VIA's CPU declaration is a mess.
The attachment WhatsApp Image 2026-09-23 at 06.46.072.jpeg is no longer available
The attachment WhatsApp Image 2026-09-23 at 06.46.07.jpeg is no longer available
The first one l0st most of the print on its heatspreader to nothing but regular thermal paste.
It looks noticeably different from the others because it has no surface components, but the bottom reveals the voltage: 1.35V can either be an Ezra or Ezra-T. POST screen does not differentiate, and default speed appears to be 800 MHz (133x6) which could be either.
I have some doubts that I will find out beyond doubt, because some tools cannot even tell the difference between Ezra and Nehemiah. Which is funny because 3DMark '99 can: with a Nehemiah, the benchmark will run fine but the frame counter will constantly count and display zero, giving you a false, abysmally low score even though the benchmark runs smooth.
The second one is the easiest. Says 1200A MHz and 1.45V. Now the 1.45V could either be Ezra-T or Nememiah, and I don't know what the A in 1200A stands for, but one thing is for sure, the Ezra-T never reached 1200 MHz, so this is a classic Nehemiah.
The third and fourth one are messy again - 800 MHz should be either a Samuel 2, Ezra, or Ezra-T. The logo was only used for Ezra and beyond. I bought them as the OEM CPU in Neoware CA2 thin clients. Those were advertised by Neoware themselves as containing Ezra 800 MHz CPUs. So the case should be clear. But it does have the pads and resistors only found on Nehemiah and Nehemiah+. The voltage is a giveaway, though: 1.25V - this can only be a Nehemiah+. And sure enough, although defaulting to 133x6 MHz, they run at 1.2 GHz at default voltage. In fact, I can set them to 1.4 GHz in SETMUL without raising the voltage at all, no issues. So I suspect these were supposed to run at 1200 to begin with.
Since they reach the 1200 MHz with .2V less than the original Nehemiah, they should overclock the best. IIRC though, both of them showed the exact same behavior both in the thin clients as well as on my Asus CUBX: could be set to 133x10.5 via SETMUL, would complete any number of benchmarks, no problem. Would instacrash whenever I set the frequency at 133x11, complete system freeze as soon as I sent the SETMUL command. TBH I don't recall how much voltage I gave them. Guess I'll try that. A good GPU for them is at my disposal, the GF4 Ti4600 can use a very low overhead 30.XX detonator driver and has no problems with the 89 MHz AGP frequency that a 133 MHz FSB does to an Intel 440BX board. Not sure about the cacheable areas, but I should have a few double sided 256MB SDR-133 modules for running 2x256.
Stay tuned!
