PCI-E was not "invented"! It was DEVELOPED!

Speaking about GPU - you forgot GMA3600, quite ok graphics blob for which they never managed to make proper driver with 3D support.

@@NickShl Invented means to "design a new process or mechanism" whereas developed means to "change with a specific direction" PCIe was invented, since even though PCI existed, PCIe was something new, not just the old but different

Lol

Came here exactly for Itanium :-P

Hard to believe they didnt call that one out.

same

@Daniel Leone cringe bot

Same

Willamette

It wasn't just the pipeline. P4 couldn't abort speculated instructions until retirement, which made it even worse, culminating in mispredictions that could end in 100 cycles of latency. Between that, the slow L2 cache, and the unorthodox L1 trace cache that replaced the i-cache, the P4 was blazingly fast right until you did anything it didn't expect, at which point performance dropped like a rock.

The other issue I remember with Netburst was that, to make branch prediction better, they dedicated a bunch of silicon to it, which put a ceiling on the speed the CPU could run. I remember people using Pentium Ms in desktop machines using socket adapters, then overclocking the bejebus out of them to get better performance than "faster" P4 chips that couldn't clock as high.

@@jbmcb Well yes. Which is why the Pentium M formed the basis for Core...

I remember having a 1.2 Ghz K7 Athlon back then, the very reason P4 existed the way it did. AMD just had such ridiculous performance AND clockspeed that Intel was back to the wall with the P3 performing well, but not that well, so they tried discrediting clockspeed as an important factor (despite being an important factor at the time) and made the P4 with its huge pipeline and branch prediction to get a higher IPC and try to prove their marketing right. But like car-junkies say: There is no replacement for displacement!

@@builder396 That's not true: P4 was very much designed to clock up high. Intel's marketing at the time was claiming that their CPUs would be clocked at 10GHz by 2005.

Only slightly less well-known than "never get involved in a land war in Asia"?

Existing???

It was the first thing mentioned. iAPX 432 was itanium.

@@MasterGeekMX No, that has nothing to do with Itanium... that was the alternative to AMD x64

@@acubleywhich is even slightly less well-known than never go in against a a Sicilian when death is on the line

Although IA-64 never made it to desktop, it had a run of over 18 years in a certain niche of the server market. I don't know if we should really consider that a failure.

@Daniel Leone rip bozo

Ah yes, the Itanic. Then AMD did a Slick Willie and published x64.

Nowadays IA-64 would be a good idea since it would cut down the instruction set quite a bit

I was shocked that this didn't make the list for the video...

Didn't the Pentium 1 60 and/or 66 Mhz models overheat to the point of melting or catching on fire?

@@ki5aok None of the processors in that generation had any thermal protection what so ever. So if you powered up the computer without a heatsink attached to the CPU they would cook themselves within seconds. Same with if the fan on the CPU heatsink died. Unless you had installed a software that could monitor the CPU temp it was very possible the processor would destruct before you knew anything was wrong. Though usually the processor would crash whatever program it was running and that would cut down on the heat enough it wouldn't kill itself. But if the he heatsink fel off, and yes it could happen if it wasn't installed properly, then the temp would spike and the CPU often would toast itself immediately. Up to the 386 this was not really a problem as they were so low powered that they didn't need a even a passive heatsink. The slowest 386 processors were also able to run without a heatsink, but they were getting toasty. 486 and onward the heatsink was mandatory for proper operation. The Pentium 3 was when the instatoast feature really took off. Up to and including the Pentium the processor was enclosed in a ceramic material. The Pentium 2 was a cartridge design with a metal plate acting a heat spreader that the heatsink was attached to. With the Pentium 3 they went back to a PGA socket but mounted the CPU core to a organic substrate much like a PCB and they had no protection for the core, but the heatsink was making direct contact to it. This also meant that the thermal mass of the CPU was minimal. That in turn meant that without a heatsink of some kind attached the CPU would self destruct in seconds. With the Tualatin revision Intel introduced the IHS. That was an attempt at cutting down on the RMA returns. The exposed die on the Coppermine was great for cooling performance, but you could crack the die when installing the CPU heatsink. It also increased the thermal mass significantly.

Its not that big of a Dud, compared with some later duds... i915/925 is a good one too...

@@ki5aokno, it wasn't that bad. they certainly ran hot, but the bigger problem with that platform was the incredibly buggy 430LX chipset.

@@lifesman1234 I heard it was, but could never confirm it, so I couldn't tell if someone was exaggerating or not. Thanks for setting the record straight.

now AV1? history repeats itself

interesting, I have some beige computers i could make some voodoo SLI systems out of but Ive actually collected 3dfx literally when they were considered useless way before they got bought out.I have 3 or 4 boxes of 3dfx cards,some voodo 5's,4's,3's all the way down voodoo rushes etc....I got like 150 or so cards I got for like dollars each and I kept them cause the voodoo 2 1000 was my "first" graphics card. I had bought a sis 6326(which I still own) cause Im old AF and I just wanted to play motocross madness then got a voodoo 2 later at gamestop for 30 bucks then was blown away by its performance and started collecting 3dfx cards since then way later when they were considered shit.I have a lot of really rare cards not even from 3dfx.I have a few ATI Fury MAXX cards(the dual gpu ones) a few non working voodoo 5 6000 cards(wish I knew the problem) and a quantum 3d aalchemy system.I forget which model but it has 32 vsa-100 chips in it which is like the voodoo 5 6000 that has 4 compared to the quantum 3d's system with 32. but performance is literally no different from a voodoo 5 5500 cause games barely even scaled back then after 2 gpu's with glide

@@richieqs7789 let’s hope history doesn’t repeat the part where intel leaves the gpu market and drops GPU manufacturing :(

I had an 8mb version for a while to play cs1.6. Man it hated fog effects lol. My current p3 retro system has the igp version, the 810 with 8mb of buffer memory on the motherboard. I plan to swap it out with a version of the motherboard that has a TNT2 16mb on board. Also, yes, the video support was great with the 810 and playing DVDs.

That's why I had a matrix mystique with rainbow runner

They had to do everything, EVERYTHING, to make sure the branch prediction failed the least possible. And even then, it couldn't win against Prescott's 31-staged pipeline. This is actually really sad.

Like a car that’s good on the straights, not so good on the corners. Maybe OK if computing loads are like USA roads.

i remember the P4 fetch stage took 20 cycles. AMD athlon XP at the time took 11 cycles. considering that on average there is a branch every 5 instructions, its a huge difference, and that is just the fetch.

@@tonyv82 I find very bizarre how we are using 15 or 20-staged pipelines in modern processors. Considering how much better did branch prediction, execution units, memory bandwidth and general architecture efficiency have all gotten better throughout the years, it appears that this is the sweet spot, and pipeline cleaning and fetching new instructions shouldn't hurt performance as much as it did in the past. Guess the first 20-staged P4s tried to implement this "sweet spot" in a very rudimentary technology and design. As for the 31-staged P4s, no. They don't get any excuse.

Even the later Pentium 4s had a bad reputation. Especially the Prescott ones, which were so power-hungry and overheated so much, that they were sometimes called "PressHot".

kzfaq.info/get/bejne/j7eEhtt9urjdqXk.html

and the iapx432

Rambus, haha. I remember almost getting sucked into that. Ended up getting an Athlon 64 instead. On a side note, I recently discovered that both Nintendo and Sony used Rambus memory.

Rambus was a massive mistake. It only had a 16-bit data path, too. Intel never saw DDR coming. I was wondering what happened to “Timna”, and it turns out that was a casualty of the Rambus fiasco, too.

I mean, at the time, Rambus did seem to be something interesting, albeit stupidly expensive. If memory serves (pun intended), they were using it in both N64 & PS2, and generally a P4 with RDRAM would be just a notch ahead of the DDR P4 in the benchmarks. It was just 3x more expensive for the RAM, and the i850 RDR chipset was like 20-30% more expensive than the i845 DDR one that eventually appeared after Intel realised how they had failed themselves. The greedy Rambus patent trolls must've been so happy back then.

I had forgotten the degrading performance of the SATA ports ... Bad luck for a friend of mine that got a recommendation for his new PC from me. Never asked me again for advice :D

Even if its going to be a racing pedal with haptic feedback and realistic car crash feedback (with a build-in rebar shooter)?

@@GeorgeWashingtonLaserMusket I haven't done any research, because my first PC was a pentium 4 on ME 😄 the second was a core 2 with 2 cores and everybody was losing their minds

@@GeorgeWashingtonLaserMusket Lol, China certainly has smaller nodes.

@@GeorgeWashingtonLaserMusket Intel released the Pentium 4 too early and half-baked when their infamous 1.13GHz Pentium III had to be recalled in 2000. In the end the architecture was utter crap even in its later years and ultimately Intel went right back to their PIII architecture with the super successful Core 2. Up until that point, it was all AMD or nothing for almost all usage scenarios between 2000-2006.

@@AlDim000 I wouldn't say it was all amd or nothing, the pentium Ds were self melting power hogs, but they were OK for the most part.

@@endless2239 The Pentium D was the first somewhat acceptable Pentium 4 CPUs they released way late in the game and they were completely obsolete the second Core 2 was released. AMD was king until the Core 2 launched. Pentium 4 was just inferior technology, but Intel had no choice but to use if for over half a decade.

Yes. Back in 1994.

@@Pandamad That was the original pentium, still was a success though

en.wikipedia.org/wiki/Pentium_FDIV_bug

@@Ashitaka0815, it was costly to recall the affected processors and they potentially lost market share to IBM and Cyrix.

@@AndrewAnthony2525 That's exactly why they hidden it even when they already knew about it. Still, it felt lie when the 166 version was released, most people either forgot about it or didn't even noticed in the first place, not like today where it would pop up instantly on your smart phones news feed.

i7-9700K is WAY worse than a 11900K (at least it did an IPC increase). i7-9700K nowadays is a stuttering fest chip on games because Intel tought it was a brillant idea to remove hyperthreading, has the same core count as a 9900K but *you lose a significant amount of cache AND HALF the threads* .

@@saricubra2867 Oh boy, yeah I didn't think of that one. I know someone using a 9700f (close enough I guess) with a 1660 Ti and he still sometimes reaches CPU bottlenecks despite the lower-midrange GPU.

@@XxGorillaGodxX The worst part, the i7-9700K CAN lose against an 8700K which is the 6 core 12 thread predecessor and Ryzen 7s with 8 cores and 16 threads at the time. Now THAT is truly a Bulldozer moment. Chips with more cores losing against chips with less cores from previous gen. i9-11900K despite having 2 less cores than a 10900K and 10850K, has higher 1% lows and fps on the extremely CPU intensive Battlefield 2042 on the 128 player mode. It's a bad flagship, but not a bad gaming CPU which was the marketing intel pushed at the time for it.

@@saricubra2867 9th gen in general really wasn't very forward thinking. The i5s only had a measly 6 threads compared to Ryzen 5's 12 while barely giving more gaming performance (and being an absolute mess in newer titles), and the i3s were just pitiful. As you said, at least 11th gen had an uplift in single thread performance, but 9th was not only a side-grade from 8th gen at best if we look at i3 and i5, but it was flat out worse if we compare the i7s. With AMD in the equation, things only got worse for 9th gen. It *was* probably the worst gen in hindsight, especially with the massive improvement 10th gen was straight afterward, doubling threads on every sku aside from the i9s.

@@saricubra2867 idk what you're on about, I've never had any stuttering in any game with my 9700k. Sure it ain't great, but it's going on 4 years old now and still keeps up with new gens.

Ask and ye shall receive: kzfaq.info/get/bejne/j7eEhtt9urjdqXk.html

The Atom was actually incredibly good, as long as you didn't try to juice ridiculous shit out of it and kept your expectations realistic.

Yes. The problem with the floating point division.

My first computer had a pentium 4 running at 3ghz with hyper threading. I should still have it somewhere

ahh yes, the infamous "waste of sand"

I agree. Compared to the Ryzen 9 5900X/5950X, the Intel Core i9-11900K was an embarrassment at worst that was uncompetitive at best before returning to proper competition with Intel 12th gen.

Most notably, it could be *two times* as fast clock-for-clock (so in practice more than 2.5x due to the higher clock speed) as a P3/Athlon with purely integer calculations (which admittedly was a situation that didn't happen often).

@@Olivyay I think it's due to the dual-issue ALU (which might have been used later for HyperThreading). However the first generation of P4s (Willamette) was bad enough that it was slower for gaming than the last generation of P3s (Tualatin). Pentium 4s really began pulling ahead steadily only after the Northwood-based models were release, especially those above 2 GHz.

Ahh, the High Temp edition

I had a Riva TNT 2. Good old Harry Potter and the Sorcerer's Stone PC worked fairly well. Then Harry Potter 2 stuttered like hell. Good card nonetheless.

Dell Dimension 4600. Those have an AGP slot.

Lol

You don't know about the Blue man ads? The Intel Pentium marketing campaign in the late 90's early 00's was the most successful CPU marketing campaign of all time. A lot of people just think Intel gained such large market share due to their OEM shenanigans alone. But those Pentium ads made sure almost every single person in the US and Europe knew the name Pentium. Even your Great Grandma (born in 1910) who had never touched a computer before knew what a Pentium was and those ads cemented Intel in the publics minds for decades to come. For most people Intel was the only CPU that existed.

@@McShave ah mate I was born in 97, so it was a bit early for me to really notice those things haha, the most I remember were ads with the Intel jingle that I liked bc the sound was cool to me

The original Pentium 60. Floating point bug. Great times when you could only afford the AMD K5 or worse, the Cyrix M1

It was the early Pentium, and it was not that it "couldn't add" (such a bug would have been spotted immediately in development) but that it *sometimes* gave incorrect results with high-precision numbers in floating point division, which really only was an issue with scientific computations. It had zero impact on typical consumer use (it took more than one year for the effect to be discovered by an academic).

@@Olivyay Wasnt there also a specific thing you could do in Excel to trigger the same bug?

@@builder396 There was probably a way to reproduce it but at least nobody spotted the problem before some researcher checked his calculations.

It's not like they didn't work, it's just that they needed double the power and also produced double the heat for the same results.

They're all Netburst as far back as Willamette. They're a totally different architecture than Pentium III, and they didn't go back to using P6 as a base for new architectures until Core.

Because it's not about Intel's worst failures, but Intel's worst products. The first release 1.13 GHz P3 was still a P3 and therefore a good product, just pushed too far too soon. Admittedly, the Northwood P4 was not a bad product either, as it had a better balance of performance and power consumption than either the previous Willamette and the subsequent Prescott, and was also competitive with the Athlon XP.

@@Olivyay The 1.13Ghz was unstable and unreliable to the point of Intel pulling all of them for a year and urgently pivoting towards an entirely different and unproven architecture to be able to have processors to sell. How can you describe that as a "good product" lol?

@@AlDim000 socket 423 P4s with 800Mhz RIMMs, early Athlons were a lot faster with cheap 128-pin DIMMs, even Durons and some late P3s (Tualatin) surpassed those early P4s

I really thought he was going to bring that one up before he said P4

That and the 11700k were just sad, but the i9 is worse for having the exact same amount of cores, sucking up a ton of power, and only being competition for the Zen 3 Ryzen 7s in practice.

I think you're conflating the original Pentiums that came out in 1993, with the Pentium II Xeons that came out in 1998. The OG Pentiums did run hot for their day, but they only had a 16 watt TDP. Nowadays, only laptop and tablet chips consume that little power. The Pentium II Xeons did come on massive cartridges, but that wasn't because the chip was massive, but because the cartridge was a complete card with both the CPU and a very large, and very fast L2 cache chip. There really wasn't a PowerPC equivalent to the use case those were designed for. I do find the whole RISC vs. CISC thing silly too. With the Pentium Pro (1995), it and all x86 chips since were themselves RISC internally with an instruction decoder that breaks CISC x86 instructions into a simpler RISC-ified x86. You still lose efficiency doing that, as the Apple ARM chips show today, but not much, especially as Motorola and IBM couldn't maintain the efficiency of PowerPC.