For me it was Quake. They said quake would run better on the Pentium. But... my overclocked 140 MHz 486 DX4 and motherboard loaded with cache actually ran quake better. At least until the higher clocked Pentiums (133,166 etc) came out and was an almost affordable upgrade. But.. those EDO RAM sticks cost me my soul, two shirts and all my lifetime savings.
Back in the day when I was gonna buy a 486dx4 100MHz system, my cousin recommended me a Pentium 90MHz instead. I opted for the Pentium. I'm pretty sure that was a good choice.. All I know is, that it ran Doom, Descent and the Second Reality demo beautifully! Don't remember the gfx card exactly, but it was a Tseng. Later, I got a Gravis UltraSound sound card (loved demos and music modules) and the Orchid Righteous 3dfx card. Tomb Raider was the first game I played with the Orchid. I could write much more, but I'm gonna stop.. :) Ah, the good old DOS days. A very special time.
i had the choice between a pentium 100 and a 486dx4 100. i picked the 486 and it was DEFINTITIVELY the wrong choice. but it did run all those games you mentioned just fine. quake was the first game to really kill it.
Actually, even 60 Mhz Pentium would outperform 486 DX4 at 100 mhz. In most cases even by significant margin (like it match about 486 DX4 at 150mhz). Proof here ua-cam.com/video/NLrKxWL73Mw/v-deo.html
1) P5 arch isn't two 486s mated together, just two EUs that have different capabilities. Kind of like ROPs in GPUs, although those can all do the same thing. 2) Some late 486s have write-back L1 cache.
@@uplink-on-yt No. All Intel CPUs up until Pentium 4 are a single core CPU. On a single core CPU such as the Pentium- you have one CPU core on the die with one set of execution units (such as ALU and FPU). Super scaler architecture design, allows the EUs on the core to simultaneously execute instructions. With a dual/multi core processor, you have multiple CPU cores on the same die. Each core has its own set of EUs along with super scaler architecture. To make an analogy, imagine a single core CPU as a office with one manager who manages only one team, and each of those team members is an EU. A dual/multi core system would be an office where there are multiple managers who are each managing their own team.
@@paulgraves1392 - I picked on the claim of the video that a Pentium was two 486 CPUs (a DX and an SX) put together on a die, and called them "pipelines".
My 486DX4/120 was able to run Diablo pretty fine. The setup program of the game was reporting the cpu as a "pentium-60" which is consistent with the notion that a Pentium offers almost double the performance of a high-end 486.
my dx4 100 also ran diablo... playably. wasn't full speed all the time but certainly didn't keep me from finishing it. things only got bad in hell with tons of projectiles and firewalls and such.
L3 cash is when you get an inheritance from some rich uncle and become a multi-millionaire overnight. L4 cash (e-Dram) is when your family name is Rothschild
the mathematical traditions are generally (but not exactly): x,y,z,r,ρ (that last being rho, not p) - coordinate/distance variables, used primarily in cartesian (x,y,z) or radial (r,ρ) a,b,c,d,h,w,l/L,R,x0,y0,z0 - constants θ,φ - angles (that's theta and phi) α,β,γ - constants in formula relating to angles (basically the same as the previous constants, just greek) t - time T,t0 - constants of time From here we get to our 'u,v'. There are a few things that use u,v. f,g,h,u,v,w - functional notation such as "f(x)" another is in the "substition method" of solving differential equations, in which you replace a complex segment of a function with "u" and solve for that with respect to u... du/dx. This of course uses the u/v variables due to its relation to the previously mentioned functional notation. I would suspect that the usage of "u/v" in these processors is a nod to this substitution method. In which the composite functionality of both "pipes" combined is the combination of the 2 pipes u and v. Since u and v don't represent a variable value, but rather a complex functionality, and those 2 complex functionalities combine to make the whole complex functionality. And I base this theory off of my 15+ years of experience in computer science where this tradition tends to hold strong when representing "functions". There are other common traditions: i,j,k - these generally mean indices, or sequences of things. For example a vector or matrix will denote each column as i,j,k. Also quaternions will denote their components as i,j,k which is borrowing from both vector notation AND the fact that complex numbers rely generally on the letter "i" (for imaginary), but since quats contain 3 complex numbers, it sequences them as i,j,k where i*j*k = -1. m,n,N - notation for any integer/number seen often in things like summations using the Sigma character Σ. And of course all of the various special constants like π and e. There are also traditions to avoid many symbols like the letter o (upper or not) and other symbols that can easily be confused for numbers or existing traditional symbols. With one weird exception being lower-case l, but that's likely just because in hidden writing the lower-case l gets little curves on its ends, and it's only confused with I (upper-case i) and 1 in computer fonts which is really only a new addition to the math realm... nevermind that maths tend to prefer handwriting to this day even with the advent of computers.
OOh - I like that you've got the equipment list there. Perhaps this is something I should start doing on my channel. Anyway - I've started soaking these up, probably backwards. Is there a playlist for these?
Aww yeah it's all about the Pentium! Had a 90 MHz skew back in the day, with a whopping 64 MB of DRAM, a Diamond Viper v770 for video and an AWE64 Gold for audio. Ran every DOS game imaginable, plus with some ems memory mapped, I had oodles of sample memory for playing cool MOD's and XM's in Fasttracker. Completely tricked out Norton Commander, QEMM memory management, customised (choice.com'd) autoexec.bat for those last few kb's of conventional memory for quake. Aah, nostalgia.. There you are :) Now I want to hunt ebay for old parts and build a proper ultimate specced 90's DOS rig. Maybe a Pentium 2 @333? Thanks for recalling the memories Mr Nostalgia Nerd!
+devjock Man, AWE64 was the nuts back then! I've got a Pentium 4 running as a 90's DOS/Windows 98 rig..... It works an absolute treat. Possibly the most powerful thing many of these games have ever come into contact with!
Aww! Norton Commander. How I miss it!!!! I STILL haven't been able to get two farting directories showing up in the same window to make all thos transfers and comparisons so easy! My Norton Commander came free with a Spanish magazine, and ended up on all my friends' and associates´ computers. I never really learned to use DOS, since I had this from the start. Long live Norton Commander! Death to Norton Antivirus!!
Love that you let that beginning clip speak for it's self, great editing decision in my opinion, I feel like too many people would try to cut it up into bits and pieces and use it through the video which I feel would lessen it's impact. Though have wonder why that guy seemed so pissed off about the letter code names, lol.
Pentium was not at all two 486s. It has two ALU pipelines, one being FPU capable. And the AMD K5 had two ALUs and one separate FPU, but it's a single core chip. Later CPUs added ALU and FPU units, and no one pretended they were somehow separate CPUs or a previous generation mooshed together.
There IS an Intel 486 DX2-66Mhz that DO have write-back technology, look up for the S-Spec "SX955", also a DX4-100Mhz with write-back, but i forgot which S-Spec exactly.
Actually the Pentium Pro was the birth of the exploits. The original Pentium did have branch prediction, however it lacked out of order execution which is required for Meltdown and Spectre exploits work.
Really interesting, but a little too brief for my liking. If you feel inclined a more lengthy analysis would be interesting, but I appreciate this isn't everyone's cup of tea.
+David Knight A video detailing how the Pentium directly influenced the way more modern processors work would be great, but his focus is primarily retro so I guess he wouldn't be interested in doing that. Maybe LGR will see this and pick up the idea :P
+David Knight I'll probably cover the x86 architecture in a future Feature video, as it warrants detail - I mean you could dedicate a whole channel to this stuff, but these Byte size videos are intended to be just that... small chunks of information in a quick mid week video.
+Nostalgia Nerd Actually your tech related videos are far more interesting (at least to me) than you playing a weird DOS game (well if it's not Elite or Syndicate of course).
The set associativity is about searching faster but about increasing the hit rate. The simplest cache is direct mapped where one memory location has only one possible location in the cache. This is simple but in worst case can lead to speed slower than no cache at all. In a two way set associative cache each memory location can have two possibilities in the cache. While still one could create code where it is worse than no cache one really would have to try. For way is better but more complex. 486 has a four way set associative cache so the statement that it was changed to two way for search speed is technically true but gives a wrong impression as it was "divided" less, not more. Modern CPUs can have up to to 16 way set associative caches.
It would have still been helpful, even to those who weren't overclocking due to limiting heat exposure and a prolonged life, so if you use a 286 today or higher, put at least a heatsink on or it may fail in a couple if years if you're enthusiastic enough to use one. Otherwise if it's below for example 10 MHz it wouldn't matter!
@@RuruFIN WTF DOES AFAIK MEAN and I wasn't stating that it would be mandatory but it would definutely help while overclocking or preserving the lifespan
This sums up why an early Pentium kicks ass for pretty much any DOS game you throw at them. Too fast on a Pentium 133? No problem, just disable the cache and you're on a level of a 386/486.
But early Pentium (Socket 4) sucks and is often outperformed by 486DX4 in everything but Quake. Those are the 60/66 Mhz heaters with the FDIV bug. I agree P133 is a wicked good CPU but that is mature Pentium era, far from early.
I loved my i486DX4-100. I used it up until 1997 after which I built a new system based on an AMD K6-233. The K6 was a pretty decent chip and I would run benches against a P-233MMX I had as well. The K6 would beat the P55C in integer calculations, but the intel would beat the K6 by a small margin in floating point calculations. I even used the same Socket 7 motherboard for both cpu's. Fun times back then.
Cache Associativity can be a pro or con. Normally the CPU (back in those days) will have to iterate through all the cache line to find something. If the caches sizes increased this caused significant slow downs. In a N-way Associative Cache the data you are looking for can only be in N spots in the cache. There for you will not need to iterate through the whole cache. The major con with this is if the associativity is low, you ran into a lot conflict misses where data you need a few instructions later could be evicted from the cache to make room for more urgently needed data. Going to RAM to re-fetch the data was very slow.
there's just nothing better than seeing a 486 computer running emm386.exe. with a full contiguous pageframe address without any issues. This is impressive for those day's good luck now day's as those fancy graphics card's tend to utilize and populate those memory addresses with uma memory so yah your pc is cheaper for a reason. loss of full functionality.
My first computer was a 486 @ blazing 25 Mhz 2 mb of ram and 75 mb hard-drive. My phone today vastly outperforms the NASA supercomputers of my childhood. Love technology :)
Not all 486 use write-through, later 486's do use write back cache. In addition, a Pentium processor executes two instructions at a time only if certain conditions apply, so the code must be optimized for it, which was not the case back then, so early pentium chips were not really much faster than high end 486's.
Excellent comment. When I upgraded from my DX4-100 rig to a Pentium 75, some games ran slower... and this was not remedied until I upgraded the Pentium 75 to a Pentium 100!
Your explanation of what a 2-way set associative cache is doesn't seem correct. It means that any address in DRAM can be stored in 2 different places within the cache. The processor will use the least-recently used of those 2 locations to store the incoming data. This is better than a 1-way set associative cache because it means fewer recently used cache locations have to be evicted to make room for incoming data.
The Cyrix 5x86 processors were pretty far ahead of their time for a socket 3 processor, while the later Intel DX4 / POD and AMD 5x86 133 mhz models held their own in the early Pentium days particularly if you had a motherboard that allowed you to enable the write-through cache mode that most of these CPU's supported. Cyrix actually quit selling their socket 3 processors out of fear it would their 686 class processors. I'd argue the Pentium II in it's quirky slot-1 form factor on the new (at the time) ATX form factor is a far more interesting leap forward in computer standards than the socket 7 / Pentium which was really just a juiced up 486 with many of the same old warts in terms of motherboard support/standards.
some 486 CPU's use write-back though in my experience it can be a tad unstable depending on the motherboard. I think the 66mhz dx2 was the first 486 to have a write-back variant produced. There are also some quirks to a 486 using write-back. generally if you have write-back enabled on a 486 any SCSI controllers you install won't work. Its due to some sort of conflict that occurs but I don't know the specifics. There are supposedly a few SCSI controllers cards that were made with a jumper that will correct this issue.
+0blivi0n100 ah, while reading memories flooded back of that old DX4-100 with write-back cache we upgraded to after an am386DX-40 which was also a beast in it's own right, but got served so damn hard by that 486 monster. It felt like that fastest thing in the world back then, I wonder what I can push it to still do today. maybe a low-res divx movie if I'm lucky? lols
+interlace the 386dx-40 was indeed a monster of a 386. You'd probably be surprised what a dx4-100 could do especially if something is written specificly for it. Heck, I recall a video somewhere where they got a movie to play on an 8088. If you really want to push it try popping on an AMD 5x86 133mhz (its not actually a 5x86 but a supped up 486) and then overclock it to 160mhz which most of those chips can do easily. That's a monster 486. Its roughly equivalent to a Pentium 90 or 100mhz except when the software was optimized for the Pentium or made heavy use of the FPU.
0blivi0n100 I had something close! My dad (i was 11 then) gave me my own cyrix 6x86 150+ that had it's weakness and strength. for retro sake built a p166mmx dos-only rig years ago that did websurfing, cd burning, mp3 and divx playing (all from dos) awesomely :) it had a voodoo5 in it. Wish I hadn't given it all away.
Well the V2 would be better with a pentium 2 or even an early pentium 3. Using this with a pentium would be underpowering the V2 (even though, a V2 is still faster than a V1 even with these cpus)
Multi-Core Processors differ from the Pentium because each of their Cores are in essence bare-bones Processors in their own right; and thus can (if developed for properly) operate in almost complete independence. The Pentium Processor added a Second ALU (Arithmetic and Logic Unit), providing a Secondary Data Pipeline... the Super-Scalar Unit (commonplace on most other contemporary CPU like the more popular Motorola 68K, IBM PowerPC, HP PA-RISC, MIPS, etc.) essentially then acts like a File Clerk, providing Tasks for each ALU to Complete. This allows for up to 2 Threads (Operations) Per Cycle instead of the 1 Thread on the i486. In Theory this meant twice the Performance, but in practise due to limitations of the x86 Architecture this actually resulted in > 1 but < 2x performance (iirc on average it was about 60% Faster) en.wikipedia.org/wiki/Intel_8086#/media/File:Intel_8086_block_scheme.svg If you take a look at that (the 8086, the earliest incarnation) ... on that Diagram (9) is the ALU, which basically does all of the grunt work of the Processor. The (5) Instruction Queue was also split with a the addition of a Superscalar Control Unit, that would determine in Pentium which ALU was to be used and in Pentium Pro would Re-Order the Instruction Stack so that both ALU were finishing their Operations at as close to the same time as possible. One of the problems with the Pentium Design was while it allowed for Superscalar Threading, it didn't allow for Independent Threading (like the 68040 and 060 Superscalar provided) ... thus how the 68060 works is much more similar to how Modern Multi-Core Processors do, albeit with Shared Registers.
super-scalar pipelines in the same core are fed by the same decode unit and can share registers. They also share the same cache (which feed fetches for both instruction and data registers). This means superscalar is cheaper to implement than dual-core because multi-core has to have cache coherency, meaning a write to a memory address in one core's cache must invalidate the data in the other core's cache if it is there, this means a lot more wires and transistors. Most multicore CPUs' individual cores are superscalar themselves because it is an effective way to execute more instructions per clock.
I've got a book (somewhere) about hard discs. Back then, in DOS days, you didn't know what "information" you needed to know, so you tried to learn everything. That's why I'm now soft in the head.
Acpi is permanent at irq9 on pentiums. It is bad for couple of isa hardware, such as sound cards. +1 to 486. Speed of pentium is too high for many old games. +1 to 486
Thank you for calling it the correct term cache ("cash") rather that than the incorrect term cachè ("cashay"). Very few people nowadays use the the correct term nowadays when referring to a processor's cache (L1, L2 or L3)
I would've bought a PR- equivalent K5 to whichever Pentium I was buying, except I was buying a Pentium-200 and the K5 wasn't exactly widely available where I live, let alone a K5-PR200 which was rarer than hen's teeth.
First time I gave thumbs down to your video - sorry, I'm not worthy" I'm not worthy"- but I really expected more full analysis than the cache thing. Like the bus-speeds and other things, like what CPU extensions, like better pipelining, new instructions, etc. it had? Btw, was it Cyrix that made these "486" or "Pentium" (not sure of latter) that basically were 486 (and Pentium possibly) that supported higher speeds and new features, nut were limited to older architectures hardware, like BUS sizes, etc? Or some other Company?
Those bloody bastards... you realize you're basically telling me the pentium was already the first "dual-core" multithreading cpu 20 years ago?!?!?! and if L1 cache is a type of ram that can run at 4+ghz on overclocked i7's, I start to wonder if it shouldn't 've beem technically possible years ago to have regular ram at those speeds, and tech release is being held back way too much to maximize profit.
Its actually not that hard to understand. Think of a processor with 10 different pipeline steps. The first step fetch instruction, the last step write the data back. There is 8 steeps between doing... stuff. It don´t really matter what the instructions between do. So for the pipline controller it think its a 10 step pipline. What it really is is two 5 step muxed pipline. So the first instruction goes to the first pipline, the second, go to the second, the third go to the first again, the forth go to the second.. and so on. The result is that for every clock the pipline does, it make two calculations.
The explanation is kind of faulty. Its just the pipeline that is double up. No the entire processor. A modern Zen processor have 4 ALU, 2 AGU and 4 FPU pipes (some with SIMD). It got 6 operation dispatcher. So the 6 operation dispatcher have 10 pipes in total to chose from. It also got two operation internalizer (jump, branch and that kind of stuff) and 4 instructions internalizer (calculate, compare and so on). So it can initialize up to 4 instruction 2 operations per clock and it can in theory do 10 pipes. But because it only can initialize 6 processes, it can only do 6 processes per clock. More realistically it can probably do 5 with near perfect code. It should not jump under 4 if the code is half decent. But of cause if there are a lot of branches it will really go to hell. This is done on two threads, but in one core. The CPU will have (usualy) 8 cores, so there would be a total of 32 ALU and 4 FPU in a 8 core processor. What the Pentium did was just have two pipes and run every second instruction on the even one and every other second on the odd on in the same thread, this way it kind of "emulate" a double clock speed with double the pipeline length. The draw back is that the simulated pipeline become really long. The upside is that its really simple making code for it, you only need one threed. The down side is that the branch prediction is really hard. That was what broke the P4.. Basically for a Pentium the CPU does two instruction per clock per pipeline steep. For 486, they make one per clock per steep. For all older they do one instruction for every two clock per step. So for a 486 running FPU code it make one 32 bit calculation per clock. A Zen can make up to 4 FPU calculation per clock per core 128 bit (or 4 x 32 bit) per core. It also have say 8 cores and a freqvency of say 4GHz. So the frequency is about 40 times higher. It can do 4 times more instructions It can do 4 floating-points per instruction It got 8 times more cores. So it should be about 5000 times faster running only FPU code. In reality its more like 1000 times faster or so. This is even true for mobile CPU:s. A ARM A57 can do 2 instruction per clock and 128bit instruction wide. Thow it is a RISC so it "loses" quite a lot of instructions.
It's not. It's actually quite different. While a dual core chip can operate on two concurrent threads at once, a superscalar cpu can only operate on one thread at a time. The difference is that there is extra hardware that tries to bunch up several instructions to be executed at once.
![Дон ДОН, Алаудинов и СБЕЖАВШИЕ из под Курска ахматовцы 😁 [Пародия]](http://i.ytimg.com/vi/dEfCf5IK26Q/mqdefault.jpg)
![Why does DOS use 8.3 Filenames? [Byte Size] | Nostalgia Nerd](/img/n.gif)
"It makes DooM run better!" was the only explanation I needed back in the day.
Doom ran perfectly on all 486 and some 386DX with a decent VGA video card anyway..
I purchased my first Pentium only for doom. As a student, I worked so hard to buy it, I think I worked a whole year doing extra stuff.
For me it was Quake. They said quake would run better on the Pentium.
But... my overclocked 140 MHz 486 DX4 and motherboard loaded with cache actually ran quake better. At least until the higher clocked Pentiums (133,166 etc) came out and was an almost affordable upgrade. But.. those EDO RAM sticks cost me my soul, two shirts and all my lifetime savings.
@@OhFishyFish oh no. on some 386 with ISA display it was lagging. Had to make the screen smaller to smoothen
more like... "It makes Terminal Velocity to run playable on "Pentium" setting".
Back in the day when I was gonna buy a 486dx4 100MHz system, my cousin recommended me a Pentium 90MHz instead. I opted for the Pentium. I'm pretty sure that was a good choice.. All I know is, that it ran Doom, Descent and the Second Reality demo beautifully! Don't remember the gfx card exactly, but it was a Tseng. Later, I got a Gravis UltraSound sound card (loved demos and music modules) and the Orchid Righteous 3dfx card. Tomb Raider was the first game I played with the Orchid. I could write much more, but I'm gonna stop.. :) Ah, the good old DOS days. A very special time.
i had the choice between a pentium 100 and a 486dx4 100. i picked the 486 and it was DEFINTITIVELY the wrong choice. but it did run all those games you mentioned just fine. quake was the first game to really kill it.
Well obviously the Pentium is much superior, but it must have also been much more expensive. It's not really a choice as much as a budget.
I guess you had the Tseng ET-4000. It was a popular and good card at the time
@@ivarfiske1913 Googled ET-4000, and those pictures of it definitely ring a bell. Pretty sure you're right. :)
Actually, even 60 Mhz Pentium would outperform 486 DX4 at 100 mhz. In most cases even by significant margin (like it match about 486 DX4 at 150mhz). Proof here
ua-cam.com/video/NLrKxWL73Mw/v-deo.html
1) P5 arch isn't two 486s mated together, just two EUs that have different capabilities. Kind of like ROPs in GPUs, although those can all do the same thing.
2) Some late 486s have write-back L1 cache.
Was Pentium the original dual core?
@@uplink-on-yt No - IBM introduced the worlds first Dual Core CPU Power4 in 2001. AMD 64 X2 in 2005 and Intel with the Pentium D in 2006
@@uplink-on-yt
No. All Intel CPUs up until Pentium 4 are a single core CPU. On a single core CPU such as the Pentium- you have one CPU core on the die with one set of execution units (such as ALU and FPU). Super scaler architecture design, allows the EUs on the core to simultaneously execute instructions.
With a dual/multi core processor, you have multiple CPU cores on the same die. Each core has its own set of EUs along with super scaler architecture.
To make an analogy, imagine a single core CPU as a office with one manager who manages only one team, and each of those team members is an EU. A dual/multi core system would be an office where there are multiple managers who are each managing their own team.
@@paulgraves1392 - I picked on the claim of the video that a Pentium was two 486 CPUs (a DX and an SX) put together on a die, and called them "pipelines".
0:43 Who the heck created that graphic? He went through all the trouble of rotating the text and pins, and ended with that perspective?
My 486DX4/120 was able to run Diablo pretty fine. The setup program of the game was reporting the cpu as a "pentium-60" which is consistent with the notion that a Pentium offers almost double the performance of a high-end 486.
my dx4 100 also ran diablo... playably. wasn't full speed all the time but certainly didn't keep me from finishing it.
things only got bad in hell with tons of projectiles and firewalls and such.
L1 cash is when u got enough money to live on ur own
L2 cash is when ur ballin
lel
L3 cash is when you get an inheritance from some rich uncle and become a multi-millionaire overnight.
L4 cash (e-Dram) is when your family name is Rothschild
"U" and "V" are used when "X" and Y" are already taken and can cause confusion.
the mathematical traditions are generally (but not exactly):
x,y,z,r,ρ (that last being rho, not p) - coordinate/distance variables, used primarily in cartesian (x,y,z) or radial (r,ρ)
a,b,c,d,h,w,l/L,R,x0,y0,z0 - constants
θ,φ - angles (that's theta and phi)
α,β,γ - constants in formula relating to angles (basically the same as the previous constants, just greek)
t - time
T,t0 - constants of time
From here we get to our 'u,v'.
There are a few things that use u,v.
f,g,h,u,v,w - functional notation such as "f(x)"
another is in the "substition method" of solving differential equations, in which you replace a complex segment of a function with "u" and solve for that with respect to u... du/dx. This of course uses the u/v variables due to its relation to the previously mentioned functional notation.
I would suspect that the usage of "u/v" in these processors is a nod to this substitution method. In which the composite functionality of both "pipes" combined is the combination of the 2 pipes u and v. Since u and v don't represent a variable value, but rather a complex functionality, and those 2 complex functionalities combine to make the whole complex functionality. And I base this theory off of my 15+ years of experience in computer science where this tradition tends to hold strong when representing "functions".
There are other common traditions:
i,j,k - these generally mean indices, or sequences of things. For example a vector or matrix will denote each column as i,j,k. Also quaternions will denote their components as i,j,k which is borrowing from both vector notation AND the fact that complex numbers rely generally on the letter "i" (for imaginary), but since quats contain 3 complex numbers, it sequences them as i,j,k where i*j*k = -1.
m,n,N - notation for any integer/number seen often in things like summations using the Sigma character Σ.
And of course all of the various special constants like π and e.
There are also traditions to avoid many symbols like the letter o (upper or not) and other symbols that can easily be confused for numbers or existing traditional symbols. With one weird exception being lower-case l, but that's likely just because in hidden writing the lower-case l gets little curves on its ends, and it's only confused with I (upper-case i) and 1 in computer fonts which is really only a new addition to the math realm... nevermind that maths tend to prefer handwriting to this day even with the advent of computers.
OOh - I like that you've got the equipment list there. Perhaps this is something I should start doing on my channel. Anyway - I've started soaking these up, probably backwards. Is there a playlist for these?
i wish i still had my 486 from the 90s
+I HOPE ALL GOOGLE BASTARDS DIE A PAINFULL DEATH You and me both!
You can have mine for free.
Aww yeah it's all about the Pentium! Had a 90 MHz skew back in the day, with a whopping 64 MB of DRAM, a Diamond Viper v770 for video and an AWE64 Gold for audio. Ran every DOS game imaginable, plus with some ems memory mapped, I had oodles of sample memory for playing cool MOD's and XM's in Fasttracker. Completely tricked out Norton Commander, QEMM memory management, customised (choice.com'd) autoexec.bat for those last few kb's of conventional memory for quake.
Aah, nostalgia.. There you are :) Now I want to hunt ebay for old parts and build a proper ultimate specced 90's DOS rig. Maybe a Pentium 2 @333?
Thanks for recalling the memories Mr Nostalgia Nerd!
+devjock Man, AWE64 was the nuts back then! I've got a Pentium 4 running as a 90's DOS/Windows 98 rig..... It works an absolute treat. Possibly the most powerful thing many of these games have ever come into contact with!
Aww! Norton Commander. How I miss it!!!! I STILL haven't been able to get two farting directories showing up in the same window to make all thos transfers and comparisons so easy! My Norton Commander came free with a Spanish magazine, and ended up on all my friends' and associates´ computers. I never really learned to use DOS, since I had this from the start. Long live Norton Commander! Death to Norton Antivirus!!
Love that you let that beginning clip speak for it's self, great editing decision in my opinion, I feel like too many people would try to cut it up into bits and pieces and use it through the video which I feel would lessen it's impact.
Though have wonder why that guy seemed so pissed off about the letter code names, lol.
+Kandi Gloss Gotta love some Minasi XD
Pentium was not at all two 486s. It has two ALU pipelines, one being FPU capable. And the AMD K5 had two ALUs and one separate FPU, but it's a single core chip. Later CPUs added ALU and FPU units, and no one pretended they were somehow separate CPUs or a previous generation mooshed together.
There IS an Intel 486 DX2-66Mhz that DO have write-back technology, look up for the S-Spec "SX955", also a DX4-100Mhz with write-back, but i forgot which S-Spec exactly.
But not every 486 was write through, there were a lot of write back ones, even DX2s.
I always loved the term super-scalar. Now we hardly use it. Bring back the 90's etc
Reminds me of a Muse song (and there's nothing wrong with that!)
Computers have come so far. And windows is still no faster
Your bite size videos... sorry, byte... are uniformly great. Thank you!
This was somehow "Recommended" to me by Google. I must have watched ALL the newer stuff! Anyway, it was, er, informative-ish.
I still don't quite understand cache but I learnt something new today. Did t realise a Pentium is just 2 486 chips in one.
Aaaah. Branch prediction. The birth of Spectre, decades ago.
Actually the Pentium Pro was the birth of the exploits. The original Pentium did have branch prediction, however it lacked out of order execution which is required for Meltdown and Spectre exploits work.
@@phoenixpoole2494 interesting! I thought that out of order was in all pentiums.
Really interesting, but a little too brief for my liking. If you feel inclined a more lengthy analysis would be interesting, but I appreciate this isn't everyone's cup of tea.
+David Knight A video detailing how the Pentium directly influenced the way more modern processors work would be great, but his focus is primarily retro so I guess he wouldn't be interested in doing that. Maybe LGR will see this and pick up the idea :P
+Kandi Gloss The 23 year old pentium isn't retro?
+David Knight I'll probably cover the x86 architecture in a future Feature video, as it warrants detail - I mean you could dedicate a whole channel to this stuff, but these Byte size videos are intended to be just that... small chunks of information in a quick mid week video.
David Knight it is, but modern chips are not.
+Nostalgia Nerd
Actually your tech related videos are far more interesting (at least to me) than you playing a weird DOS game (well if it's not Elite or Syndicate of course).
When I was a student, my Computer Science teacher told me the 486 was really a bunch of 8086's bunched up to work togeather.
In a way he's right but not that simple
The set associativity is about searching faster but about increasing the hit rate. The simplest cache is direct mapped where one memory location has only one possible location in the cache. This is simple but in worst case can lead to speed slower than no cache at all. In a two way set associative cache each memory location can have two possibilities in the cache. While still one could create code where it is worse than no cache one really would have to try. For way is better but more complex. 486 has a four way set associative cache so the statement that it was changed to two way for search speed is technically true but gives a wrong impression as it was "divided" less, not more. Modern CPUs can have up to to 16 way set associative caches.
I generally miss the times when PCs did not need cooling, especially the CPU.
It would have still been helpful, even to those who weren't overclocking due to limiting heat exposure and a prolonged life, so if you use a 286 today or higher, put at least a heatsink on or it may fail in a couple if years if you're enthusiastic enough to use one. Otherwise if it's below for example 10 MHz it wouldn't matter!
@@jammehrmann1871 AMD DX2 66 can work without any cooling AFAIK
@@RuruFIN WTF DOES AFAIK MEAN and I wasn't stating that it would be mandatory but it would definutely help while overclocking or preserving the lifespan
@@jammehrmann1871 AFAIK = as far as I know
@@RuruFIN thx btw did u understand what is was saying
My 486 had a Coast , or Cache On A Stick i think it was L3 cache and boosted performance almost doubling speed
Still an L2 cache; just with the convenience of being on a stick instead of individual chips
This sums up why an early Pentium kicks ass for pretty much any DOS game you throw at them.
Too fast on a Pentium 133? No problem, just disable the cache and you're on a level of a 386/486.
But early Pentium (Socket 4) sucks and is often outperformed by 486DX4 in everything but Quake. Those are the 60/66 Mhz heaters with the FDIV bug. I agree P133 is a wicked good CPU but that is mature Pentium era, far from early.
I still use my Compaq Amanda 7370. Basically just for email and the occasional word processing task, most comfortable computer to type on ever.
You'd be even more comfortable typing on the keyboard.
I loved my i486DX4-100. I used it up until 1997 after which I built a new system based on an AMD K6-233. The K6 was a pretty decent chip and I would run benches against a P-233MMX I had as well. The K6 would beat the P55C in integer calculations, but the intel would beat the K6 by a small margin in floating point calculations. I even used the same Socket 7 motherboard for both cpu's. Fun times back then.
Small margin? Dude, the P233MMX is almost 50% faster in Quake.
Informative video.
Cache Associativity can be a pro or con. Normally the CPU (back in those days) will have to iterate through all the cache line to find something. If the caches sizes increased this caused significant slow downs. In a N-way Associative Cache the data you are looking for can only be in N spots in the cache. There for you will not need to iterate through the whole cache. The major con with this is if the associativity is low, you ran into a lot conflict misses where data you need a few instructions later could be evicted from the cache to make room for more urgently needed data. Going to RAM to re-fetch the data was very slow.
there's just nothing better than seeing a 486 computer running emm386.exe. with a full contiguous pageframe address without any issues. This is impressive for those day's good luck now day's as those fancy graphics card's tend to utilize and populate those memory addresses with uma memory so yah your pc is cheaper for a reason. loss of full functionality.
Speed, more transistors, and cache? Wait, wait...it's two 486 processors?! I didn't know that!
My first computer was a 486 @ blazing 25 Mhz 2 mb of ram and 75 mb hard-drive. My phone today vastly outperforms the NASA supercomputers of my childhood. Love technology :)
I got an intel ad before this video...
This whole video is an Intel ad. :D
@@GiuseppeGaetanoSabatelli Or an ad for adblock, etc. Can't remember the last time I saw an ad.
Whatever was the iComp index and why don't we still use it. Take it that it gave you an idea of its performance did it?
in this video we understand cash is everything
i had a 486/33 that had a 83mhz pentium overdrive chip. im wondering how good that actually was
Not all 486 use write-through, later 486's do use write back cache. In addition, a Pentium processor executes two instructions at a time only if certain conditions apply, so the code must be optimized for it, which was not the case back then, so early pentium chips were not really much faster than high end 486's.
Excellent comment. When I upgraded from my DX4-100 rig to a Pentium 75, some games ran slower... and this was not remedied until I upgraded the Pentium 75 to a Pentium 100!
Your explanation of what a 2-way set associative cache is doesn't seem correct. It means that any address in DRAM can be stored in 2 different places within the cache. The processor will use the least-recently used of those 2 locations to store the incoming data. This is better than a 1-way set associative cache because it means fewer recently used cache locations have to be evicted to make room for incoming data.
The Cyrix 5x86 processors were pretty far ahead of their time for a socket 3 processor, while the later Intel DX4 / POD and AMD 5x86 133 mhz models held their own in the early Pentium days particularly if you had a motherboard that allowed you to enable the write-through cache mode that most of these CPU's supported. Cyrix actually quit selling their socket 3 processors out of fear it would their 686 class processors. I'd argue the Pentium II in it's quirky slot-1 form factor on the new (at the time) ATX form factor is a far more interesting leap forward in computer standards than the socket 7 / Pentium which was really just a juiced up 486 with many of the same old warts in terms of motherboard support/standards.
some 486 CPU's use write-back though in my experience it can be a tad unstable depending on the motherboard. I think the 66mhz dx2 was the first 486 to have a write-back variant produced. There are also some quirks to a 486 using write-back. generally if you have write-back enabled on a 486 any SCSI controllers you install won't work. Its due to some sort of conflict that occurs but I don't know the specifics. There are supposedly a few SCSI controllers cards that were made with a jumper that will correct this issue.
+0blivi0n100 ah, while reading memories flooded back of that old DX4-100 with write-back cache we upgraded to after an am386DX-40 which was also a beast in it's own right, but got served so damn hard by that 486 monster. It felt like that fastest thing in the world back then, I wonder what I can push it to still do today. maybe a low-res divx movie if I'm lucky? lols
+interlace the 386dx-40 was indeed a monster of a 386. You'd probably be surprised what a dx4-100 could do especially if something is written specificly for it. Heck, I recall a video somewhere where they got a movie to play on an 8088. If you really want to push it try popping on an AMD 5x86 133mhz (its not actually a 5x86 but a supped up 486) and then overclock it to 160mhz which most of those chips can do easily. That's a monster 486. Its roughly equivalent to a Pentium 90 or 100mhz except when the software was optimized for the Pentium or made heavy use of the FPU.
0blivi0n100
I had something close! My dad (i was 11 then) gave me my own cyrix 6x86 150+ that had it's weakness and strength. for retro sake built a p166mmx dos-only rig years ago that did websurfing, cd burning, mp3 and divx playing (all from dos) awesomely :) it had a voodoo5 in it. Wish I hadn't given it all away.
nice, id say a V1 or 2 is a better fit for a p166 in DOS but that's still a nice rig. v5's are getting pretty scarce and more sought after.
Well the V2 would be better with a pentium 2 or even an early pentium 3. Using this with a pentium would be underpowering the V2 (even though, a V2 is still faster than a V1 even with these cpus)
So how is this different from a dual-core?
Multi-Core Processors differ from the Pentium because each of their Cores are in essence bare-bones Processors in their own right; and thus can (if developed for properly) operate in almost complete independence.
The Pentium Processor added a Second ALU (Arithmetic and Logic Unit), providing a Secondary Data Pipeline... the Super-Scalar Unit (commonplace on most other contemporary CPU like the more popular Motorola 68K, IBM PowerPC, HP PA-RISC, MIPS, etc.) essentially then acts like a File Clerk, providing Tasks for each ALU to Complete.
This allows for up to 2 Threads (Operations) Per Cycle instead of the 1 Thread on the i486. In Theory this meant twice the Performance, but in practise due to limitations of the x86 Architecture this actually resulted in > 1 but < 2x performance (iirc on average it was about 60% Faster)
en.wikipedia.org/wiki/Intel_8086#/media/File:Intel_8086_block_scheme.svg
If you take a look at that (the 8086, the earliest incarnation) ... on that Diagram (9) is the ALU, which basically does all of the grunt work of the Processor. The (5) Instruction Queue was also split with a the addition of a Superscalar Control Unit, that would determine in Pentium which ALU was to be used and in Pentium Pro would Re-Order the Instruction Stack so that both ALU were finishing their Operations at as close to the same time as possible.
One of the problems with the Pentium Design was while it allowed for Superscalar Threading, it didn't allow for Independent Threading (like the 68040 and 060 Superscalar provided) ... thus how the 68060 works is much more similar to how Modern Multi-Core Processors do, albeit with Shared Registers.
super-scalar pipelines in the same core are fed by the same decode unit and can share registers. They also share the same cache (which feed fetches for both instruction and data registers). This means superscalar is cheaper to implement than dual-core because multi-core has to have cache coherency, meaning a write to a memory address in one core's cache must invalidate the data in the other core's cache if it is there, this means a lot more wires and transistors. Most multicore CPUs' individual cores are superscalar themselves because it is an effective way to execute more instructions per clock.
Back when Quake came bundled with a wingman joystick!!
Mark minasi. I remember he wrote a book about harddisk and I used to read them. Oh old times
I've got a book (somewhere) about hard discs. Back then, in DOS days, you didn't know what "information" you needed to know, so you tried to learn everything. That's why I'm now soft in the head.
1:40 laserdisc with disc rot?
Acpi is permanent at irq9 on pentiums. It is bad for couple of isa hardware, such as sound cards. +1 to 486. Speed of pentium is too high for many old games. +1 to 486
Wasn't a 486 133 released by AMD faster then the first gen Intel Pentium 60?
+Slavomir G yes, because they had a bigger cache size than the first gen Pentiums.
The fastest one was AMD 5x86 160MHz, it's still an enchanced 486 despite its name.
Thank you for calling it the correct term cache ("cash") rather that than the incorrect term cachè ("cashay"). Very few people nowadays use the the correct term nowadays when referring to a processor's cache (L1, L2 or L3)
I have a Pentium SX 994 at 120mhz and unfortunately I do not have any socket 7 motherboard...
I would've bought a PR- equivalent K5 to whichever Pentium I was buying, except I was buying a Pentium-200 and the K5 wasn't exactly widely available where I live, let alone a K5-PR200 which was rarer than hen's teeth.
Yorkshire, then?
Super scaler architecture froam soviet elbrus. Vladimir Pentkovsky worked on all of them pentium.
First time I gave thumbs down to your video - sorry, I'm not worthy" I'm not worthy"- but I really expected more full analysis than the cache thing. Like the bus-speeds and other things, like what CPU extensions, like better pipelining, new instructions, etc. it had?
Btw, was it Cyrix that made these "486" or "Pentium" (not sure of latter) that basically were 486 (and Pentium possibly) that supported higher speeds and new features, nut were limited to older architectures hardware, like BUS sizes, etc? Or some other Company?
Does someone know the Music name?
l1 to l3 are on die set now.
cache me outside. how bout dat
lol i see what u did there
Those bloody bastards... you realize you're basically telling me the pentium was already the first "dual-core" multithreading cpu 20 years ago?!?!?! and if L1 cache is a type of ram that can run at 4+ghz on overclocked i7's, I start to wonder if it shouldn't 've beem technically possible years ago to have regular ram at those speeds, and tech release is being held back way too much to maximize profit.
+interlace Yup! Tell me about it XD
What i don't understand is how they kept the cpu running only one thread despite multiple pipelines
Its actually not that hard to understand.
Think of a processor with 10 different pipeline steps. The first step fetch instruction, the last step write the data back. There is 8 steeps between doing... stuff.
It don´t really matter what the instructions between do. So for the pipline controller it think its a 10 step pipline.
What it really is is two 5 step muxed pipline. So the first instruction goes to the first pipline, the second, go to the second, the third go to the first again, the forth go to the second.. and so on.
The result is that for every clock the pipline does, it make two calculations.
matsv201 that's assuming there are no data dependencies
The explanation is kind of faulty. Its just the pipeline that is double up. No the entire processor.
A modern Zen processor have 4 ALU, 2 AGU and 4 FPU pipes (some with SIMD). It got 6 operation dispatcher. So the 6 operation dispatcher have 10 pipes in total to chose from.
It also got two operation internalizer (jump, branch and that kind of stuff) and 4 instructions internalizer (calculate, compare and so on).
So it can initialize up to 4 instruction 2 operations per clock and it can in theory do 10 pipes. But because it only can initialize 6 processes, it can only do 6 processes per clock. More realistically it can probably do 5 with near perfect code. It should not jump under 4 if the code is half decent. But of cause if there are a lot of branches it will really go to hell.
This is done on two threads, but in one core. The CPU will have (usualy) 8 cores, so there would be a total of 32 ALU and 4 FPU in a 8 core processor.
What the Pentium did was just have two pipes and run every second instruction on the even one and every other second on the odd on in the same thread, this way it kind of "emulate" a double clock speed with double the pipeline length. The draw back is that the simulated pipeline become really long.
The upside is that its really simple making code for it, you only need one threed. The down side is that the branch prediction is really hard. That was what broke the P4..
Basically for a Pentium the CPU does two instruction per clock per pipeline steep. For 486, they make one per clock per steep. For all older they do one instruction for every two clock per step.
So for a 486 running FPU code it make one 32 bit calculation per clock.
A Zen can make up to 4 FPU calculation per clock per core 128 bit (or 4 x 32 bit) per core. It also have say 8 cores and a freqvency of say 4GHz.
So the frequency is about 40 times higher.
It can do 4 times more instructions
It can do 4 floating-points per instruction
It got 8 times more cores.
So it should be about 5000 times faster running only FPU code. In reality its more like 1000 times faster or so.
This is even true for mobile CPU:s. A ARM A57 can do 2 instruction per clock and 128bit instruction wide. Thow it is a RISC so it "loses" quite a lot of instructions.
L3 cache?
OK but i miss my 486 sx25 and DX2/80.
Still?
“Algorivvum” 😂
U and V are coordinates that define a Parametric Surfaces in 3D space or UV used in in 2D Texture Coordinates.
Hm, so it's basically a hyperthreaded 486...?
Not really, though smt is a way to keep the pipelines busy
Because it Superscalable
This sounds an awful lot like dual core
It's not. It's actually quite different. While a dual core chip can operate on two concurrent threads at once, a superscalar cpu can only operate on one thread at a time.
The difference is that there is extra hardware that tries to bunch up several instructions to be executed at once.
Dual core???
calvinatorzcraft no. It's very different
that's one worn pentium chip lol
"Well loved"....
Why Pentium Kicks 486 Ass - XD
Pentium 60 vs 486dx4
try it and you will see.
In quake, P60 will win, anything else - 486.
I would never use a 60/66 Pentium because of the fact it often cannot even do division correctly.
😲
8GB Cache
hi