Intel 8088 vs 8700K

[Hermann]

Hey guys I was just wondering if you could possibly help me out. I am wondering what the performance increase between one of the very first CPU's (in this case the Intel 8088) and the Intel 8700K would be approximately. I used the website Userbenchmark and tried finding the oldest CPU they had which happened to be an Atom Z520 and the results were pretty staggering.

http://cpu.userbenchmark.com/Compare/Intel-Core-i7-8700K-vs-Intel-Atom-Z520/3937vsm4161

Apparently according to real world speed the 8700K was 3301% faster than the Atom Z520, but I would like to know around how much more powerful the 8700K is compared to the Intel 8088. 

In case you are wondering, this is for a school project. Any help or input would be greatly appreciated.

[NelizMastr]

A good chance to learn about reliability and validity of sources as well. Hint, userbenchmark doesn't do well at either.

 

That aside, performance is measured over many variables. You can compare on cores, clock speed, cache, integer score, floating point score etc. There isn't just one metric that determines all performance a chip has.

 

 

 

 

[panther420]

17 minutes ago, Hermann said:

Hey guys I was just wondering if you could possibly help me out. I am wondering what the performance increase between one of the very first CPU's (in this case the Intel 8088) and the Intel 8700K would be approximately. I used the website Userbenchmark and tried finding the oldest CPU they had which happened to be an Atom Z520 and the results were pretty staggering.

http://cpu.userbenchmark.com/Compare/Intel-Core-i7-8700K-vs-Intel-Atom-Z520/3937vsm4161

Apparently according to real world speed the 8700K was 3301% faster than the Atom Z520, but I would like to know around how much more powerful the 8700K is compared to the Intel 8088. 

In case you are wondering, this is for a school project. Any help or input would be greatly appreciated.

It's.. Not even a comparison.

 

The main problem is they are so different you would have trouble figuring out a benchmark that could actually run on both, while simultaneously taking advantage of the 8700k's multithreaded power. To start, 8088 is a 16-bit CPU with all 16-bit registers, but with an 8-bit external data bus according to wikipedia. That means it's probably best suited running DOS (I don't even think it's capable of running Windows 1.0!). In stark contrast, the 8700k is a 64-bit CPU whose 64-bit implementation of X86, the 32-bit version, is loaned from AMD to some extent. It of course, can run any 64-bit operating system with some exceptions for older stuff like WinXP. Given this large discrepancy in a property that fundamental, it really shows that these are extremely different CPUs.

 

But let's say you wanted to do it anyway. The easiest way I'd say to show a raw, painfully simplified comparison would be to calculate the difference in clock speed. An 8700k runs at a base of 3.7Ghz, compared to the maximum of 10mhz on a 8088. Convert the Ghz to Mhz on the 8700k and you have 3700mhz. 3700/10 = 370 => 8700k has a 370x bump in clock speed. Just keep in mind that there are tons of other factors to keep in mind in any benchmark.

 

For your school project. I'd personally recommend you outline that it really isn't a fair comparison, and add a bit if possible to show how much things have advanced in 40 years in the computer industry. Also, you might want to compare that to another industry - like cars, where you can almost always drive a car from 40 years prior if kept clean and working properly. You can't do that with an 8088! You should probably also indicate that the 8700k is so much more powerful, it can emulate a fairly complete 8088 system without breaking a sweat. I haven't researched if 8088 emulators exist, but I'd honestly be surprised if they didn't and besides, it will still be able to do it in theory.

 

Hope this can point you in the right direction.

[straight_stewie]

49 minutes ago, Hermann said:

Hey guys I was just wondering if you could possibly help me out. I am wondering what the performance increase between one of the very first CPU's (in this case the Intel 8088) and the Intel 8700K would be approximately

In case you are wondering, this is for a school project. Any help or input would be greatly appreciated.

Well, this is a perfect example to learn about some of the real metrics used to rate microprocessors.

What we want to do is compare the processor on multiple points, some of the data points may be unavailable for either processor, which makes this very difficult. Fortunately for us, getting data about the 8088 is very easy: The 8088 is identical to the 8086 except for where the 8086 has a 16 bit external data bus, the 8088 has an 8 bit external data bus. This does make accessing full words slower, because it requires to memory addresses. Both are still 16 bit processors. (There's an interesting piece of trivia here, you can follow the development of x86 general purpose registers all the way back to the 8008). 

Ideally, we would compare the processors to each other on these items:

• FLOPS (which weren't even available in hardware on the 8088, except by use of an external co processor called the 8087)
  • Multiplication
  • Division
  • Addition
• IMOPS (A term I invented just now, it means Integer Math Operations Per Second. We are concerned with ALU speed here)
  • Multiplication
  • Division
  • Addition
  • Bit Shift
  • Bit Rotate
  • Bitwise Xor
• IPS (A hard metric to get right on CISC architectures)
• Cache optimized memory access speed (Hard to get right, but will tell us the speed improvement offered by modern caching)
• None cached memory access speed (Very, very hard to get right, but will tell us the speed hit of a missed cache).
• Branching on random input (A hard one to get right as well, since this must be adjusted to account for IPS. Tells us the speed hit of speculative execution)
• Branching on patterned input (Same as above, except for this tells us the speed benefit of speculative execution)
• Merge Sort on a very large data set (this will tell us the speed improvement of multiple cores. Requires atleast a rudimentary OS).
• Multithreaded Matrix Multiplication (this will tell us the speed improvement of HyperThreading, requires a rudimentary OS).

I'm sure that you can see that coming up with a fair comparison of the two is no easy task. It will require you to have access to both processors, and know how to effectively program them. If you're the luckiest person alive, after a few minutes (or hours) of digging you might accidentally find an old standardized benchmark score for some system with the 8088 in it, and the same benchmark for some system with the 8700. This is still an imperfect method, but it's the best you can hope for to even come close to getting a correct answer.

In shorter words; A comparison is not impossible, just very, very difficult. I would suggest seeing if you can change the topic of your report.

 

[Hermann]

31 minutes ago, panther420 said:

It's.. Not even a comparison.

 

The main problem is they are so different you would have trouble figuring out a benchmark that could actually run on both, while simultaneously taking advantage of the 8700k's multithreaded power. To start, 8088 is a 16-bit CPU with all 16-bit registers, but with an 8-bit external data bus according to wikipedia. That means it's probably best suited running DOS (I don't even think it's capable of running Windows 1.0!). In stark contrast, the 8700k is a 64-bit CPU whose 64-bit implementation of X86, the 32-bit version, is loaned from AMD to some extent. It of course, can run any 64-bit operating system with some exceptions for older stuff like WinXP. Given this large discrepancy in a property that fundamental, it really shows that these are extremely different CPUs.

 

But let's say you wanted to do it anyway. The easiest way I'd say to show a raw, painfully simplified comparison would be to calculate the difference in clock speed. An 8700k runs at a base of 3.7Ghz, compared to the maximum of 10mhz on a 8088. Convert the Ghz to Mhz on the 8700k and you have 3700mhz. 3700/10 = 370 => 8700k has a 370x bump in clock speed. Just keep in mind that there are tons of other factors to keep in mind in any benchmark.

 

For your school project. I'd personally recommend you outline that it really isn't a fair comparison, and add a bit if possible to show how much things have advanced in 40 years in the computer industry. Also, you might want to compare that to another industry - like cars, where you can almost always drive a car from 40 years prior if kept clean and working properly. You can't do that with an 8088! You should probably also indicate that the 8700k is so much more powerful, it can emulate a fairly complete 8088 system without breaking a sweat. I haven't researched if 8088 emulators exist, but I'd honestly be surprised if they didn't and besides, it will still be able to do it in theory.

 

Hope this can point you in the right direction.

Thank you so much for the input and help I will definitely take this into consideration for my project.

[Kredbu]

Comparing older processor speeds to new ones is a bit tricky. 

 

Their instruction sets and how many bits per instruction will be different, so you won't be able to do a direct comparison easily.

 

As a hypothetical example...

Let's say processor gen2 has "add" and "multiply" instructions and runs at a speed of 1hz (1 operation per second)

 

If you wanted to do a 4x4  calculation, it would use the multiply instruction and complete in 1 second.

 

If you wanted to do a 3+3 calculation, it would also take one second, using the add function

 

Now let's say processor gen1 only has the "add" instruction in its set of instructions but also runs at 1hz.

 

If you wanted to run the 3+3 "benchmark", it would use the same "add" function as the newer processor, and since it also runs at 1hz, it would complete in 1 second... So you may conclude they are the same speed.

 

If, however, you tried running the 4x4 "benchmark" on the older processor, you could have two possible results.

 

1. The benchmark application will be compiled to run on gen2 only and will try to use the "multiply" instruction. The processor won't know what to do with it, and will fail to run (or it may run the wrong instruction if it has another different one with the same code)

 

2. There may be a separate version of the benchmark compiled to run on the gen1 processor. Since it knows there is no multiply on gen1… it will compile 4x4 into three addition instructions... 4+4, result of that (8) + 4 and result of that (12) + 4

 

So when that runs, it needs to go through the processor 3 times. At 1hz, that means the multiplication "benchmark" will take 3 seconds to run.

 

So now you have two different benchmarks. One says that the new processor is 3x faster, the other says they are the same speed.

 

That is a super simplified example, it skips a lot of things the processor would have to also do as part of the calculation (such as loading and storing)  which would take extra time, but hopefully it illustrates why the comparison is a hard one. Different benchmarks will produce different results, and some may not be able to run due to the vast difference in capabilities of the two.

 

Other things to consider:

-Clock rate comparison of the two

-IPC (Instructions that can be executed per clock)

-How big the maximum number it can deal with calculations on (IE. 8bit Vs 64bit)

 

[Apepa]

1 hour ago, NelizMastr said:

You can compare on cores, clock speed, cache, integer score, floating point score etc.

 

Cores and clock speed are irrelevant unless you're comparing CPUs from the same product line.

 

Also, floating point score is a bit of an issue in this context too, as x86 CPUs relied on separate Math Co-Processor chips until built in FPUs became common in the 90s.

[porina]

I vaguely remember using Landmark to test an Atari ST with software PC emulator. It didn't do well... but I wonder how this bench would score on a modern system, if it runs at all.

https://dosbenchmark.wordpress.com/research/landmark/

 

[Mira Yurizaki]

If you're going to do a comparison between these two processors, the only thing you can really compare is the instructions per second on a standardized benchmark. And since the 8088 had no FPU, it'd have to be on a integer-only based benchmark. Supposedly Dhrystone is good enough for the industry to use as a benchmark to cross-compare different CPUs and Wikipedia has a table from the 8086 (which is the same as the 8088, save for the 8-bit external data bus) to the Ryzen 1800X. I don't know the specifics of Dhrystone other than that, but I'm going to assume it was designed in such a way to test just the CPU, rather than rely on the system itself. i.e., it fits entirely in the CPU's cache and the performance of the rest of the system is not a factor.

 

One thing to note is that CPUs execute things in stages. So theoretically you could come up with a performance comparison if you knew:

• How many stages there are
• How many clock cycles it takes to complete each stage
• How many execution units can be scheduled in parallel
• If it pipelines instructions or doesn't start executing the next one until the previous one is completed.

Note that some features, like speculative execution, superscalar pipelining and simultaneous multithreading may not exactly come into play here if you're assuming the best case scenario. In which case, the code is linear, simple, and predictable, rendering those features a non-factor.