So, did not watch the video yet, but I suppose "it depends on what you do with your PC", lol. Depends on how multi-threaded the specific thing you're using your PC for is. For example Maya and 3ds Max are multithreaded but not very well still (even in 2018 yeah), so single core performance matters for viewport performance. ARMA3 servers are also poorly multithreaded, so single core clock speed is important there as well. ETC.
The design of a cpu is truly amazing. we take for granted how much stuff a cpu has to do to compute very complex task all within microseconds. it's truly amazing what people can do when we work together and create something soo complex like a cpu.
Speaking of dropping: Massdrop! Massdrop makes products with input from their members and give them a place to connect, learn and shop with people who share their interests
If it is 32bit then it could be great for old games/programs though I doubt there will be a large market for it without support for the old PCI and AGP slots.
I'll never forget the days when my Intel 120 MHz CPU. When I tried to run games, I get a message says "Your Processor doesn't have MMX technology" or some like that. 😁
Heh, I had a 133 Mhz Pentium, and somehow managed to make it run a game designed to be run at least by a Pentium II, so what ended up happening was that the game ran, but proportionally slowed down. It wasnt a problem as long as I was racing against the AI, but the third race was against the clock, which, unlike the game, ran at normal speed. It was impossible.
Reminds me of the moment when i tried to run a p3 small pc game on my p1 pc and it loaded completely but it was rendering the 16 bit hi rez graphics in 256 colors frame per minute. Later, I was playing with disk cleanup utility from win95 on the same pc and deleted the whole windows folder. Got a really nice message on restart.
"Megahertz Myth" was a term used in marketing by Apple! This was back when they were on the PowerPC CPU platform, and they and their partners (IBM and Motorola/Freescale) had put relatively more work into increasing the instructions per cycle, compared to cranking up the clock speeds, rather than the opposite priorities from Intel at the time. They did product demos in which, for example, a 500 MHz G4 Mac completed a Photoshop script even faster than a 1 GHz Pentium III machine. So, what changed? Well, Intel began to really master the instructions per cycle, too, in search of faster mobile processors with low battery consumption. Meanwhile, the PowerPC makers had stalled out on clock speed! Intel's upcoming products were so much better at performance-per-Watt, that Apple brought out the Intel version of Mac OS X they had going on in back rooms.
That's some interesting history! The first time I noticed differing performance despite clock rate differences was when I had an AMD Sempron 1.75GHz overclocked at 2GHz beating my friends 2.6GHz (might have been 2.4 or 2.5) Intel Celeron. But then the Core2Duos came out and I think a 2.4GHz dual core was equivalent to my new AMD Athlon 3GHz dual core... never really looked into why and just generally assumed intel had the better technology or implementation at that point, something that seemed to be the case up until Ryzen I believe. Stuck with intel and nvidia ever since and never been disappointed, the i7 13700K I have now is undervolted by 0.200 and underclocked and never uses more than 20% in games it's crazy, wish graphics cards were that far ahead
yes, almost all terms were mentioned which doesn't give you any deeper insight. You're probably just joking 😃 I like the video, it's very well explained but the scope is (luckily) not comparable to any lectures about the topic I attended.
There are some small time retailers (who definitely have NO IDEA about PC components) that list CPU frequency MULTIPLIED by the number of cores. If it's a 3GHz 4-core CPU they will advertise it as a 12Ghz! CPU lol
*releases video, about processor with just higher clock speed* Linus: y’know what would be a good idea.. *releases video about importance of clock speed*
@@problemschueler Well in comparison they were weaker, but also a lot cheaper. You'd get 25-70% more performance on a i5 3570k compared to a fx 4300, but the fx 4300 was 80$ while the i5 3570k was 240$, so 1.25x - 1.7x more performance for 3.0x the price. The fx 4300 had 500 million transistors for all 4 cores/2 modules, the i5 3570k had 1.3 billion transistors for all 4 cores. 2.6x more transistors for up to 1.7x better performance, I'd say quite good performance per transistor. And up until 2016-2017 the FX CPUs could play pretty much 99.99% of the games 60fps+, rare exceptions were the ones that even intel CPUs were struggling. So they were indeed weaker in comparison but good enough. The problem of the FX CPUs is that they went with many small cores, if the highest core count was 6 cores/3 modules, but the cores were 30-33% bigger with probably 5-10% lower clock speeds, it would have been much better and would have costed about the same since the die size would be consistent.
@@badass6300 that 80$ is easly 200+ these days because it's soo shit that most people probably changed their CPU by now while Intel users are still waiting for performance gains worth the money. Also meanwhile electricity wasn't free and it might look like in reality Intel indeed WAS a better choice
@@badass6300 and you are comparing fx 4300 to Ivy Bridge (3rd gen) unlocked i5 ao your argument is even more invalid. AMD during FX era was just shite. Deal with it. In the long run Intel was MUCH better performing, cooler, pretty much didn't use any power comparing to the arc furnace of team red and like I said: actually looking today was even a cheaper option! Today tho I pick Ryzen 9/10 times
That cupholder example totally missed the point. Clock speed is like engine's maximum RPM, a rotation of crankshaft is like a CPU cycle and it sort of takes part in defining final performance but almost nobody buys a car taking RPM into account. Also, I'm pretty sure most of Linus' viewers have no idea what MMX is.
Very informative. Thank you for explaining this. I was wondering how a modern CPU is faster without a faster clock speed? Now, at least I have an idea. For my work loads, I’ll stick a lot of cores(8+) and CPUs with mid range clock speeds.
yes, when comparing the same architecture, clock speed means a lot. a pentium 4 at 3ghz will get wrecked by a modern 3ghz cpu though, and that's just assuming single threading on both. when you get into multithreading, it depends on how much of the software workload can be parallelized... amdahl's law and stuff, there's a lot of old multi-headed server beasts that only recently got outpaced by 6-core processors last year when it comes to raw multithreading. clock speed scales performance nicely, given the memory system can keep feeding it, the diminishing returns on power:performance are pretty awful though.
My old Dell OptiPlex has a Pentium 4, 3GHz (technically 3.06GHz) CPU. For such an old computer with a 1 core processor, it's actually lightning quick, and I have a gaming laptop with a 2.6GHz 6 core CPU. I can admit that the Pentium 4 cannot multitask very well. 😅
Missed a better metaphor opportunity with the car-could have said, "just like you can't determine how much power an engine produces solely based on RPM or solely on displacement" (e.g. two cylinders with 1L of displacement per cylinder (i.e. 2L displacement) running at 8000 RPM might produce the same amount of power as four 1L displacement cylinders (i.e. 4L displacement) at 4000 RPM, ceteris paribus), where frequency is to RPM as IPC is to displacement.
I remember back in 2008 when I wanted to buy a new PC to replace then-current Pentium 3 PC. My family wasn't a rich one, so the best thing I could buy is a Pentium Dual Core. But then I saw the clock speed. The Dual Core has 2.8 GHz clock speed. But then I saw a Pentium 4 with 3.2 GHz. I was a kid and I was an idiot. I decided to go with the P4 3.2 GHz. Two weeks later, I regretted everything.
@ateb3 that it is but it is still a security exploit because of just how it works. Everything has to be unencrypted to work with as data on the CPU execution cores so they get to be so but doing so means unencrytped data. It is a security exploit that can take that unencrypted data out of the cache before being caught by the checking math built into the chip. So it is a good thing for consumers who do not have to worry about being hardware security aware but a bad thing if your computer is a possible target of attack. The "worst" versions are most able to happen on servers due to admin control being able to be given to off site clients by default for ICMP.
now to counter my own point it also means that 1 they have the username and password for the ICMP login 2 they can use a ICMP login which means they are on the LAN already or VLAN anyhow and then 3 you are already screwed by that point as in the event they have access to the ICMP login of one computer they most likely have it to more than one to which your sever(s) are screwed due to it being better to just install something that looks harmless onto the system and does minimal traffic to steal the info instead of a hardware exploit to gather them all slowly
Just saw the video. If you guys are going to mention branch prediction in CPUs, you need to mention about the Spectre and Meltdown attacks. Branch prediction opened avenues of vulnerability and required cumbersome security patches that considerably slowed down (at least 5% and more averaging to 15%) CPUs that used the aforementioned feature.
Linus overclocks i9s with cooling systems that cost thousands of dollar so that he can reach 5 ghz plus stable clock speed then says it has nothing to do with CPU performance.
overclocking and cooling upgrade is trying to make whatever run better at the lowest possible temp., comparing clockspeeds of different architectures mindlessly is another thing.. eg: i have a quad core , 1.6 ghz i5 8th gen laptop variant with hyper threading that can turbo upto 3.4 ghz and an amd fx that runs 3.6 ghz stock , thatd , if i try to overclock on air would easily hit 4 ghz at the very least.. i know the i5 is faster. even if its a low power laptop version.
GHz x IPC is the best metric, at least from a gaming perspective. 4-6 cores with max clocks x IPC is much better than any other configuration, as far as I'm concerned. Correct me if I'm wrong, I guess. But yeah, clocks are half of the equation used to find performance, assuming the workload doesn't scale infinitely with more cores. If I was a content creator, I might multiply that number by the number of cores, or maybe even add decimal modifiers for each core that doesn't reach the same clock as the primary one. (for example, 16 core cpu that hits 4GHz on primary core, and 3Gz on remaining cores, (IPC)(4)[1+15(.75)]. )
@@bueb8674 have you? If I look for the IPCs of my current cpu, a 4670K, the first resulta are forum posts asking about the performance and how much better IPC Intel has over an AMD FX, a cpu boss comparison and not much else, really.
@@ppsarrakis That's right. Since my original post I did some more research and that was the preferred method to find out, yet, IPC varies depending on the workload, so it's not a static number... Anyway, the problem with that method is that you have to own both CPUs to run the tests, and most people want to know before purchasing so they can choose better. :p
It's a monitor stand, and how many regular consumers do you know who have bought one? It's not designed for a basic consumer like you or me; the Mac Pro and the Pro Display XDR and accessories are designed either for enthusiasts or actual professionals who do things like UA-cam full time and need the processing power built into a single machine.
This clears up some confusion I've been having for a long time. I picked up a budget gaming laptop with an Intel Core i5 @2.3 GHz quad core. It's weird because I've seen i3's at over 3 GHz and i7's at 1.7 GHz. I've also seen really old computers with Vista with 2 GB memory but a 2.5 GHz Core 2 Duo or something like that. Does that mean that old machine could potentially be faster than my gaming PC?
TL:DW All the CPU's from the past 5-7 years in the desktop market are good enough for most end users. Except FX chips. Joking aside, Explanation: One major issue was the memory controller and cache. For comparison intel chips were and still generally hit 40-45ns latency speeds when sending commands from the cpu to the memory controller; and thus only slightly longer to get data from the RAM. AMD FX was hitting 65ns+. Thats near 50% penalty *per* memory instruction. so anything the fx chip couldn't keep in cpu cache (called a cache miss) killed the performance. FX chips *should* have been doing much better but were *literally* sitting around (skipping cycles) waiting for data from the main system memory. FX also had many other issues that gimped the cpu causing it to basically be sitting there with a usb cord up it's ethernet port doing nothing.
@@wangkevin3431 Well. It had 4 instruction decoders per pair of cores. So effectively 2 per core, meanwhile intel had 3 decoders per core. so when the cpu was fully loaded it started to get bottlenecked waiting for instructions. however this impacted the performance a minor amount. Next was the prediction. they finally built a branch instruction predictor on par with intel's the down side was the length it took for instructions to get all the way through the cpu, the 'pipeline' was a bit longer. see cpu's can put instruction after instruction into the 'pipeline' like linus said on a assembly line. the issue is cpu's do many different things, so the assembly line needs to be 'ready' for the next instruction to be executed. As in the predictor needs to guess the correct next instruction + data to pull to add to the pipeline. If it doesn't ... the cpu has to wait for the pipeline to basically finish and find out from the previous command what it needs. For example, data sets and server stuff can be predicted keeping the cpu busy however, user input, i.e. games, can *not* be predicted as easily thus again the cpu is sitting around waiting for things to figure out what to do and get fed data. Next is yes, the shared floating point unit that gets overwhelmed when both cores are trying to use it. this shouldn't be an issue but it was because of windows. Back in the day you could and had to manually patch windows to treat the fx chip like a intel chip with hyperthreading. That fixed windows scheduling two heavy tasks 'next' to each other on the same shared resources chip. So with that patched, windows would schedule programs like an Intel chip and space out the programs first using every even core (0,2,4,6) then every odd core. That got a lot of general performance back relieving stress on both the decoders and floating point. Finally, the biggest issue was the cache. the FX's L2+ cache was slower then intel's. So if the working data set and instructions can't be kept in L0 or L1 ... then the CPU again was waiting for data from L2 or L3 or L4. L3/4 is the RAM depending on if the cpu had a L3 on die cache as well or not. Each level is bigger and includes more cores with the highest level cache in the cpu being accessible by all the cores, it's generally how they share data with each other. So finally with the cpu waiting all the time for data and the most important subsystem it has to wait on being slower than the competitors ... the FX (Bulldozer) was pretty garbage from an engineering standpoint. For users looking for single threaded performance, intel chips at the time could perform, clock for clock, near 2x the instructions on average. That is to say a 4ghz amd core performed as well as a intel 2ghz core; in the worst case. average would be closer to 4ghz amd ~~ 3ghz intel. And in workloads where the amd's shortcomings didn't gimp it ... it was only 10-20%slower clock for clock while you had twice as many cores for much cheaper. ... surprisingly similar to how it currently is except for a important difference; AMD caught up and Rysen and Threadripper don't have huge crippling flaws. And amd is maybe ~5% behind clock for clock while they have twice the cores for half the price.
You obviously haven’t used a laptop with a Celeron processor or a processor that starts with a letter and has 4 digits after it. They still make cheap CPUs that are terrible and still would be terrible with a 8 GHz OC.
i'm still running an AMD FX-6100 overclocked to 4,2Ghz with an aio water cooler and it's stable with games running up to 60 Fps so it's not so bad and it's been running for about 5 years
Yes and no. The IPC of Zen+ CPUs (Ryzen 2000 series) is virtually the same as the current Intel CPUs (~3% slower if I remember correctly). Where AMD falls short compared to Intel is pretty much clock speed only (and a bit of latency because of CCX-to-CCX communication). So IPC is not really the issue with Ryzen, as I said, it's almost the same as Intel. The issue is really clock speed and latency. To get better single core performance than Intel, AMD need to EITHER increase clock speed, or increase IPC, or do both. You can be almost 100% sure they'll try to do both and increase both IPC and clock speed. Rumor has it Zen 2 CPUs will have clock speeds of 4.7 Ghz or better and ~15% better IPC compared to Zen. This puts it on par or slightly better in single core performance compared to i9 9900k and MUCH better in multi-core performance. Not to mention it's going to be HELL OF A LOT easier to cool because it's built on the 7nm node, not a mini furnace like 9900k which is basically a factory overclocked CPU.
@@shawnpitman876 You are thinking about gaming where you'd be right, Intel is 10-15% faster at same clocks there. I'm talking productivity where the IPC of Ryzen is basically the same as Intel (and even beats it in some tests). Check the article about this over at Techspot: www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/
they still wouldnt if Intel had their way 4 core 4 thread now because of spectre and well no reason to increase core count. AMD woke the blue giant we will soon see in the next 5 years or so if it has had a real affect or will Intel just continue as they always have.
Nice video like always 😉 Comparing CPU frequency to cars engine would more be like comparing the maximum rpm the engine can reach. That's an important information but not a performance indicator at all. A S2000 2.0L 240HP can reach 9500rpm, but il a lot slower than a BMW 435D 3.0L diesel wich devellops 313HP
Yes and no Only yes if it's a huge overclock (like 3.0 to 4.5) and if you are bottlenecking and every single MHz counts But overall the gains might not be worth extra money/temp/power usage
Yeah you're right.I overclocked my Core2duo from 2.5 to 4ghz and it pretty much doubled it's cinebench score and made games like doom and GTA V playable.
for real the most important Ghz number is the one that cancel your bottleneck (cpu), is the only way you can see like... double performance by just modify a single number, even if it is only +200/300 mhz, if with that number the bottleneck disappear.. voalá 30%+ performance appear
For the most part it’s an important stat that’s always taken out of context. But in the case of the i9, the cheaper and faster model is better for gaming... even though you should really just get an i5 if you’re only gaming. A full size truck may have 500hp, but that doesn’t mean it’s going to accelerate faster than a 300hp coupe. There’s so much more that goes into it. But just like cars, is it in your budget, and does it do what you need it to best? You can’t haul furniture and tow a 5th wheel in a Bugatti or Model S, and a half ton isn’t going to be setting any records on the Nürburgring.
Slumpdog like I said, it depends on how it meets your needs. Most mainstream and consoles adapted games run equally fine on a top end i5 for a few hundred less. Me personally, I edit a lot of photos and the occasional video with some games on the side, so I use an i7 for now, and the base model i9 when I need to upgrade.
Actually electric vehicle tows really well And don't forget that when it comes to bugatti it's still an 8 liter 16 cylinder quad turbo 1500 bhp fucking drain for fuel (easly 1k+ torques) I'd compare it more to a Japanese cars. Huge power high revving, but almost 0 torque (like motorbikes)
@@matthewprather189 yes it is , a honda S2000 has 240hp/1tq and 0-60 in 5.7sec compared to a heavier and lower powered 85 corvette has 230hp/340tq but 0-60 in 5.5sec, torque matters even more on a race track with turn because torque lets you accelerate out of the turn faster and overtake easier because torque is instant
he LITERALLY answered the question and said that he would go in detail after the first 40 seconds absolutely pointless comment, or i'm guessing you didn't even watch it
Thanks for making, yet another, incredibly informative video! 4:55 My car has ten cup holders! 😆 It wasn't a deciding factor though. I didn't even find out until after I bought the car.
It depends a lot on what you are doing, and again, depends on how much your CPU can do in one cycle. Generally speaking, more cores doesn't affect much the performance, because most things are still being programmed to use only one core, but higher clock not necessarily means that it will be better, to really compare 2 different CPU's you should see benchmarks of each one and decide which one has a better performance in the tasks you are going to use it, there is no easy way to compare CPU.
What's the difference between ""Instructions Per Cycle"" and ""Hyper Threading/Simultaneous Multithreading""? Both seem to do the same thing, from the video's explanation...
@@Lucian_Andries if you did know the answer and wanted to make a sarcastic comment. you should make that clear in the comment. dont go and be negative to someone who made a good answer to your question
@@gamingwithegoon Did you pay attention of what I actually wrote? Or you just wanted to make a "sarcastic comment"? I know what HT and SMT is and their difference!!!!!!!!!!!!!!!!! I wrote the question very clear, it's not my fault you can't read. :(
@@gamingwithegoon thanks. a fair portion of the tech community are fairly demeaning like that. i'm not sure why someone would go to an entry video on an educational topic, make a loosely related response, and then respond with hostility when clarification is offered; but it happens. i rewatched the video to see if hyperthreading was even brought up. it wasn't. idk what the guy was motivated by but he was posting on a video unrelated to one of the topics he brought up. if you understand a thing, and aren't using the videos as teaching material for yourself or others, why question in a non-time-sensitive field? i deleted the original response. if the person's question was disingenuous, then it may lead others into greater confusion through introducing new topics. if the question was sincere, and my explanation was confusing and inadequate, it similarly bears deleting. besides, his grammar in his seccond response is unintelligible to me. so it may be the case that my dialect is unintelligble to him and perhaps more generally, those who are struggling with this particular problem. gaming with egoon, you have my appreciation all the same. keep being awesome and you can be sure good things will come.
@@morthim Does nobody know how to read anymore, nowadays??! :O I asked what's the difference between IPC and HT(SMT for AMD). I DID NOT ask what is the difference between HT and SMT!!!!!!!!!!!!!!!!!!!!!!!!!!! Pay attention or learn to read, don't shit on me for your idiocy! :( *EDIT:* Yes, it was a very beautiful explanation, props to you! But it wasn't what I asked. And if you pay attention, you would see I've written it very good as well. There is nothing confusing about it, unless you don't pay attention, or the other thing... Even a blind man could understand it... I even added quotation marks, to make the words stand out more... Added some more spaces to eliminate confusion... But you guys are worse than 5 year olds! :(( By the way, you misspelled HT and SMT in that beautiful explanation... :p
Isn’t it the most important part statistic of a processor? Even core count above four cores does not matter too much for gaming, but speed does. Edit: Guess I was wrong. Interesting video!
Instructions per cylce (IPC) is what really matters, but it is rather complicated to calculate it precisely, and it doesn't take into account what kind of program you are running, so a cpu with good IPC may be really fast at calculating mathematical formulas and stuff, but it may be really awful with video or other kind of stuff. So that is why we have benchmarks today, benchmarks allow you to have an estimative of how your cpu behaves in the context you want to use it. The thing is that benchmarks are also not perfect, because a cpu manufacturer, like AMD or Intel, could just create a CPU really good at running a handful of the most famous benchmarks and suck at other things ( although I don't think this happens today), so thats why people like Linus, have test everything in different setups and different benchmarks before posting a review. Edit: actually is IPC not CPI, thanks for correcting me.
Times the CPU is waiting for data to come from the memory, or picked the wrong branch on a conditional branch, or ran a weirdass instruction that locks the CPU for a few cycles, or was unable to fill all the pipelines. It was pretty much the thing that made the pentium 4 CPU be slower than the atlhon. It had a BIG pipeline, and on every branch prediction fail, it had to wait 40 cycles, while an atlhon had to wait only 20.
@@eaglefat9398 Also an fx8320 can be overclocked to 9590 speeds (4.7Ghz all core) for £100 while a 1500x was £140 Not to mention you'll get more done on an fx8320 in 2012 than you will by waiting for another 5 years and buying a 1500x. Things have to be taken into perspective. Clock speeds dont matter IPC doesnt matter Cores dont matter They only matter when configured together in... IPC*clocks*cores/scaling/(amdahls law)
What is the best netbook that is low end but has really solid keyboard and good typing while being good enough for youtube video watching and running emulators.
Please i just subscribed to your channel please help me which should I choose for gaming between hp 15 core i5 8250u 1.60-1.80ghz 8gbram or HP pavilion 15 core i5 7200u 12gb ram 2.5ghz
So may I get a translation here. So given the information that CPU vendors give, besides base and boost clock what other key things should I be looking for in the specs sheet for any given processor? Thank's in advance. :-)
and again the answer is probably *"it depends"*
It matters if youre comparing same cpu architecture and same number of cores, otherwise it doesnt matter.
It depends on how you 💩 in toilet.
@@surge5240 i read this while taking a sht lol
So, did not watch the video yet, but I suppose "it depends on what you do with your PC", lol. Depends on how multi-threaded the specific thing you're using your PC for is. For example Maya and 3ds Max are multithreaded but not very well still (even in 2018 yeah), so single core performance matters for viewport performance. ARMA3 servers are also poorly multithreaded, so single core clock speed is important there as well. ETC.
It was a joke...
The design of a cpu is truly amazing. we take for granted how much stuff a cpu has to do to compute very complex task all within microseconds. it's truly amazing what people can do when we work together and create something soo complex like a cpu.
Multipeople processing
So correct. Its all evolution. They created one simple machine and keep enhancing it every year :)
Cpu is my Dad
The evolution of the CPU is also amazing.
@@creative2z then who created us?
I think Linus is the perfect one to explain that, because when he touches a computer *the frequencies drop* .
Drop joke in comments: check!
@@lurts9820 how much $ in the check?
I *drop* a like on this one
the frequency that the actual PC itself drops is actually a lot higher when linus touches it.
Speaking of dropping: Massdrop! Massdrop makes products with input from their members and give them a place to connect, learn and shop with people who share their interests
Wait... You guys _DON'T_ buy cars based on how many cup holders they have..?! 😓
My dad does, that's why his 2 seat pickup truck has 13 cupholders
of course, it has to have at least 2 out front, max 4. Done. If more, than you're missing out on something else.
How the hell im gonna live my life now that I have none on my car :
I think Doug DeMuro does.
My entire car is a cupholder
Soon: Intel core i1
Single core, no Hyper threading, **All the Clocks**
5THz
If it is 32bit then it could be great for old games/programs though I doubt there will be a large market for it without support for the old PCI and AGP slots.
gnihton
5THZ
150 gigawatt tdp
So the 10Ghz pentium4's they talked about 14 years ago?
Cant run any hotter than their current chips
@@tomstech4390 10? i swear intel was claiming we'd get to 20 ghz cpu's.
3:25 "CPU with better branch prediction can have significantly better performance"
*Laughs in Spectre*
Not many people care
*Curse in Meltdown*
Rogeh It does, there’s nothing wrong with that statement.
SonicSP he wasn’t rebutting you. He was referencing the spectre virus, which you should check out also on this channel
@@OfficialJervin, Spectre is a vulnerability not a virus.
I'll never forget the days when my Intel 120 MHz CPU. When I tried to run games, I get a message says "Your Processor doesn't have MMX technology" or some like that. 😁
Heh, I had a 133 Mhz Pentium, and somehow managed to make it run a game designed to be run at least by a Pentium II, so what ended up happening was that the game ran, but proportionally slowed down. It wasnt a problem as long as I was racing against the AI, but the third race was against the clock, which, unlike the game, ran at normal speed. It was impossible.
Reminds me of the moment when i tried to run a p3 small pc game on my p1 pc and it loaded completely but it was rendering the 16 bit hi rez graphics in 256 colors frame per minute.
Later, I was playing with disk cleanup utility from win95 on the same pc and deleted the whole windows folder. Got a really nice message on restart.
another way to say "im old"
Used to have a 266 MHz CPU which I could switch to 166 MHz using the turbo button. And Win95 on a 2GB HDD. 😅
@@Ultrajuiced Rip turbo button and firewire
That wasn't very cache memory of you
this comment is very under liked
Nice
"Megahertz Myth" was a term used in marketing by Apple! This was back when they were on the PowerPC CPU platform, and they and their partners (IBM and Motorola/Freescale) had put relatively more work into increasing the instructions per cycle, compared to cranking up the clock speeds, rather than the opposite priorities from Intel at the time.
They did product demos in which, for example, a 500 MHz G4 Mac completed a Photoshop script even faster than a 1 GHz Pentium III machine.
So, what changed? Well, Intel began to really master the instructions per cycle, too, in search of faster mobile processors with low battery consumption. Meanwhile, the PowerPC makers had stalled out on clock speed! Intel's upcoming products were so much better at performance-per-Watt, that Apple brought out the Intel version of Mac OS X they had going on in back rooms.
That's some interesting history! The first time I noticed differing performance despite clock rate differences was when I had an AMD Sempron 1.75GHz overclocked at 2GHz beating my friends 2.6GHz (might have been 2.4 or 2.5) Intel Celeron. But then the Core2Duos came out and I think a 2.4GHz dual core was equivalent to my new AMD Athlon 3GHz dual core... never really looked into why and just generally assumed intel had the better technology or implementation at that point, something that seemed to be the case up until Ryzen I believe. Stuck with intel and nvidia ever since and never been disappointed, the i7 13700K I have now is undervolted by 0.200 and underclocked and never uses more than 20% in games it's crazy, wish graphics cards were that far ahead
you just explained half semester of my Computer Science "Processors architecture" in 5 mints,
If you think this then you didnt pay attention in your classes
@@penguin1714 dude the only classes I ever paid attention to were computer related and programming and I am a peogrammer now lol
Except after 5 minutes you just heared all the terms and after the lecture you could draft a simple architecture if you payed attention.
@@MaEtUgR you guys don't get it, xD
besides being the diagrams,
it was actually almost everything.
yes, almost all terms were mentioned which doesn't give you any deeper insight. You're probably just joking 😃
I like the video, it's very well explained but the scope is (luckily) not comparable to any lectures about the topic I attended.
There are some small time retailers (who definitely have NO IDEA about PC components) that list CPU frequency MULTIPLIED by the number of cores.
If it's a 3GHz 4-core CPU they will advertise it as a 12Ghz! CPU lol
technically right because the CPU does 12 billion instruction cycles per seconds, but completely misleading
they do that on purpose for tech illiterate people - it's an oversimplification, but it can be said in five seconds
@@kuhluhOG not technically correct lol
I've seen someone advertise a 4 core CPU with 8 core integrated GPU as a 12 core CPU. I was upset a bit.
if it says 12GHz with a single core then it should be fine
"people are drawn to *large* numbers" some do, but we all don't like *LARGE* prices.
Apple users disagree.
@@andredeketeleastutecomplex well you were clearly drawn to large usernames
If you put the playback speed to 1.5 you can hear Linus' real, unedited voice. He is a chipmunk
*releases video, about processor with just higher clock speed*
Linus: y’know what would be a good idea..
*releases video about importance of clock speed*
Laughs in AMD FX
Well, when overclocked the FX CPUs weren't much higher clocked than Intel's Sandy and Ivy Bridge, especially Sandy.
@@badass6300 no but they sucked big time against the intel counterparts with the same clockspeeds
@@problemschueler Well in comparison they were weaker, but also a lot cheaper. You'd get 25-70% more performance on a i5 3570k compared to a fx 4300, but the fx 4300 was 80$ while the i5 3570k was 240$, so 1.25x - 1.7x more performance for 3.0x the price.
The fx 4300 had 500 million transistors for all 4 cores/2 modules, the i5 3570k had 1.3 billion transistors for all 4 cores. 2.6x more transistors for up to 1.7x better performance, I'd say quite good performance per transistor.
And up until 2016-2017 the FX CPUs could play pretty much 99.99% of the games 60fps+, rare exceptions were the ones that even intel CPUs were struggling. So they were indeed weaker in comparison but good enough.
The problem of the FX CPUs is that they went with many small cores, if the highest core count was 6 cores/3 modules, but the cores were 30-33% bigger with probably 5-10% lower clock speeds, it would have been much better and would have costed about the same since the die size would be consistent.
@@badass6300 that 80$ is easly 200+ these days because it's soo shit that most people probably changed their CPU by now while Intel users are still waiting for performance gains worth the money. Also meanwhile electricity wasn't free and it might look like in reality Intel indeed WAS a better choice
@@badass6300 and you are comparing fx 4300 to Ivy Bridge (3rd gen) unlocked i5 ao your argument is even more invalid. AMD during FX era was just shite. Deal with it. In the long run Intel was MUCH better performing, cooler, pretty much didn't use any power comparing to the arc furnace of team red and like I said: actually looking today was even a cheaper option!
Today tho I pick Ryzen 9/10 times
"You can immagine what kind of phone I'm using according to your preference"
Nokia 3310.
LUL
That cupholder example totally missed the point. Clock speed is like engine's maximum RPM, a rotation of crankshaft is like a CPU cycle and it sort of takes part in defining final performance but almost nobody buys a car taking RPM into account.
Also, I'm pretty sure most of Linus' viewers have no idea what MMX is.
A bit simplified.
But a good video. =)
I like how Linus and co have improved their content lately.
Dude you make this so easy to understand. Like I’ve learned more from you than any Teacher in my life...
Very informative. Thank you for explaining this. I was wondering how a modern CPU is faster without a faster clock speed? Now, at least I have an idea. For my work loads, I’ll stick a lot of cores(8+) and CPUs with mid range clock speeds.
yes, when comparing the same architecture, clock speed means a lot.
a pentium 4 at 3ghz will get wrecked by a modern 3ghz cpu though,
and that's just assuming single threading on both.
when you get into multithreading, it depends on how much of the software workload can be parallelized... amdahl's law and stuff, there's a lot of old multi-headed server beasts that only recently got outpaced by 6-core processors last year when it comes to raw multithreading.
clock speed scales performance nicely, given the memory system can keep feeding it,
the diminishing returns on power:performance are pretty awful though.
My old Dell OptiPlex has a Pentium 4, 3GHz (technically 3.06GHz) CPU. For such an old computer with a 1 core processor, it's actually lightning quick, and I have a gaming laptop with a 2.6GHz 6 core CPU. I can admit that the Pentium 4 cannot multitask very well. 😅
Please make a video about LoRaWAN.
Really interesting topic, I would like to know more about!
Missed a better metaphor opportunity with the car-could have said, "just like you can't determine how much power an engine produces solely based on RPM or solely on displacement" (e.g. two cylinders with 1L of displacement per cylinder (i.e. 2L displacement) running at 8000 RPM might produce the same amount of power as four 1L displacement cylinders (i.e. 4L displacement) at 4000 RPM, ceteris paribus), where frequency is to RPM as IPC is to displacement.
Short answer: Yes
Long answer: Yeah
I remember back in 2008 when I wanted to buy a new PC to replace then-current Pentium 3 PC. My family wasn't a rich one, so the best thing I could buy is a Pentium Dual Core.
But then I saw the clock speed. The Dual Core has 2.8 GHz clock speed. But then I saw a Pentium 4 with 3.2 GHz. I was a kid and I was an idiot. I decided to go with the P4 3.2 GHz.
Two weeks later, I regretted everything.
should have gone with a core 2 duo instead lol, better than both at any clock speed
I made an even worse mistake, a Celeron
I was suprised they sold Pentium 4 in 2008 at that time it was a really outdated CPU
he forgot that branch prediction lead to spectre CPU hardware level exploits
@ateb3 that it is but it is still a security exploit because of just how it works. Everything has to be unencrypted to work with as data on the CPU execution cores so they get to be so but doing so means unencrytped data. It is a security exploit that can take that unencrypted data out of the cache before being caught by the checking math built into the chip.
So it is a good thing for consumers who do not have to worry about being hardware security aware but a bad thing if your computer is a possible target of attack. The "worst" versions are most able to happen on servers due to admin control being able to be given to off site clients by default for ICMP.
now to counter my own point it also means that 1 they have the username and password for the ICMP login 2 they can use a ICMP login which means they are on the LAN already or VLAN anyhow and then 3 you are already screwed by that point as in the event they have access to the ICMP login of one computer they most likely have it to more than one to which your sever(s) are screwed due to it being better to just install something that looks harmless onto the system and does minimal traffic to steal the info instead of a hardware exploit to gather them all slowly
it mgiht be stil an security exploit but it stil has fuck all to do with wehat the video is about
I thought this gonna be some benchmarks and statistics comparing 2 vs 3 vs 4 ghz in games or editing
When gaming and/or streaming i can say overclocking my 8700k definetly improved my framerate ingame while the stream got more stable.
You missed the point of the video.
I used to have an amazingly fast 3.46GHz CPU back when that was a ridiculous lot. Except it was a Celeron D, and it really sucked.
Hey, nice 6502 at the 1:00 mark!
My 1st CPU! And best electronic experience ever..., thanks Woz
Visual6502.org
Is this the commodore 64 CPU?
The C64 CPU have some custom features, but the instruction set is the same.
How many channels do you have I actually am curious.
Really good video. Helped me with my assignment
If you play games and unity games in perticular its VERY important to have a fast singlecore speed
Also processor cache. Bigger is better.
*_Just install more RAM bruh_*
Duh, you can just download them tho.
It’s all fun and games until a gamer pirates ram
I would love to see this updated?
I had an it management ad before the video and it had linus in the ad so I thought it was the video lol
It depends.......Think of it as a an 18 wheeler Truck vs a fast 2 seater car. Both have their uses.
Yes! We must have ALLLLL the Gigahertz! 3.5 GHz, 4 GHz, 4.5 GHz, MORE, MORE!!
Just saw the video. If you guys are going to mention branch prediction in CPUs, you need to mention about the Spectre and Meltdown attacks. Branch prediction opened avenues of vulnerability and required cumbersome security patches that considerably slowed down (at least 5% and more averaging to 15%) CPUs that used the aforementioned feature.
All very well but what should you do to decide what computer to choose?
Linus overclocks i9s with cooling systems that cost thousands of dollar so that he can reach 5 ghz plus stable clock speed then says it has nothing to do with CPU performance.
overclocking and cooling upgrade is trying to make whatever run better at the lowest possible temp.,
comparing clockspeeds of different architectures mindlessly is another thing..
eg:
i have a quad core , 1.6 ghz i5 8th gen laptop variant with hyper threading that can turbo upto 3.4 ghz
and an amd fx that runs 3.6 ghz stock , thatd , if i try to overclock on air would easily hit 4 ghz at the very least..
i know the i5 is faster. even if its a low power laptop version.
@John Identical architectures, form the same version with the same microcode. But must not be the same CPU.
GHz x IPC is the best metric, at least from a gaming perspective. 4-6 cores with max clocks x IPC is much better than any other configuration, as far as I'm concerned. Correct me if I'm wrong, I guess. But yeah, clocks are half of the equation used to find performance, assuming the workload doesn't scale infinitely with more cores. If I was a content creator, I might multiply that number by the number of cores, or maybe even add decimal modifiers for each core that doesn't reach the same clock as the primary one. (for example, 16 core cpu that hits 4GHz on primary core, and 3Gz on remaining cores, (IPC)(4)[1+15(.75)]. )
How do I know the IPCs of a CPU, tho?
@@Deses try searching it?
@@bueb8674 have you? If I look for the IPCs of my current cpu, a 4670K, the first resulta are forum posts asking about the performance and how much better IPC Intel has over an AMD FX, a cpu boss comparison and not much else, really.
@@Deses you clock both your 4670k and the amd fx at the same mhz and benchmark them at 1 core,higher number has more IPC.
@@ppsarrakis That's right. Since my original post I did some more research and that was the preferred method to find out, yet, IPC varies depending on the workload, so it's not a static number...
Anyway, the problem with that method is that you have to own both CPUs to run the tests, and most people want to know before purchasing so they can choose better. :p
I hit that bell already, its how I'm notified when Linus Tech Tips, Techquikie, and TechLinked release new content :)
why was lines holding a cup with a string coming out of the bottom ? idk if you can get apps for that phone
Well, I hope it matters!
Else overclocking would be useless..
"People are drawn to large numbers"
Say that to the people who brought 999$ mac book stand.
It's a monitor stand, and how many regular consumers do you know who have bought one? It's not designed for a basic consumer like you or me; the Mac Pro and the Pro Display XDR and accessories are designed either for enthusiasts or actual professionals who do things like UA-cam full time and need the processing power built into a single machine.
This clears up some confusion I've been having for a long time. I picked up a budget gaming laptop with an Intel Core i5 @2.3 GHz quad core. It's weird because I've seen i3's at over 3 GHz and i7's at 1.7 GHz. I've also seen really old computers with Vista with 2 GB memory but a 2.5 GHz Core 2 Duo or something like that. Does that mean that old machine could potentially be faster than my gaming PC?
03:40 Thanks Linus. Now I feel old.
"High budget videogame" *shows csgo gameplay*
csgo is more cpu heavy than your typical AAA game who rely more on gpu
@@soniablanche5672 typical AAA game? lol don't tell me that your pc can only run potato games lul
TL:DW
All the CPU's from the past 5-7 years in the desktop market are good enough for most end users.
Except FX chips.
Joking aside,
Explanation:
One major issue was the memory controller and cache.
For comparison intel chips were and still generally hit 40-45ns latency speeds when sending commands from the cpu to the memory controller;
and thus only slightly longer to get data from the RAM.
AMD FX was hitting 65ns+.
Thats near 50% penalty *per* memory instruction. so anything the fx chip couldn't keep in cpu cache (called a cache miss) killed the performance.
FX chips *should* have been doing much better but were *literally* sitting around (skipping cycles) waiting for data from the main system memory.
FX also had many other issues that gimped the cpu causing it to basically be sitting there with a usb cord up it's ethernet port doing nothing.
Should have had 1 controller per "core" instead of 1 per 2
@@wangkevin3431 Well. It had 4 instruction decoders per pair of cores.
So effectively 2 per core, meanwhile intel had 3 decoders per core. so when the cpu was fully loaded it started to get bottlenecked waiting for instructions. however this impacted the performance a minor amount.
Next was the prediction. they finally built a branch instruction predictor on par with intel's
the down side was the length it took for instructions to get all the way through the cpu, the 'pipeline' was a bit longer. see cpu's can put instruction after instruction into the 'pipeline' like linus said on a assembly line. the issue is cpu's do many different things, so the assembly line needs to be 'ready' for the next instruction to be executed.
As in the predictor needs to guess the correct next instruction + data to pull to add to the pipeline. If it doesn't ... the cpu has to wait for the pipeline to basically finish and find out from the previous command what it needs.
For example, data sets and server stuff can be predicted keeping the cpu busy however, user input, i.e. games, can *not* be predicted as easily thus again the cpu is sitting around waiting for things to figure out what to do and get fed data.
Next is yes, the shared floating point unit that gets overwhelmed when both cores are trying to use it.
this shouldn't be an issue but it was because of windows.
Back in the day you could and had to manually patch windows to treat the fx chip like a intel chip with hyperthreading. That fixed windows scheduling two heavy tasks 'next' to each other on the same shared resources chip.
So with that patched, windows would schedule programs like an Intel chip and space out the programs first using every even core (0,2,4,6) then every odd core. That got a lot of general performance back relieving stress on both the decoders and floating point.
Finally, the biggest issue was the cache. the FX's L2+ cache was slower then intel's.
So if the working data set and instructions can't be kept in L0 or L1 ... then the CPU again was waiting for data from L2 or L3 or L4.
L3/4 is the RAM depending on if the cpu had a L3 on die cache as well or not.
Each level is bigger and includes more cores with the highest level cache in the cpu being accessible by all the cores, it's generally how they share data with each other.
So finally with the cpu waiting all the time for data and the most important subsystem it has to wait on being slower than the competitors ... the FX (Bulldozer) was pretty garbage from an engineering standpoint.
For users looking for single threaded performance, intel chips at the time could perform, clock for clock, near 2x the instructions on average.
That is to say a 4ghz amd core performed as well as a intel 2ghz core; in the worst case.
average would be closer to 4ghz amd ~~ 3ghz intel.
And in workloads where the amd's shortcomings didn't gimp it ... it was only 10-20%slower clock for clock while you had twice as many cores for much cheaper.
... surprisingly similar to how it currently is except for a important difference; AMD caught up and Rysen and Threadripper don't have huge crippling flaws. And amd is maybe ~5% behind clock for clock while they have twice the cores for half the price.
@@agenticex ah ic thanks for the info
You obviously haven’t used a laptop with a Celeron processor or a processor that starts with a letter and has 4 digits after it. They still make cheap CPUs that are terrible and still would be terrible with a 8 GHz OC.
i'm still running an AMD FX-6100 overclocked to 4,2Ghz with an aio water cooler and it's stable with games running up to 60 Fps so it's not so bad and it's been running for about 5 years
Clear, concise and helpful. Thanks!
Can you tell me why most of the good looking cases don't have external bay for a optical drive?
Clock speeds are equally as important as IPC, but there needs to be the right balance between both, depending on the process node and the architecture
When you're comparing different processors using the same architecture, yeah. But when comparing an AMD and an Intel CPU? Clock speed is irrelevant.
@@phazonclash both amd and intel are x86-64 and thus both have the same instruction set, and clock speeds always matter, along with IPC, CPI and IC
@@badass6300 here's a video you should definitely watch ua-cam.com/video/bVMkbACYLG8/v-deo.html
Everyone would be attracted by a big buffet tho
Will you please made a video on how hosting server are made and work
Hi Linus. Great video! Thank you!
So for AMD to get a major boost over Intel is it right they need to improve the IPC for each core ?
Eastyy English please
yes, but they are already working and about to launch something with IPC gains, ryzen 3 coming in 2019.
Yes and no. The IPC of Zen+ CPUs (Ryzen 2000 series) is virtually the same as the current Intel CPUs (~3% slower if I remember correctly). Where AMD falls short compared to Intel is pretty much clock speed only (and a bit of latency because of CCX-to-CCX communication). So IPC is not really the issue with Ryzen, as I said, it's almost the same as Intel. The issue is really clock speed and latency. To get better single core performance than Intel, AMD need to EITHER increase clock speed, or increase IPC, or do both. You can be almost 100% sure they'll try to do both and increase both IPC and clock speed. Rumor has it Zen 2 CPUs will have clock speeds of 4.7 Ghz or better and ~15% better IPC compared to Zen. This puts it on par or slightly better in single core performance compared to i9 9900k and MUCH better in multi-core performance. Not to mention it's going to be HELL OF A LOT easier to cool because it's built on the 7nm node, not a mini furnace like 9900k which is basically a factory overclocked CPU.
@@bgtubber you are wrong, it is zen that is 3% slower than zen+ cpu's. zen+ cpu's are still about 7-10% behind current intel CPU's in IPC.
@@shawnpitman876 You are thinking about gaming where you'd be right, Intel is 10-15% faster at same clocks there. I'm talking productivity where the IPC of Ryzen is basically the same as Intel (and even beats it in some tests). Check the article about this over at Techspot: www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/
Cores didn't really matter back in the Kaby Lake days.
they still wouldnt if Intel had their way 4 core 4 thread now because of spectre and well no reason to increase core count. AMD woke the blue giant we will soon see in the next 5 years or so if it has had a real affect or will Intel just continue as they always have.
So how to really compare cpu perf ? Benchmark ? Because i don't feel like my i7 6 core sandybridge at 5ghz need upgrade ...
Hi Linus do you think geek bench scores are replacements to the clock speeds if no why ?
what happened to taran's keyboards?
I mean he has like 3-4 of them but i only saw one.
Syed SherShah that’s a very old clip. That was when they just moved.
Nice video like always 😉
Comparing CPU frequency to cars engine would more be like comparing the maximum rpm the engine can reach. That's an important information but not a performance indicator at all.
A S2000 2.0L 240HP can reach 9500rpm, but il a lot slower than a BMW 435D 3.0L diesel wich devellops 313HP
so, now i need a video to tell me what i *should* consider when selecting a cpu...
I learn so much in this channel
Yes and no
Only yes if it's a huge overclock (like 3.0 to 4.5) and if you are bottlenecking and every single MHz counts
But overall the gains might not be worth extra money/temp/power usage
Dude this is the 3rd tech video I've seen you on in the last 15-20 minutes.
Commenting as a blue tick wanker is good advertisement
Yeah you're right.I overclocked my Core2duo from 2.5 to 4ghz and it pretty much doubled it's cinebench score and made games like doom and GTA V playable.
@@Tomboo Not advertising, I just saw his comments on some 2200g reviews and now here.
for real the most important Ghz number is the one that cancel your bottleneck (cpu), is the only way you can see like... double performance by just modify a single number, even if it is only +200/300 mhz, if with that number the bottleneck disappear.. voalá 30%+ performance appear
For the most part it’s an important stat that’s always taken out of context.
But in the case of the i9, the cheaper and faster model is better for gaming... even though you should really just get an i5 if you’re only gaming.
A full size truck may have 500hp, but that doesn’t mean it’s going to accelerate faster than a 300hp coupe. There’s so much more that goes into it. But just like cars, is it in your budget, and does it do what you need it to best? You can’t haul furniture and tow a 5th wheel in a Bugatti or Model S, and a half ton isn’t going to be setting any records on the Nürburgring.
i5's in big modern games like Star Citizen cant keep up. So it does matter if games program to push or use the entire chip.
Slumpdog like I said, it depends on how it meets your needs. Most mainstream and consoles adapted games run equally fine on a top end i5 for a few hundred less. Me personally, I edit a lot of photos and the occasional video with some games on the side, so I use an i7 for now, and the base model i9 when I need to upgrade.
Actually electric vehicle tows really well
And don't forget that when it comes to bugatti it's still an 8 liter 16 cylinder quad turbo 1500 bhp fucking drain for fuel (easly 1k+ torques)
I'd compare it more to a Japanese cars. Huge power high revving, but almost 0 torque (like motorbikes)
@@GewelReal Well torque isn't really a meaningful spec
@@matthewprather189 yes it is , a honda S2000 has 240hp/1tq and 0-60 in 5.7sec compared to a heavier and lower powered 85 corvette has 230hp/340tq but 0-60 in 5.5sec, torque matters even more on a race track with turn because torque lets you accelerate out of the turn faster and overtake easier because torque is instant
Interesting, but than what do you need to look for in the specs?
Are just look at benchmark results?
a better analogy than cup holders would be compression ratios in car engines
Clockspeed is not at all what makes a chip, look at Bulldozer vs Kaby Lake at 5 GHz.
Bulldozer - the one that got Bulldozered
Detective Pikachu is actually Deadpool 😂
LOOOOOL
*surprised pickachu meme*
Ryan Reynolds yea.
Imagine if there were people who didn't know this lol
I imagined you holding a Motorola razr and i dont believe freshbooks will work on it so well.
So where can we find all this information about the CPU?
Wan show
Wait, theres a core i9 now? I just bought a core i7
Don’t worry, the i9 is just a rebrand of the i7 so they can increase the price.
i9 isn't worth it for most people. Don't worry.
i9 has existed for a bit now. You're probably thinking of 9th gen
If you want an extra room heater get the new i9
@@QuincyIsCrispy i9 has existed in mainstream since 9th gen. So they have existed for the same amount of time....?
Facebook blocked me from liking posts so I came here and liked this video... and I'm into hitting the bell linus 😂
so clock speed is microprocessor rpm?
4:49 "based on its engine displacement alone" would be a better analogy. Cup holders? Seriously Linus?
Yes, but let's make a 6 minute video out of it.
IKR
Techquickie is for people who arent too tech savvy.
Then lets comment on it
he LITERALLY answered the question and said that he would go in detail after the first 40 seconds
absolutely pointless comment, or i'm guessing you didn't even watch it
Welp, we literally had almost everything in this video in q1 of this school year so it's a pretty good video. Explaining everything really clear
Thanks for making, yet another, incredibly informative video!
4:55 My car has ten cup holders! 😆 It wasn't a deciding factor though. I didn't even find out until after I bought the car.
finally tech for a long time ;)
So does more cores matter than CPU speed?
Linus woud say: depends
Software optimisation > raw hardware power
@@mariozenarju6461 Always. Look at consoles. Impressive to see what the PS4 is able to handle.
@@Koeras16 Heard of Android-x86? Now that's a beast
It depends a lot on what you are doing, and again, depends on how much your CPU can do in one cycle. Generally speaking, more cores doesn't affect much the performance, because most things are still being programmed to use only one core, but higher clock not necessarily means that it will be better, to really compare 2 different CPU's you should see benchmarks of each one and decide which one has a better performance in the tasks you are going to use it, there is no easy way to compare CPU.
Well if i overclock it, it does matter xD
Did you watch the video? He mentions overclocking
So uh is ryzen 3 2200 with a Mhz 3500 in my recent built computer inside bios really good?
Can memory keep up? Can you fit your application in the L3 cache? How many ALUs does your pipeline have?
What's the difference between ""Instructions Per Cycle"" and ""Hyper Threading/Simultaneous Multithreading""? Both seem to do the same thing, from the video's explanation...
@@morthim You're kidding, right?
Look, if you can not read, then please do not bother to answer.
@@Lucian_Andries if you did know the answer and wanted to make a sarcastic comment. you should make that clear in the comment. dont go and be negative to someone who made a good answer to your question
@@gamingwithegoon Did you pay attention of what I actually wrote? Or you just wanted to make a "sarcastic comment"?
I know what HT and SMT is and their difference!!!!!!!!!!!!!!!!!
I wrote the question very clear, it's not my fault you can't read. :(
@@gamingwithegoon thanks. a fair portion of the tech community are fairly demeaning like that. i'm not sure why someone would go to an entry video on an educational topic, make a loosely related response, and then respond with hostility when clarification is offered; but it happens.
i rewatched the video to see if hyperthreading was even brought up. it wasn't. idk what the guy was motivated by but he was posting on a video unrelated to one of the topics he brought up. if you understand a thing, and aren't using the videos as teaching material for yourself or others, why question in a non-time-sensitive field?
i deleted the original response. if the person's question was disingenuous, then it may lead others into greater confusion through introducing new topics. if the question was sincere, and my explanation was confusing and inadequate, it similarly bears deleting.
besides, his grammar in his seccond response is unintelligible to me. so it may be the case that my dialect is unintelligble to him and perhaps more generally, those who are struggling with this particular problem.
gaming with egoon, you have my appreciation all the same. keep being awesome and you can be sure good things will come.
@@morthim Does nobody know how to read anymore, nowadays??! :O
I asked what's the difference between IPC and HT(SMT for AMD).
I DID NOT ask what is the difference between HT and SMT!!!!!!!!!!!!!!!!!!!!!!!!!!!
Pay attention or learn to read, don't shit on me for your idiocy! :(
*EDIT:*
Yes, it was a very beautiful explanation, props to you!
But it wasn't what I asked. And if you pay attention, you would see I've written it very good as well. There is nothing confusing about it, unless you don't pay attention, or the other thing... Even a blind man could understand it...
I even added quotation marks, to make the words stand out more... Added some more spaces to eliminate confusion... But you guys are worse than 5 year olds! :((
By the way, you misspelled HT and SMT in that beautiful explanation... :p
Isn’t it the most important part statistic of a processor? Even core count above four cores does not matter too much for gaming, but speed does.
Edit: Guess I was wrong. Interesting video!
Instructions per cylce (IPC) is what really matters, but it is rather complicated to calculate it precisely, and it doesn't take into account what kind of program you are running, so a cpu with good IPC may be really fast at calculating mathematical formulas and stuff, but it may be really awful with video or other kind of stuff. So that is why we have benchmarks today, benchmarks allow you to have an estimative of how your cpu behaves in the context you want to use it. The thing is that benchmarks are also not perfect, because a cpu manufacturer, like AMD or Intel, could just create a CPU really good at running a handful of the most famous benchmarks and suck at other things ( although I don't think this happens today), so thats why people like Linus, have test everything in different setups and different benchmarks before posting a review.
Edit: actually is IPC not CPI, thanks for correcting me.
It's part of the equation pretty much.
Clock speed * IPC * stalls * number of cores
Times the CPU is waiting for data to come from the memory, or picked the wrong branch on a conditional branch, or ran a weirdass instruction that locks the CPU for a few cycles, or was unable to fill all the pipelines.
It was pretty much the thing that made the pentium 4 CPU be slower than the atlhon.
It had a BIG pipeline, and on every branch prediction fail, it had to wait 40 cycles, while an atlhon had to wait only 20.
That's indeed the best idea.
Otherwise you will be trying to simulate what the program need in your head and that's quite messy.
4:05 OMG my old g11
Yes but ony if you also take into considaration intructions per clock and core count
Ofcourse it matter. What happened to you?
nothing u idiot hes just talking about it, calm down.
sound one ain't watch the video lol
back in 2013 you could buy a AMD FX9590 that ran at 5.0Ghz stock but a ryzen 1500x at 3.4Ghz is faster because it has 52% better IPC
@@eaglefat9398 Also an fx8320 can be overclocked to 9590 speeds (4.7Ghz all core) for £100 while a 1500x was £140
Not to mention you'll get more done on an fx8320 in 2012 than you will by waiting for another 5 years and buying a 1500x.
Things have to be taken into perspective.
Clock speeds dont matter
IPC doesnt matter
Cores dont matter
They only matter when configured together in... IPC*clocks*cores/scaling/(amdahls law)
I'm pretty sure Raghavendra Singh was just joking....
Yo no views but there are 3 comments hmmmmmmm...
I think you will be interested in this video.
ua-cam.com/video/RY_2gElt3SA/v-deo.html
Hey teckquikie is it possible for u to do a video abt computer components since we r doing then at school and would like to get it. Appreciate it.
Soooo what do I look for beside cache, cores, clock speed?
Thanks very much!
So what do I look for if not GHz for gaming?
Re: the 5:55 mark, funny you mention that since "LightScribe" (CD Burning explanation) is recommended to me!
4:08 - an original Logitech G15? Damn I thought I was the only one still using one of those!
What is the best netbook that is low end but has really solid keyboard and good typing while being good enough for youtube video watching and running emulators.
Please i just subscribed to your channel please help me which should I choose for gaming between hp 15 core i5 8250u 1.60-1.80ghz 8gbram or HP pavilion 15 core i5 7200u 12gb ram 2.5ghz
But Linus, my car has 8 cup holders! Do you know how much change I can stick in them?!
So may I get a translation here. So given the information that CPU vendors give, besides base and boost clock what other key things should I be looking for in the specs sheet for any given processor? Thank's in advance. :-)