TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
A 2026 report says a modernized Stockfish engine, adapted to a 200 MHz Pentium Pro, reached an estimated 3,180 Elo at 30 seconds per move. In a 100-game test, it beat a 1996 version of Crafty running on modern hardware, but it was not tested against Garry Kasparov; the claim that it could have beaten him is an estimate, not a match result.
A modern chess engine adapted to a 200 MHz Pentium Pro reached an estimated 3,180 Elo at 30 seconds per move, according to a 2026 report by Nicholas Carlini. The tests suggest software advances could have made a 1996 desktop far stronger at chess than contemporary programs, but the engine was not tested against Garry Kasparov, so the claim that it would have beaten him remains an inference.
Carlini reports that he ported and tuned Stockfish, a modern engine, to run within the limits of a Pentium Pro system with 64 MB of RAM and 32-bit processing. In his tests, the adapted engine scored an estimated 2,846 Elo at one second per move, 3,109 at 10 seconds and 3,180 at 30 seconds. The report gives uncertainty ranges for those estimates; they are experimental ratings, not results from games against Kasparov.
The report also compared the adapted engine with Crafty version 11.13, a competitive engine from 1996, running on a modern Ryzen 7 computer. In a 100-game match, Carlini says the Pentium-based modernized engine won 78 games, drew 17 and lost five. This establishes a result against that specific software opponent under the test conditions, not against a human world champion.
In a separate test, the original Crafty build scored an estimated 1,999 Elo on the Pentium Pro at one second per move and 2,383 at 10 seconds. On the Ryzen 7, its one-second rating rose to 2,614. Carlini also reports that adding modern search and evaluation methods to Crafty raised its estimate on the Pentium Pro to 2,718 at one second and 2,974 at 10 seconds. These figures support his argument that better algorithms, not just faster processors, account for much of the improvement.
Software Gains on 1996 Hardware
The experiment separates two sources of chess-engine strength: computing speed and software design. A Pentium Pro cannot examine positions at anything close to Deep Blue’s reported rate, but modern search and evaluation techniques can help an engine use its limited time more effectively. The findings make the historical question less about whether a desktop could match Deep Blue’s hardware and more about what software available in 1996 could do.
That distinction matters because Kasparov’s 1996 match was a landmark in human-versus-computer chess. Deep Blue defeated him in one game but lost the six-game match; IBM’s upgraded machine beat him in their 1997 rematch. Carlini’s tests do not revise those results. They offer a counterfactual: with later engine methods, an ordinary period computer might have been much stronger than the programs people could actually run at the time.
The reported rating estimates also need careful interpretation. Elo predicts relative performance across a defined rating pool and conditions; it does not guarantee a particular result in a short match. A projected advantage over a player rated just under 2,800 is suggestive, but it is not a substitute for a direct, controlled match against Kasparov.
Products Worth Considering
Bose S1 Pro+ All-in-One PA Speaker
Amazon Kindle Scribe (64GB) with Premium Pen, Metallic Jade
Bose SoundLink Plus Portable Bluetooth Speaker, Wireless Outdoor Speaker, Up to 20 Hours Battery Life, Waterproof and Dustproof, Black
As an affiliate, we earn on qualifying purchases.
Deep Blue and the Kasparov Matches
Deep Blue’s first match against Kasparov took place in 1996. It won the first game, marking the first time a computer beat a reigning world champion in a standard match game, but Kasparov won the match. IBM’s system combined general-purpose processors with custom chess chips designed for tasks such as move generation and position evaluation. Carlini’s report says the upgraded Deep Blue used 30 processors and 480 custom chips and could analyze more than 200 million positions per second.
The report contrasts Deep Blue with estimates for engines across eras: roughly 2,300 Elo for strong commodity-hardware programs available in 1996, 2,700 for Deep Blue and about 3,700 for modern Stockfish. Those ratings are presented in Carlini’s analysis and depend on the rating methods and test conditions used. His experiments ask how much of the gap can be recovered by applying modern programming ideas to old hardware.
No Kasparov Match Was Run
The report does not provide a match between the Pentium-based engine and Kasparov, nor does it establish how the engine would perform under the exact conditions of the 1996 Deep Blue match. The claim that it could have beaten him is an extrapolation from rating estimates and engine tests against other software. The ratings include uncertainty ranges, and their meaning depends on testing conditions, including time controls and the opponents used.
The supplied report excerpt also does not give enough detail to independently assess every experimental choice, such as the full game settings and rating pool for each estimate. The Pentium Pro system is described as a later test machine, and the engine uses methods developed after the historical match. It is therefore a retrospective software experiment, not evidence that such a program existed or could have been prepared for Kasparov in 1996.
Further Testing Would Clarify the Claim
The report presents the findings as an experiment rather than a historical match reconstruction. Further evidence would require more complete documentation of the test setup and repeated matches across a broader set of opponents, time controls and hardware configurations. A direct Kasparov comparison is not reported, and no such match is announced in the source material.
For now, the supported takeaway is narrower: Carlini’s tests indicate that modern chess-engine techniques can substantially improve play on a Pentium Pro, including against a 1996 engine running on newer hardware. Whether that particular system would have beaten Kasparov in 1996 remains untested.
Key Questions
Did a modern engine on a Pentium Pro actually play Kasparov?
No. The report describes engine tests and a 100-game match against 1996 Crafty running on a modern Ryzen 7. It does not report a game against Kasparov.
How strong was the adapted Stockfish?
Carlini estimates it at 2,846 Elo at one second per move, 3,109 at 10 seconds and 3,180 at 30 seconds on the Pentium Pro. These are estimates from the reported tests, not a Kasparov match rating.
What happened in the reported 100-game test?
The modernized engine on the Pentium Pro scored 78 wins, 17 draws and five losses against Crafty version 11.13 running on a Ryzen 7, according to Carlini.
Does the experiment show that hardware did not matter?
No. The report says faster hardware helped the original Crafty engine, while modern search and evaluation methods also raised its strength on the Pentium Pro. Carlini’s conclusion is that programming improvements account for most of the Elo gains in his comparison, not that computing power was irrelevant.
Source: nicholas.carlini.com
