From Forum Signatures to FLOPS: A Two-Decade Look at PC Hardware summary: Comparing my 2008 rig (as posted in a psicofxp.com signature) with current hardware, to explore what changed, what didn’t, and what’s actually “planned obsolescence.”
My 2008 Rig (by psicofxp.com Forum Signature)
Athlon64 3200+ Venice s.754
MSI K8N V v2.0 (nForce3 250)
Sapphire Radeon 9550 @ 472 / 472
Markvision + Spectek DDR-400 2x512MB
Creative Audigy 7.1 24-bit
WD 40GB ATA100 7200 RPM
Euro case with LEDs and window
"550W" generic PSU (soon to be replaced)
15" generic CRT @ 1152x864
Aiwa NSX-500 mini component system
Tools of the Time
- ATI Tray Tools
- RaBiT
- ATiWinflash / ATiFlash
- Physical mods: side windows, cold cathodes, front intake holes
- CRTs rescued from e-waste — because LCDs were still a luxury
Spec Comparison: Then and Now
Component | 2003: Radeon 9550 | 2004: X850 XT | 2007: HD 2600 Pro | 2013: R9 270X | 2017: RX 560 | 2021: Radeon 6600 XT |
---|---|---|---|---|---|---|
Architecture | RV350 (DX9b) | R480 (DX9c) | RV630 (DX10.1) | Pitcairn XT | Polaris 21 | Navi 23 (RDNA2) |
Process | 130nm | 130nm | 65nm | 28nm | 14nm | 7nm |
VRAM | 128–256MB DDR | 256MB GDDR3 | 512MB DDR2 | 2GB GDDR5 | 4GB GDDR5 | 8GB GDDR6 |
Bus width | 128-bit | 256-bit | 128-bit | 256-bit | 128-bit | 128-bit |
Clock (MHz) | ~250 | ~520 | ~600 | ~1000 | ~1176 | ~2359 (boost) |
GFLOPS | ~4.4 | ~83 | ~200 | ~1971 | ~2600 | ~10,600 |
TDP (watts) | ~30W | ~75W | ~45W | ~180W | ~80W | 160W |
FLOPS/W | ~0.15 | ~1.1 | ~4.4 | ~10.9 | ~32.5 | ~66 |
MSRP (launch) | ~$90 | ~$499 | ~$149 | ~$199 | ~$130 | ~$379 |
FLOPS/$ | ~0.05 | ~0.16 | ~1.34 | ~7.65 | ~10.0 | ~28 |
Note: These values are approximations based on reference models and typical clock speeds. Values can vary by SKU and manufacturer.
The Fast Decay of Old Hardware
Some parts of that rig are still usable (hello, Aiwa). Others aged out due to:
- Instruction set deprecation
- Graphics APIs
- Storage protocols
- Secure boot and TPM
These are not all examples of planned obsolescence. Some are the result of:
- Physics and economics (Moore’s Law, thermals)
- Software complexity
- Developer productivity vs. backward compatibility
Planned obsolescence means designing something to break, the flaws being introduced in the design by intention. That’s not the same as aging. Some will say, “Oh, my Athlon XP 2000+ with 512MB DDR ran Windows XP more smoothly than this bloated phone runs Android, even with 8 ARM cores and 6GB of RAM. I'm sure it has something like if(rand()%2==0) sleep(1);
buried somewhere inside.
I'm not here to defend all software choices (I'm sure there’s bloat) but hardware really has improved. Here's a look at raw performance per dollar over time. The numbers speak for themselves:
So, it’s worth distinguishing between “not supported” and “made to break”.
Something similar happens about performance per dollar. Without considering the USD inflation rate:
Conclusion: A System’s Value Over Time
That 9550 gave me 4+ years of usable gaming. It booted Linux fast. It taught me VBIOS editing. I learned more from that build than I have from some modern plug-and-play systems. Today’s hardware is faster, but maybe not as personal (?).
That's why not upgrading every generation isn’t about resisting progress but it’s about timing. In areas like GPUs, CPUs, and SSDs, waiting a few months can improve performance per dollar. Prices drop, new generations emerge, and a lower-tier model from a newer family might outperform a mid-range part from the previous one with better efficiency. For instance, trading a RX 560 for a RX 6400 gets you more performance at lower power and MSRP.
That’s playing the long game 😸
So, you want to compare your old GPU against your current one? That doesn't sound fair, you should start by measuring performance and power consumption, then divide.
Next steps: I’ll add screenshots of the X850 XT and HD 2600 Pro, maybe include a gallery/
folder in the repo with captured BIOS dumps or tray tool screenshots.