiessenthe mod

I gotta say, that's one weird-ass auto-correct/auto-suggest for "before the". 😳

Makes me wonder if L2 cache provided a better enhancement on older boards

Quite a hefty improvement (0kb vs 512kb).
https://www.youtube.com/watch?v=FQ6WtQMFCSM

Most likely fast SDRAM is the main reason why L2 cache absence is not so much noticeable on 430TX.

LeFlash wrote:

System seems also stable at 350MHz, getting 70fps in Quake timedemo. 😉

One could also mod the CPU and not the board, because the MMX has trough-hole pins. That's the next thing i'll do to keep my MoBo stock and test the CPU in other boards.

I have a 300mhz Tillamook (SL34N), would this trick also enable this cpu to work correctly with cache? 😲

In theory - yes.
The Tillamook PGA are alle the same, at least the datasheet says that.

BTW: I moved the mod from the MoBo to the CPU itself. Works! Unmodded CPU does not boot, modded one does - last test successful!

feipoa wrote:

Nice! Any chance this mod will make the CPU/cache work in a 430TX board?

I'm waiting for the answer to this one too.

Anonymous Coward wrote:

feipoa wrote:

Nice! Any chance this mod will make the CPU/cache work in a 430TX board?

I'm waiting for the answer to this one too.

Give me a board and i'll report 😉