Comparing Old and New Processors – How technology has changed since 1997
Why am I digging out the past?
This post is inspired by a recent discovery of some old, and then very old processors that I found in my stash of hardware. It’s a glimpse into how the technology has changed between 1997 to 2007, and then a quick comparison with more modern processors. I will briefly describe each of the 3 products, give some real-world examples of their capabilities, and eventually draw a comparison between them.
Today’s processors are very much different types of beasts, hugely more capable, and perhaps deserve their own post at some point in the future. So we will only use their specifications for comparison.
Cyrix 6x86MX – 233MHz
Cyrix. Does anyone even remember that name? A relatively short-lived company in the world of processors. Founded in 1988, merged into National Semiconductor in 1997, bought out by Texas Instruments in 2011. To an IT enthusiast since at least 1997, like myself, the only name recognisable in the home computing industry will always remain to be Cyrix. The history started with the release of their Intel 386, 486 and Pentium ranges, albeit being cheaper, but mostly less powerful than their Intel counterparts. Still, those were quite popular in the market of domestic built PC by enthusiasts around the world.
The Cyrix 6x86MX P233 that I have here, is the last CPU released before they were bought out by National Semiconductor (which led to various issues due to changes in agreements etc). Which makes it the greatest product of Cyrix, before (maybe) arguably they have started their demise into the realm of forgotten computer parts manufacturers. An interesting press release of the 6x86MX range can be found here.
How was its performance? Well, despite being Cyrix’s best, it wasn’t very good at keeping up with the demands of the then-starting 3D gaming market. Back when Quake was becoming really popular, being played at every LAN party or IT class in schools (yep, I remember once we finished assignments we were given some ‘free time’), the 6x86MX, even in it’s most powerful form, the 233MHz package, couldn’t keep the gaming enthusiasts satisfied.
Intel Pentium 4 – 2.40 GHz
Another CPU from my stash was a 2002 Intel Pentium 4, with a 2.4GHz frequency clock. Only five years have passed since the Cyrix with 233MHz clock was released, but speeds skyrocketed up tenfold! This was the age of the fastest rising clock speeds, before the advent of dual or quad-cores, hyperthreading, and other technologies used today to improve performance without raising the clock speed exponentially.
Intel was by then well established, and in close competition with AMD. Both companies were trying to outperform each other with the AMD always chasing Intel’s performance, but making up with lower costs (it has stayed this way until today, although AMD isn’t always doing the chasing, and often offers processors that rival those of Intel). One annoying thing about the package they were provided in? The breather hole in the bottom-left corner. Always getting filled up with the thermal paste, making swapping or cleaning up for re-application a difficult process. Glad those are gone!
Even though the 2.4GHz processor wasn’t very special for its times (still roughly around 10 times more powerful as the 1997 Cyrix), another interesting comparison can be made in the technologies they were made with. Intel’s Northwood architecture was made using 130nm lithography process, while Cyrix 6x86MX was manufactured with 350nm processes. This was a huge reduction in the size of lithography used on processor dies, and it’s mainly what has allowed such speed increases in a relatively short period. We’ll compare those values to the next processor, and at the end, we’ll run those past modern CPUs of 2019 to draw some conclusions.
Intel Core2 Duo E4500 – 2.20 GHz
Released in 2007, I can actually remember using this CPU for a long time for my work computer. It offered enough power to run applications available at that time and performed really well. Despite its clock being slower than the 2002 Intel P4 2.4GHz? How was that possible?
Say hello to dual-core processors. Combining 2 logical CPUs on a die, despite its comparable clock speeds, those were capable of much more operations allowed by its double-power. Combine that with improvements in lithography (we’re now down to 65nm – so halved from the technology used 5 years earlier), improvements in architecture and commands used in processors.
That’s very well explained if you look at a comparison of the 2 graphs below. On the left is the performance of a single core compared for both CPUs. While still showing an increase of around 45%, it’s nowhere near the ten-fold increase observer in the previous 2 examples. On the right, a comparison of the performance of the whole package, showing an increase of 557%.
Soon, hyper-threading will arrive. Speeds will slowly creep up thanks to further improvements in lithography. And another CPU range will be retired and abandoned, making space for even faster processors, in the industry where going faster is always the goal.
Comparison of all 3 processors, plus 2 more recent examples
Below you’ll find a table summarising some of the specs for the 3 processors presented above. To keep the compared periods comparable lengths, i.e. around 5 years apart, I’ve added a 2012 Intel i7-3770K and a 2019 AMD Ryzen 3900X. The gap at the last interval is bigger, but it’s an interesting CPU to add for a couple of reasons. It’s the first AMD we’re looking at (remember, how AMD used to be chasing Intel or just sticking to being the cheaper alternative?) that overtakes Intel’s latest i9 chips in performance tests. Also, it boasts an incredible amount of cores, virtual cores and transistors. It’s a beast!
- Cores/threads: 1/1
- Lithography: 350nm
- CPU Mark: 18.55*
- Transistors count: 3m
IntelIntel Pentium 4
- Cores/threads: 1/1
- Lithography: 130nm
- CPU Mark: 228
- Transistors count: 55m
IntelCore2 Duo E4500
- Cores/threads: 2/2
- Lithography: 65nm
- CPU Mark: 1,271
- Transistors count: 167m
- Cores/threads: 4/8
- Lithography: 14nm
- CPU Mark: 15,928
- Transistors count: 1.48B
- Cores/threads: 12/24
- Lithography: 7nm/12nm
- CPU Mark: 31,870
- Transistors count: 19.2B**
What’s next ?
With the exponential growth shown on these graphs, can we expect performance scores 10 times those of today’s in the next 5 years? Who knows. With the constant push for higher speeds, better architectures and better lithography processes, we probably will. And then there’s quantum computing, still an emerging field, but likely to break all the rules we’ve known so far.
* – As there are no CPU Mark tests for the Cyrix 6x86MX architecture (the tests available back then were different so can’t compare them directly), I have derived the APPROXIMATE CPU mark score of 18.55 by comparing that CPU’s Mark 99 score with the Mark 99 score of the next processor in our review (13.1*228/161=18.55)
** – AMD Ryzen gen 3 chip has 19.2B of transistors, but the particular CPU reviewed reportedly is only using 4.8B of those. Meaning that there’s a lot more room for better performance packages based on that chip.