j^aws wrote on 2022-02-20, 20:37:
I did these tests years ago, and the conclusion for 'best' slow platform with the largest CPU range without using tools like Mos […]
Show full quote
I did these tests years ago, and the conclusion for 'best' slow platform with the largest CPU range without using tools like Moslow and Throttle:
Most flexible from slowest:
1) K6III+ with a S7 supporting appropriate and active Turbo switch (not a SS7)
2) Slot 1/ S370 with an EZRA-T overclocked to 1200+ MHz
3) S7 and K6III+ with board that can change write thru from write back cache for L1 (an extra parameter)
After this, you can use a second board with an unlocked Core2Duo S775 that scales from 600MHz to 3000+MHz, covering 3D games.
Or an unlocked PIII Tualeron/ Tualatin instead of the above, from 200 MHz to 1400+ MHz. Or an AMD Athlon flavour.
A) Nehemiah and Slot1 / S370
B) SS7 and K6III+
C) Unlocked PII and Slot 1, with ability to switch write thru from write back L1 cache (extra parameter)
Thank you, @j^aws. Yes, that's basically my experience as well.
Anyway, this is the main idea of this thread: I'm trying to prove that, no matter how one looks at it (from a flexibility/performance/stability standpoint, etc), the Ezra-T + 440BX is superior to a SS7 build.
That does not mean that SS7 builds are bad, far from it. As I said, I also love this platform and have a bit too many SS7 systems myself. However, having used both the Ezra-T/440BX and a multitude of SS7 PCs, I can attest without any doubt in my mind that the former is just better in every way.
Anyway, will get to the flexibility comparisons in a future video, but for now I want to focus on the performance side of things.
As I mentioned in a previous post, I will now switch to SS7 for one or two videos, so I have built the following system:
FULL SYSTEM SPECS:
MB: Asus P5A rev 1.04
CPU: AMD K6-2+ 550 MHz OC @ 633 MHz / FSB105 & 6 multi
RAM: 128 MB SAMSUNG SDRAM PC133 (there are actually 2 x 128 MB modules, but only half the size is detected by the motherboard; using more than 128 MB of RAM decreases performance by 5 - 10%, even when using a chip with on-die cache).
VIDEO: Asus V7700Ti GeForce 2 Ti (OC @ GeForce 2 Ultra clocks)
SOUND: Creative Sound Blaster Live 5.1 SB0220
SOUND2: ESS AudioDrive ES1688F (non-PNP / irrelevant for this test).
HDD: Seagate 40 GB IDE/PATA
I'm planning on doing the benchmarks sometime next week, so I figured that I should first ask for a few suggestions regarding the testing procedure, to avoid being accused that I'm favoring one platform over the other 🤣
A bit of context: this is my second fastest SS7 build. My fastest build also has an Asus P5A rev 1.04, with a K6-3+ 400 running @ 633 MHz (FSB115 x 5.5 multi), however it's currently in storage ~300 km away from my current location. That build is generally ~10% faster than this one - so a faster SS7 build than this one is possible but only if one is very, very lucky with both the CPU and the motherboard.
Having said that, the build I'm using for the upcoming tests is also extremely fast for a SS7! Unfortunately, this particular motherboard does not like FSB speeds higher than 105 (it does work @ 112, but it's not 100% stable, there will be an occasional crash in certain games - particularly Unreal). After investigating, I've concluded that the motherboard cache is the issue (when disabling the external cache it's perfectly stable, but slower).
Still, it's a very fast platform as it is, but I should mention (and I can't stress this enough): the speed of this build is not the norm for SS7, not even close. In fact, I have yet to find a faster motherboard than the Asus P5A (and, boy, have I tested many!). Bottom line, 99% of SS7 builds out there will be slower than this one (some by quite a lot), so keep that in mind.
The CPU I'm using is very overclockable (better than average), so I'm going to ignore one of the rules that I set for myself (I mentioned in the initial thread that overclocks are totally fine - and recommended, actually - since they allow us to better understand what a CPU's true potential is, but I will stick to clocks which should be achievable with most/all CPUs of that particular type). So, yeah, just this one time, I will ignore this rule and go with a chip that is more overclockable than usual.
So, before doing the actual video, I'm open to suggestions regarding what other games I should test. I do have a very important rule: the game must have a benchmark mode that is repetitive and easy to run.
One idea is to test all games with and without sound (because, especially on SS7, there is a HUGE difference between the two). Of course, my main focus since the beginning has been testing with sound, for obvious reasons (people who want to buy these PCs are interested in the real-world performance numbers, and that certainly includes sound as well).
As a reminder, this is the current list of games/benchmarks (in the exact order I've been running them):
- GLQuake (low resolution)
- Quake 2 (low resolution / high resolution / software mode)
- 3DMark99 (default settings)
- 3DMark2000 (default settings)
- MDK2 (low resolution / high resolution)
- Expendable (low resolution)
- Unreal (low resolution)
- Quake 3 (with / without sound + low resolution/high resolution)
If the list of tested games changes dramatically, I will of course revisit the Ezra-T to run the new benchmarks.
2 x PGA132 / 5 x Socket 3 / 9 x Socket 7 / 12 x SS7 / 1 x Socket 8 / 14 x Slot 1 / 5 x Slot A
5 x Socket 370 / 8 x Socket A / 2 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 7 x LGA775 / 1 x LGA1155
Current PC: Ryzen 7 5800X3D
Backup PC: Core i7 7700k