World Record: Haswell CPU Overclocked to 7193.8 MHz

[N.Broekhuijsen]

Chi-Kui Lam from Hong Kong has set a new overclocking world record.

World Record: Haswell CPU Overclocked to 7193.8 MHz : Read more

 

[derekullo]

If you disable hyperthreading you have an i5.
If you disable half the cores and hyperthreading you have a highly advanced Pentium D.

 

[LostAlone]

So... This isn't actually a world record, it's just a record for a Haswell? Because those are some seriously different things, and I'm certain that a bunch of AMD chips have been taken past 8GHz, and that was a couple of years ago.

It doesn't stop this being cool and all, just that if you're going to call something a world record, it really needs to be the world record for all comparable things. It's like saying that a station wagon set a world speed record, then finding out that it was only for station wagons.

 

[g00ey]

If you disable half the cores and hyperthreading you have a highly advanced Pentium D.[/quote]

Are you sure? As I understand it, the Pentium D was a true dual core Pentium 4 with hyper threading disabled and the P4 was architecturally quite different from the prior Pentium 3. IIRC Intel scrapped the P4 and went back to the design philosophy of the Pentium 3 when they designed the Core Duo CPUs and their successors.

 

[TechyInAZ]

If you disable hyperthreading you have an i5.
If you disable half the cores and hyperthreading you have a highly advanced Pentium D.

No not a Pentium D, a Pentium G3258 20th Anniversary Edition.

 

[InvalidError]

If you disable hyperthreading you have an i5.
If you disable half the cores and hyperthreading you have a highly advanced Pentium D.

No not a Pentium D, a Pentium G3258 20th Anniversary Edition.

Yup, that was exactly the first thing I wanted to post when I read the "news."

No point in getting an i7 to overclock if you are going to downgrade it to Pentium-level feature-wise to get there.

 

[StarBound]

I know the point of suicide overclocking is to push a cpu or gpu to the absolute max but when are we going to get to the point where we say the standard is to have the system then run a game or benchmarking tool for an hour and stable before we accept the overclock numbers?

 

[pills161]

Doesn't mean anything really if you have to gimp the chip to get there. I'm with Starbound, being able to boot into windows and open CPU-Z doesn't mean shit to me, let's see you overclock a chip to 7193.8 without partially disabling it and then run some games/benchies and then I'll be impressed.

 

[jasonelmore]

If you disable hyperthreading you have an i5.
If you disable half the cores and hyperthreading you have a highly advanced Pentium D.

not really true. Cache is still bigger per core, still has more instruction sets (dedicated hardware).

Cache differentiates Core cpu's more than anything besides Hyper Threading.

 

[mrmotion]

"Of course, there is no question that this is a completely useless exercise for everyday purposes, but it remains fun to see what certain hardware combinations are capable of."

Looks like you guys quit reading before the end of the article...

 

[wireframed]

So really, this is about as fast as a 3GHz i7? Doesn't seem worth it, losing 60% of the performance.
Of course if you only care about singlethreaded apps, but how many of those are really that demanding? Everything I'm waiting on is multithreaded (and almost perfectly parallel).

'Course he didn't do it for the performance, I get that. It just seems odd since he's only focusing on one small part of what performance is. Really nails how pervasive the "frequency = performance" is.

 

[heero yuy]

world record? for intel cpus? because i'm pretty sure someone has had an AMD CPU up to/over 8GHz before

 

[wireframed]

world record? for intel cpus? because i'm pretty sure someone has had an AMD CPU up to/over 8GHz before

That's another problem with focusing on MHz exclusively - a 5GHz AMD CPU is in many ways equivalent to a 3-4GHz Intel CPU (depending on what strengths the application plays to).

Really, this could be considered a "14.4GHz CPU" (7.2GHz on two cores), which makes it a lot less impressive if you take a 3.9GHz i7 which would be a 15.6GHz CPU. Again, I'm a bit biased, because my most demanding applications scale almost linearly with the number of cores (and CPUs and workstations, for that matter).
So yeah, two fast cores, that a stock i7 from a couple years ago could beat in many applications...

 

[InvalidError]

world record? for intel cpus? because i'm pretty sure someone has had an AMD CPU up to/over 8GHz before

Either way, wake me up when overclocking world-records translate into world-records in either benchmarks or real-world applications instead of merely large numbers in CPU-Z that cannot be used for any actual work.

 

[Saso Mange]

I see many of you talk about other CPU's which got over 8GHz mark before... I ask you, did OS boot completely with such OC ?

 

[oxiide]

I know the point of suicide overclocking is to push a cpu or gpu to the absolute max but when are we going to get to the point where we say the standard is to have the system then run a game or benchmarking tool for an hour and stable before we accept the overclock numbers?

There already is a standard. It has to boot into the OS and be able to screenshot a CPU-Z window.

Competitive overclocking is about the competition itself, it has nothing to do with real-world benchmark performance. Take it or leave it as it is. Yes, it is abstracted from the reality of how we use our processors, but then Usain Bolt's sprinting doesn't reflect how most of us use our legs, either.

 

[InvalidError]

Yes, it is abstracted from the reality of how we use our processors, but then Usain Bolt's sprinting doesn't reflect how most of us use our legs, either.

But Bolt's legs do show a practical application of what the human body is capable of with extensive training and genetic luck-of-the-draw.

"Suicide overclocking" on the other hand is more like bodybuilders who cheat using steroids and end up ripping their muscles because they lacked sufficient conditioning from regular training to keep up with the strain... big muscles that cannot do any more useful work than people with half the muscular mass.

 

[rishiswaz]

Guys you do realize that dividing a frequency by the amount of cores doesn't mean anything right? 7.2GHz dual core does not mean its a 14.4GHz CPU, there are thing like parallelism and multithreading which don't really allow the CPU to scale linearly like that. The CPU will still balance threads even if there are only fewer than normal, it just means each individual thread will have a larger load.

 

[rishiswaz]

Guys you do realize that dividing a frequency by the amount of cores doesn't mean anything right? 7.2GHz dual core does not mean its a 14.4GHz CPU, there are thing like parallelism and multithreading which don't really allow the CPU to scale linearly like that. The CPU will still balance threads even if there are only fewer than normal, it just means each individual thread will have a larger load.

 

[Dustin Mock]

Anyone know what the true cpu clock speed world record really is? Not these types of competitions world record, but something along the lines of an "MIT custom built CPU operating at .5 degrees above absolute zero" kind of world record.

 

[InvalidError]

Guys you do realize that dividing a frequency by the amount of cores doesn't mean anything right? 7.2GHz dual core does not mean its a 14.4GHz CPU

That is not the point.

The point is that those overclock records are doing nothing more than managing to boot into Windows without crashing. If you ran Prime95 or even real-world applications or game on them, they would likely crash before getting in-game proper... they are not proper overclocks; they are just-barely-working overclocks. They are not usable for anything beyond claiming a world record.

 

[InvalidError]

Not these types of competitions world record, but something along the lines of an "MIT custom built CPU operating at .5 degrees above absolute zero" kind of world record.

If someone rebuilt the 8086 on 14nm, they would likely be able to run it at over 15GHz... but I doubt anyone would want to use it since it is in-order, takes 3-4 cycles just to load instructions from memory and another 2-5 cycles to actually execute them so it would perform about on par with a modern chip running at 300MHz without any of the newer chips' capabilities.

If you want to run circuits really fast, you have to make individual pipeline stages really simple. Intel tried that with the P4 and failed.

 

[Dustin Mock]

Anyone know what the true cpu clock speed world record really is? Not these types of competitions world record, but something along the lines of an "MIT custom built CPU operating at .5 degrees above absolute zero" kind of world record.

Nevermind, found what I was looking for. It's 798 GHz on a Silicone-Germanium 130 Nanometer chip running on 1.7 volts at 4 degrees Kelvin using liquid Helium.

http://www.news.gatech.edu/2014/02/17/silicon-germanium-chip-sets-new-speed-record

 

[Bondfc11]

I agree with all the comments about this being a "bogus" record. This is an absolutely useless exercise. I understand there initially is a wow factor involved, but if you cannot run anything stable before the thing implodes how is this worth anyone's time? The real world record should be as many people have said - OC that sucker and run some Prime, etc. Then I will be impressed. This - not so much (really not impressed at all).

 

[oxiide]

They are not usable for anything beyond claiming a world record.

But no one made such a claim. Anywhere. It is literally just about claiming a world record, and I don't think this poor, unsuspecting news article has tried to advertise it as anything more. So I don't understand what everyone is arguing so hard against. You either care about this subject matter, or you don't.