I am still puzzled why neither Intel or AMD tries to make a speed record in simple thread computing

I am talking about 10 000 MHz CPU, it doesnt exist but it should

Today we get things like 3800 MHz quadcore or even more cores. I get it why, there is a lot of multithreading today.

But there are also problems that cannot be properly multithreaded.

And there isnt even one CPU that runs single thread at incredible speed.

We need a dualcore CPU built with highest possible single thread speed in mind.
Why not quadcore? Tempreature issues.
Dualcore will not have temperature issues as such. Also omit the GPU part from the CPU die.
Dual processing CPU core, no GPU, at 10 000 Mhz, thats all.

It would be a bit more than double the speed of a 3800 MHz CPU (8 core CPU) when running its single thread.

GENETIC RESEARCH software dont work well as multithreaded apps, the single threaded version is the only one reliable. This has been so since the year 2007 and things arent changing.

For a researcher, a lot of research would cut research times with 10 000 MHz dualcore CPU paired with 64 GB of DDR 5 RAM or whatnot.

anyway a less silly question: now im wondering at what point "electrons cannot move any faster" becomes the issue. in current context in mhz.

>>105719698 (OP)
Educate yourself
https://en.wikipedia.org/wiki/NetBurst

>>105719698 (OP)
Heat, retard
You want more speed you need more current. You burn more current you have more heat. Heat removal is the limiting factor on IC speed. Even with the best liquid cooling system known to man you are still limited by how quickly heat can diffuse our od the silicon
Electrons speeds are not an issue, at least not yet. Electricity is more about moving electric fields rather than moving actual electrons around

>>105719815
>netburst

you are trying to derail the thread
netburst is not capable of going past 3800 MHz
it was flawed tech

however with Ryzen or Intel core i7 Gen.3. or later, there is no such obstacle as with netburst

>>105719836
>heat

there is heat in modern CPUs because they have 4-8 processing cores

dualcore would be considerably less heat

>>105719868
The scaling is not linear, having 4 cores at 2GHz requires a lot less current and produces a lot less heat than one core at 8GHz. Also the cores are physically separated from one another in the chip, so you have 4 temperature centers to dissipate heat from rather than one ultra-high-temperature point

>>105719857

>>105719914
LGA775 CPUs are considerably slower than Nehalem tech

>>105719969
>year 2000 cpu is slower than 2008 cpu
source?

>>105719698 (OP)
>I am still puzzled why neither Intel or AMD tries to make a speed record in simple thread computing
heat
as simple as

Hasn't there been some really huge overclocks reaching near those speeds but the problem is that you need shit like liquid nitrogen to keep the cpu from melting? I saw some pretty massive overclocks on old amd cpus from Bulldozer era.

>>105719698 (OP)
because multi tasking is more important than countless of operations in a queue.

>>105720246
yes but there could be dedicated computers for niche use to speed up the speed of research, however there arent because computer manufacturers do not make them

>>105720318
Sure. Just buy a quad core arrow lake and OC it

>>105720330
they cant be overclocked

>>105720342

>>105719857
>netburst is not capable of going past 3800 MHz
>>105719698 (OP)
>Today we get things like 3800 MHz quadcore
Keep going lil buddy you almost got it

>>105719698 (OP)
has to do with design complexity and chip yields,

large cores are expensive to design and very complicated, and defective single core chips cant be salvaged like multicore chips (aka if you have 8 cores and 2 are bad, you can just disable those and still sell the chip,)

last big core intel designed was the cancelled tejas and jayhawk chip, which was an enormous single core chipdesigned to clock to 5-7ghz, it cost intel alot of money and was a huge loss. ridiculously powerhungry and would have been slower than the core 2 duo designs intel ended up shipping instead.

>>105719857
what are you smoking? netburst is exactly what the op is asking about , its a chip designed for high clock speeds and high IPC (inferior to later chips but Netburst was not as low ipc as people say, for its time), they do infact run faster than 3.8ghz ive litterally posted a 65nm cedar mill at 5.3ghz without suicide voltage on air, the problem is higher clocks = more power consumption and the laws of physics simply do not allow ultra high frequencies currently on silicon technology without insane powerdraw. so chasing 10ghz is not feasible, if you do , you end up with netburst. intel was not the only company to try this either, ibm did exactly the same thing with power 5 and power 6, reaching 5ghz in 2007 but had horrible efficiency and is why apple dropped them for intel cpus back in the day.

the aforementioned cedarmill draws 65w at 3.2ghz and like 250w at 4.6ghz, at 5.3 it trips the ocp on my motherboard, but its probably in the >300w range and is nearly uncoolable without refrigeration. even with all of the advancements if intel were to make a giant core designed to clock to insane speeds, it would draw >500w guaranteed if it were hitting anywhere near 10ghz, would unironcially be a house fire.

>>105719698 (OP)
We have 6 Ghz single threaded chips.

>>105721302
>65nm cedar mill at 5.3ghz without suicide voltage on air, the problem is higher clocks = more power consumption and the laws of physics simply do not allow ultra high frequencies

that stuff is useless, you can run dualcore sandy bridge at 5000 MHz and I guarantee its would be equal to 8000 MHz cedar mill if cedar mill was able to do such speeds (which it aint)

>>105721313
how about 10 ghz

>>105719836
>Electricity is more about moving electric fields
Will you ever have to deal with magnetism effects?

have you heard of a concept called heat density?
150W 9700X runs hotter than a 250W 9950X because the heat is concentrated into a smaller area
and what you're describing is already being done by modern CPUs
when all other cores are idle or parked, one core gets pumped with 1.5V so it can clock at 5.9-6.2GHz
you cannot get to 10GHz on current CPUs even with LN2 and all other cores disabled
to make them clock higher you need extra logic on the chips and it's not always worth it, one example is AMD Vega which had a lot of silicon dedicated to allowing it to clock higher than Fiji which ended up backfiring because core clock wasn't the bottleneck in that uarch

complete fucking retard OP, never post again

>>105719698 (OP)
Because increasing performance by 2x by having 2 threads is infinitely cheaper than increasing performance by 1.5x while having a single thread.

>>105721375
Packaging size constraint limits the heat dissipation

>>105721412
Actually now that I think about it, its likely not that. I mean it is for consumer devices since we get a usable cpu package so we have to work from that level of limits to begin with. But from chip designers perspective, you could probably engineer a 10 Ghz single core chip but that would make it an unsuable chip for sale due to power budget being eaten up by single chip which has no real world use case anymore.

>>105719914
Just imagine the amount of wasted cycles waiting on I/O....

The reason is complicated but the circuitry becomes unstable past a certain frequency.

>>105721379
We do, self-induction is always a problem in high frequency, high current applications, especially at the supply bond wires where you have a large amount of current flowing through a single thick wire

>>105719698 (OP)
Intel was going to under Gelsinger and Keller.
Royal Core design was their third attempt at 'reverse hyper threading' instead of individual cores, with multiple execution ports off a usr, you'd just have a huge massive ~100mm2 core with hundreds of execution ports, some big, some small.
'cores' would be virtual/amorphous - not really smt or cmt, more like partitions.
You could have theoretically run the whole run as a huge single core.

Most of it died with Keller being pushed out by Raja.
It fully died with Gelsinger's firing.

>>105719698 (OP)
Fast cores are worthless if you can't keep them fed and cooled. Clock speed isn't everything. What really matters is having a balanced design to maximize instructions per clock cycle, and that's almost entirely determined by the micro architecture and how well it can manage the cache hierarchy while doing dynamic frequency scaling. Highly parallel designs with a good software scheduler end up being superior because when one core gets too hot and needs to downclock, your workload can shift to a different core and maintain higher speeds.

>>105722662
but hey this only means computers arent so great for biology work

>>105721929
what do you mean

Why worry about clock speeds ? Just make an asyncronous CPU instead

The suits ofc fear change like the plague

>>105721366
your seem to have missed the point, power scaling applies to all cpus, if you run a sandy bridge at 5 ghz+ it will consume huge amounts of power, secondly cedarmill on 32nm soi would clock much higher than sandybridge, and is an ancient design, but its design principles follow exactly what the op describes, if intel were to build up a netburst like core with 4x as many transistors like sandybridge and much higher ipc as a result, it would have absolutely monsterous power draw.

>>105719698 (OP)
Alignment algos being a pile of fucking garbage will not be resolved by whatever impractical CPU design and associated custom socket/mobo etc would have to be used for 10gHz
if even snapgene is a slow pile of shit you know there's no hope. pray that IDT gets into the open source software dev market (lmao) or write your own improvements. geneticists and biofags in general are notorious for being nocoders. you could make a better program for alignment and literally get rich, I can personally attest that industry would pay the toll for a better software solution. pretty easy to interop between software too since the files and ref# etc are all standardized