This is the video transcript. Read the original article with all the details →
Good morning and welcome back to the ValorosoIT channel! Here, we deal with vintage computers, but not only today! We also include a couple of modern computers. What do I want to talk to you about in this video? This video is in response to some questions you asked me, in particular when I replaced the processor of this computer, IBM PS/2 model 30 with an Intel 8086 processor, which I replaced with an NEC V30 to increase its performance. I had upgraded my computer! You asked me: But how much has this performance improved? Can you now run Fortnite on it? Well, I have some doubts about whether Fortnite can be played on it, but in any case you will find the answer yourself if you follow this video to the end. But why an entire video to measure the performance of all these computers, which I will now present to you one by one, and even modern computers? Wasn't it enough to put two numbers: this computer goes to 10, this other one goes to 50, so you could avoid all this video brothel? Well, this question is very interesting and, in fact, I would have liked to ask it too to my Italian teacher, when she expected me to memorize all the first pages of The Betrothed, when Manzoni described in great detail, That branch of Lake Como, which turns towards midday... And my question is: a little drawing?. That is, you made a little drawing, if you didn't know how to do it, you had a friend of yours do it, and we had solved it and I didn't have to learn these pages by heart.
Then, among other things, the topic is also quite interesting. I don't mean Manzoni who will surely be interesting too, but I mean the speed of these computers, because I realized that it is not enough to take the Norton SysInfo and measure the speed of the computer, to calculate the CPU benchmark, because that utility does not take into account the presence or absence of the mathematical coprocessor (FPU), which I had installed both on this IBM and on this one below which has an Intel 286 processor.
The mathematical coprocessor, otherwise known as FPU (floating point unit) is that integrated circuit that is placed alongside the processor and is used to carry out floating point calculations in a hardware manner and not emulated via software by the processor. Here, Norton SysInfo does not take into account the presence of this mathematical coprocessor. I had to find software that also measured the performance of the CPU combined with the math coprocessor, because obviously that helps greatly in floating point calculations. Another difficulty I encountered was measuring the performance of two modern computers, to compare the results of these computers from the late 80s with the current ones, just to see how much technology has advanced in the meantime. In the past I had already made a video Commodore 64 VS modern computer - I really liked that video - there I also took into account other situations inside the computer, so I also measured the capacity of the hard disk, data transfer speed. Instead today we are only concerned with CPU performance.
To measure the speeds of the CPUs of modern computers, Roy Longbottom was very useful to me: he is a computer scientist who has measured the speeds of, I think, all the CPUs existing in the world. He has created a series of programs that are used precisely to measure the speed of CPUs. I got the program for Windows, it actually works under DOS, but it is able to overload all the processors of the modern computer, because modern computers do not have just one processor like the vintage ones: each processor has many CPUs inside, 8, 16, we don't know how many.
In short, it depends on the processor model, and each CPU is also able to carry out multiple operations at the same time, so we should load the CPU according to the specific threads of that CPU itself.
As a unit of measurement for computer speed we use WIPS, which are Whetstone Instructions per Second. In practice, the Whetstone algorithm carries out a series of mathematical calculations, which are repeated N thousand times, or N million times depending on the speed of the CPU, and in the end this index comes out, which we are going to compare between vintage computers and even modern ones.
Obviously, huge brackets can open here, that is, it is difficult to find a program that measures in the same way. Let me explain: if a program is compiled with a certain programming language, if it works in MS-DOS rather than Windows, obviously all these things affect the final result. Also because the Whetstone test, which is the one we are going to use, in theory would also serve to measure the performance of the software, not just the performance of the hardware.
And now I present to you the competitors in this CPU speed race. IBM PS/2 model 30, Intel 8086 processor and 640 KB of RAM. We are going to measure this computer with the original Intel 8086 processor, with the compatible pin to pin processor, which is the NEC V30. And then we also make measurements with and without the Intel 8087 mathematical coprocessor. Below, with the white switch, we have the IBM PS/2 model 30, but with the Intel 286 processor. In this case we are going to make two measurements, so both the one with and without the Intel 287 mathematical coprocessor.
I would have liked to measure the performance of this Commodore PC10, but unfortunately I have difficulties, in the sense that I still don't have a tool that allows me to write 5 1/4 inch floppy disks. I'm buying it, I'm buying a Greaseweazle, if you pronounce it that way, which is used to connect the 5 and 1/4 drive but also the 3 and 1/2 drive to a modern computer via USB port, then I'll keep you updated on this thing, which seems rather interesting to me.
Other computer in the competition: Olivetti M200. It is equipped with a NEC V40 processor and 640 KB of RAM. The NEC V40 processor would be very similar, but not pin to pin compatible to the Intel 80188, which is a processor that has never been used much for computers, but rather has been used for industrial applications. It is a fairly economical processor because it is true that it works at 16 bits, but in reality the external data bus is 8 bits.
Here we are, another competitor, Frael King, with NEC V20 processor and 512 KB of RAM. The NEC V20 is a clone of the Intel 8088. This NEC V20 is a processor from 1984, although in reality the computer must be much more modern, it must be from the late 80s, and I wonder, who, in the late 80s when there were beautiful computers with double color monitors, bought a green phosphor monitor.
Here, the same reasoning as before applies: not having the possibility to write to diskettes yet, we cannot go and measure this IBM 5155.
Finally, we land on this last vintage computer. The last of the vintage ones, because as I was telling you, we also have two modern computers whose speed I would like to measure.
And it's an Olivetti Prodest PC1. As a processor it has the NEC V40, like the Olivetti M200 we saw before, but it only has 512 KB of RAM. It is a computer from 1987. The first of the two modern computers that we are going to measure is this HP Z420, although it is not exactly modern because it is a computer that is already several years old. It has an Intel Xeon E5 processor with 8 CPUs and 16 threads. The other computer is an assembled one, Leo and I assembled it for gaming, and it has an Intel i5-13500 with 14 CPUs and 20 threads.
Well, come on, let's start measuring. I hope the Commodore doesn't get mad because I put the IBM monitor on it.
Now let's open the first computer in the competition: IBM PS/2 model 30 and check which processor I had left inside it the last time I used it. I don't know if it reads well, but it's the Intel 8086 and the math coprocessor isn't installed on the side. In fact, I have here the sachet with both the Intel 8087 mathematical coprocessor and the NEC V30, and then I actually have two NEC V30s, one also as a spare, which we will then install in place of the 8086 to measure the performance of both processors.
Let's turn it on: while it does the self-test, I present to you the little programs: CheckIt, which is the one that will also measure performance with the mathematical coprocessor, the IBM PC DOS with which we will start all the computers. In fact, I insert it before I go looking for the diskette. And then, just out of curiosity, let's also run SysInfo from Norton Utilities, which, as I told you before, does not have the possibility of measuring the performance of the FPU, which is the mathematical coprocessor.
Out with DOS. Let's start with the SysInfo of Norton Utilities version 6. Here it is, here is a summary of the computer's characteristics: let's go to Benchmark, CPU Speed. And it gives us 2.2. Now, I will transcribe all these values and put them in the article that I will write to accompany this video. I will publish the article on my website www.valoroso.it. So, please, then go and see all the performances we measure in this video.
We enter CheckIt. We always go to Benchmark, Main System. Here he is doing the measurement. As for the CPU Speed, he has already measured it, the Dhrystone test, which in practice are fixed point operations, and instead the Whetstone test is the one with floating point operations and, as you notice, he already notices that there is no mathematical coprocessor. As for the CPU, it gives us a value of 801 Dhrystone and 16.4 K (so thousand) Whetstone, so it will be 16.4 K WIPS, when we compare it with the other computers later. And even modern ones.
Let's now install the mathematical coprocessor inside this IBM PS/2 model 30 with Intel 8086 processor. The mathematical coprocessor (FPU) is always original Intel and is the 8087. I had already shown how to mount it in some other videos, even the short ones, it must obviously be mounted following the correct instructions indicated.
Sorry if you can't see it very well from the video, but the position isn't the best. We turn it on, it always starts with IBM DOS version 3.3. I didn't show it in the video, but evidently before putting on the mathematical coprocessor, I turned off everything and obviously all these replacement operations, which I cut a little in the video editing, but in reality they must be done safely. Then turn off the computer, turn off the monitor and change the coprocessor, do all the operations, and then turn it back on only afterwards.
Norton Utilities SysInfo. Here, as you can see, the value is exactly the same as before, despite having installed the mathematical coprocessor. It was this fact that stopped me from using Norton SysInfo to do all the testing. I'm showing it to you, like this, out of your curiosity. But in reality the measurement is not accurate because with the mathematical coprocessor, now we see it with CheckIt, which uses the Whetstone test, you will realize that, compared to before, it will certainly improve. Otherwise what was the point of installing a math coprocessor, which cost a lot of money at the time. CheckIt has identified the Intel 8087 and let's see what it offers us as a test.
The Dhrystones are perhaps slightly lower, but obviously there is a certain tolerance, but they have essentially remained unchanged around 800. Instead, the Whetstones have gone from 16,000, therefore 16 KWIPS, to an abundant 239 KWIPS. Evidently the mathematical coprocessor made a big difference, bringing the index from 16K to 239K and this is why it was actually necessary to also install the mathematical coprocessor at the time.
I have saved you the purely operational part in which I was going to remove the Intel 8086 processor and the Intel 8087 mathematical coprocessor and now we are going to put the NEC V30 in place of the Intel 8086 processor. Okay, let's push it in. Let's try with Norton's SysInfo. Here, with the NEC V30, instead of with the Intel 8086, we went from 2.2 to 2.9 so there is an increase in performance. Let's see what changes with CheckIt. And yes, even without the mathematical coprocessor because we removed it, we have 1023 Dhrystones against the 800 or so we saw before and 20.4 K Whetstones, therefore K WIPS, against the abundant 16K we saw before.
Then, still with the computer turned off, I go to install the mathematical coprocessor, which I must not confuse with the processor, because otherwise we will make a computer with a dual processor. Here we go, yes the little pins are all in so I can push. Well! NEC V30 with Intel 8087 mathematical coprocessor, let's see the performance? 2.8 therefore, with the mathematical coprocessor, the NEC V30 has dropped by 0.1 compared to before. And instead, with CheckIt, oh well yes, with CheckIt you see how much more preponderant is the speed of this computer with the mathematical coprocessor.
Instead, I was noticing that, even with CheckIt, and the Dhrystones went down compared to before where there was no math coprocessor. It's as if, by installing the math coprocessor, fixed point operations are slightly slower, perhaps because CPU performance is affected by the speed of the coprocessor? Honestly, I couldn't explain it to you, but if there are any of you who can explain it to me, you can gladly do so in the comments, because we've already been doing two measures now. With the math coprocessor, fixed-point calculations are slightly slower, while floating-point calculations are obviously extremely faster.
I put aside the IBM PS/2 30 with 8086 processor that we just used to do the experiments. I always opened the IBM PS/2 model 30, but with an Intel 80286 processor. We see that the mathematical coprocessor is not installed because I have it here to install later. Shall we test on this computer?
So, with SysInfo we see that we also have VGA and start the benchmark. Even the graphics are prettier, eh, on VGA, in fact. 5.6, eh, almost double compared to the previous computer. Almost double if we consider the NEC V30, more than double if we consider the Intel 8086.
We already see almost double the Dhrystones and 37,000 Whetstones without a math coprocessor. So, in short, it is a decidedly more powerful processor than the Intel 286 compared to the 8086 or the NEC V30.
Here is our mathematical coprocessor that we are going to install following the correct orientation. So, are all the pins on? We see. Yes, we can push.
The most attentive will have noticed this component here, nwX287. I also made a video about it: it is an alternative to the Dallas DS1287, the clock and also saves the system configuration. In fact, very often, these computers arrive with the internal battery of the DS1287 which is exhausted and you have errors when starting up, typically 161 and 163. By replacing this component with a modern alternative and reconfiguring the computer with the IBM Starter Disk, the errors disappear. Now, in fact, we have to use the IBM Starter Disk, because we need to configure the presence of the mathematical coprocessor.
Let's turn it on and put this disk here, which is the configuration disk. Obviously, I found the images of the disks on the internet and then went to write them onto diskettes, to have them, to be able to use them on these computers. Gives the error. Indeed, 162, which should be the configuration error. Start with the Starter Disk. Here we are, it has indeed found a configuration error and we can use the automatic configuration. Because, alternatively you can manually go into the configuration parameters and enable the mathematical coprocessor. In this case he did everything, so we remove the Starter Disk and put the IBM PC DOS on it.
Norton Utilities, again out of curiosity. Exactly, still 5.6, the CPU, despite having the mathematical coprocessor installed.
And now let's try CheckIt. The Dhrystones remained the same as before. Now let's see… the Whetstones. Eh, we have 248 K Whetstone. Not bad, right? One thing that surprises me a little is that the speed of the floating point operation of this computer with the Intel 80287 as mathematical coprocessor is slightly lower than that of the NEC V30 with the mathematical coprocessor which, if you remember before, we had a slightly higher number, I think 270 versus 248.
Third computer in the competition, Olivetti M200, the one that a friend of mine found in the landfill and brought to me. Then I made four mini videos of me cleaning it and trying it out. You liked these mini videos a lot because I saw that they got a lot of views. NEC V40, 640 KB RAM. We have SysInfo, 1.8, very sad, 1.8. We see CheckIt in black and white because, unfortunately, this monitor is black and white. But why did someone have to buy a computer with a black and white monitor in the late 80s? Even though it says NEC V20 here, it's actually a V40. But okay, who knows, the mysteries of these programs that measure CPU performance. 682 Dhrystone and 13.2K Whetstone. In short, compared to the performances we saw before with the 286 and with the mathematical coprocessor, these are decidedly inferior.
Frael King, let's start it. The story of this computer is also quite nice. Because I was at the hairdresser, it turns out I liked vintage computers. There was a gentleman who wanted to sell it, but he thought that this computer was worth hundreds and hundreds of euros, but then I found out, it was worth very little. I told him, listen, it's worth little, or sell it to me cheap, otherwise you'll sell it to someone else. And luckily he accepted. Let's go with the SysInfo, as before.
The gentleman who sold it to me said: Eh, but I bought it for my son, he hadn't used it much. Oh, damn, I believe it, it's a computer from the late '80s, you can buy it with a green phosphor monitor! In short, already the Commodore 64, even the VIC-20 was in color. In short, we also need to make these gifts look good for the kids, don't we? 1.7.
Oh, after a few attempts which I saved you because it gave an error when reading the diskette, I managed to load CheckIt. Let's measure these blessed performances: 618 Dhrystone and 12.2K Whetstone. The slowest computer encountered so far wins the prize.
Olivetti Prodest PC1, we start it with the Olivetti DOS itself. Because that other guy from the IBM PC, who knows why, doesn't like him. NEC V40 processor. And well, 512K, right? Let's see… 512K of RAM. No, he doesn't like that either!
Oooh! After trying N thousand disks to get it to boot, I succeeded with yet another disk. He has his own problems. In short, we get it. I thought he wanted to pull out of the race.
Let's see if I can read SysInfo. Mmmhhh… So, SysInfo didn't want to read it, but CheckIt at least reads it, let's see if we can do the benchmark.
Oh, at least he uploaded it. This computer here wanted to get out of the race, eh! 614 Dhrystone. So, he's also a bit poor, eh, he's on the low end.
And 11.7 K Whetstone, maybe, maybe, maybe the worst ever.
And here enters the race an almost modern computer, HP Z420, with Intel Xeon E5-4650 processor, 8 CPUs and 16 threads. With 16 threads, according to the indications of Roy Longbottom himself, who created this software, you have to start it twice because each of these instances of software can load 8 threads. So 1 and 2. In the meantime, a log file is also being created in which we will later see what the result is.
The two programs are working, loading the processors and measuring what are then the mega WIPS, in this case. Indeed giga WIPS. However, we only have to take into account the final result that will come at the end, because he starts loading a processor, two processors now, four. In fact, you see, 1, 2, 3, 4, 1, 2, 3, 4, etc., until you end up loading eight. On the indications of Roy Longbottom himself, we add the last value, therefore the one made on 8 processors, of instance one and instance two and we obtain the definitive value of the measurement of this computer.
Here they are, they're done, let's go and see. So, this is for the eight threads and the M WIPS, so Mega WIPS we find them on this column here.
So 27911 for instance 1 and 27843. We can take the calculator, so 27843 + 27991, we have 55754 MWIPS (mega), which is 55.7 GWIPS (giga). We have seen on vintage computers, which went to 16K, 20K, or almost 300K with the math coprocessor. Here we have almost 56 gigabytes. Finally, the last computer, this truly modern one, 13th generation Intel i5-13500 processor, with 14 CPUs and 20 threads. I ran three instances of Roy Longbottom's software to make sure I overload all the CPUs and I'm going to add the final values. Let's take the calculator, 50190 + 48714 + 48198, and see 147 GWIPS (giga)!
Now, if we want to compare the results of Leo's computer, therefore 147 GIGA WIPS, therefore 147 billion times it did the Whetstone cycle in one second, with the crappiest computer we saw before, which was the Olivetti Prodest PC1, we can do some math. Therefore, 147102 mega, i.e. 147 G, or 147102 mega, divided by 11700, which are the 11.7K of the Olivetti Prodest PC1, equal to 12,572,820. That is, in practice, the gaming computer with an Intel i5 processor is 12 and a half million times more powerful than the Olivetti Prodest PC1.
Now, the question that one of you asked me, that is, if Fortnite can run on the IBM PS/2 model 30 with an Intel 8086 processor, but transformed to the NEC V30 and with the mathematical coprocessor installed, let's see if it could run. Let's see how much more powerful Leo's computer is than the IBM PS/2.
So, 147,102,000,000 (performance, therefore of the Intel i5 13500 CPU) divided, let's recover what was the performance of the IBM PS/2 model 30 with the NEC V30 processor and the mathematical coprocessor: 270,200. We have more than half a million times. The gaming computer is more than half a million times faster than the IBM PS/2 model 30, even with the upgraded processor and math coprocessor. So I'll ask you the question, to those who asked me previously: but, in your opinion, if a modern computer gives you 100 FPS on Fortnite, this one here which is half a million times slower, how many FPS does it give you?
Even assuming he managed to load it. In short, it doesn't even have all those gigabytes on the hard disk, it doesn't even have all that RAM, etc. So, give yourself an answer and write it to me below in the comments.
Well, I hope that this video has also been useful for comparing the performances of vintage computers but also of modern ones. What interests me is also to show you how much computers have evolved from about 40 years ago, just under 40 years ago, to today. We have seen that it is not an evolution, which is said Ah yes, it has improved 100 times, 1000 times. No! we are talking about improvements in the order of millions of times, ranging from half a million to even 12 million times more, compared to computers of the time. So, if you managed to get this far to see the video and you liked it, you can give me a like, you can write it in the comments and please subscribe to the channel, activate the notification bell, because it doesn't cost you anything, but it really makes me very happy. See you in the next video, see you soon. Bye bye.