As j already pointed out, it is hard to give any number. An unscientific measure might be Moore's Law, which would mean an increase of computing power of 2^(17/1.5) over the past 17 years, which comes out at ca. 2500 times faster.

Generally speaking, there are a few factors which have changed over the past 17 years which are worth mentioning here, so you can get an image of how much faster it should be (assuming a desktop computer):

• In 2000, there was no x86_64 architecture, which means that a computer from 17 years ago would have to spend more time on handling big numbers
• In 2000, the Intel Pentium III was still in use, the Pentium 4 made its debut at the end of the year, so I will not mention it. The Pentium III was single-core/single-thread and had a clock rate of 450MHz - 1.4GHz. On the AMD side, there was the Athlon with 500MHz - 1.4GHz, also single-core/single-thread.
• The above point means that at the time, you could only use one thread to calculate prime numbers. There are very easy and potent algorithms to run in one thread, however today, we have algorithms which can make scale to multiple threads to quickly calculate prime numbers. That means, that if you use a monster, like an Intel Broadwell-E with 20 Threads or an AMD Ryzen with 16 Threads, you will likely outperform the single-threaded approach, especially when it comes to greater numbers.
• Also, the number of operations per second per core, as well as the processor frequency, were increased in newer architectures, so even if you used a single threaded algorithm, today's CPUs would outperform the old ones easily.
• Storage access is important, and with newer generations of busses, which connect the CPU to the RAM, and newer RAM technology, the speed has increased significantly. Not only that, even L1 and L2 cache were increased ca. by factor 4 and a L3 cache was added

Unfortunately, I am missing precise numbers for performance comparison of the hardware (I might have to dig a little more), but I generally expect, that today, we are 1000-2000 times faster than in the year 2000.

Don't forget, that it is highly difficult to compare different algorithms due to Single-Threaded-ness vs Multi-Threaded-ness.

this depends on the architecture of the CPU and is not that trivial to compare you would need to be more specific because in 2017 I still could use a low throughput CPU.

Even between 2016 and 2017 you would need to specify the architecture. Multi-threading or Single-threading, do you remain in the CPU - Caches or do you have to go to the RAM ? What is speed of the FSB, RISC or CISC. And usually u try to use the GPU for mass computations.

The only thing that I am willing to say ... in theory it could be faster ;).