Intel Ivy Bridge Processor Hits 7GHz Overclock Record

MojoKid writes "Renowned Overclocker HiCookie used a Gigabyte Z77X-UD3H motherboard to achieve a fully validated 7.03GHz clock speed on an Intel Core i7 3770K Ivy Bridge processor. As it stands, that's the highest clockspeed for an Ivy Bridge CPU, and it required a steady dose of liquid nitrogen to get there. HiCookie also broke a record for the highest memory speed on an Ivy Bridge platform, pushing his G.Skill Trident X DDR3-2800 memory kit populated in four DIMM slots to 3,280MHz. Not for the faint of heart, the record breaking CPU overclock required that HiCookie pump 1.956V to the processor, according to his CPU-Z screenshot. The CPU multiplier was set at x63."

Comments

Cool

[Lord Lode]

So Ivy Bridge can do something after all! :D

Re:Cool

[Anonymous Coward]

ooh burrn on ivy bridge!

Re:Cool

[Anonymous Coward]

It still doesn't do anything. It just does nothing that much quicker.

Re:Cool

[smash]

You mean other than kicking the crap out of AMD?

Re:Cool

[Anonymous Coward]

http://hothardware.com/News/AMD-Breaks-Frequency-Record-with-Upcoming-FX-Processor/

Intel makes 7Ghz. Yawn.

Come back with your snappy come-back when you hit 8.4Ghz.

Re:Cool

[smash]

Because they both have the same IPC?

Re:Cool

[beelsebob]

Come back when the AMD CPU hits 14Ghz, because as things stand, it'll need to get there to beat the Intel at 7Ghz

Re:Cool

[solidraven]

True story!
As it is right now with the huge pipelines the indicated clock frequency doesn't matter all that much. Especially as the clock is often divided into sub cycles. The real issue still is - as it always has been - talking to the memory. And while ever smarter compilers combined with better out of order execution does help. It's still a hassle that you can't directly talk to the memory and need to wait. It really drags down the efficiency of your pipeline if you made a wrong prediction. The actual value we should be looking at is the amount of instruction launched per second. If possible half floating point, half integer. That should lead to far more interesting performance evaluations than the just the clock frequencies. Might lead to funny results and a lot of sad gamers.

Re:Cool

[Anonymous Coward]

Excuse me how do you figure 7ghz on Intel equals 14ghz on an AMD. If you are thinking of Hyper threading then you do not understand how a CPU works at all and you need to go back to the books and read up on how CPU's work and what is really the difference between AMD and Intel chips.

Re:Cool

[beelsebob]

I figure it because that's what the benchmarks show –AMD's CPU cores are currently roughly half the speed of Intel's at the same clock rate.

One core, two threads?

[ZeroSumHappiness]

Can someone explain why it's reporting one core, two threads?

Is this:
1. Set to one core to get a better heat profile?
2. Only using one core for the test?
3. Using all cores for the test but only reporting one core's results?

Because if it's 1 or 2 I think I see some problems with this benchmark.

Re:One core, two threads?

[Anonymous Coward]

Generally they disable all cores but one to achieve these clock speeds.

Re:One core, two threads?

[Anonymous Coward]

Look at any record setting benchmark, Intel or AMD. Cores are disabled on purpose for the OC to be achievable.

Re:One core, two threads?

[h4rr4r]

Then it is not really an i7, might as well over clock an i3.

Re:One core, two threads?

[Anonymous Coward]

So... what's the advantage to having one core running at 7GHz versus four at standard clock speeds, assuming that whatever you're running takes full advantage of all the threads?

Re:One core, two threads?

[Anonymous Coward]

.../facepalm

Re:One core, two threads?

[Anonymous Coward]

I don't think you really understand the purpose of this. He's not "running" anything, so it doesn't matter how many threads. An overclock like this is only done to say that it can be done and that you did it, it's not practical in any way shape or form. You would never try to run an actual application on this, odds are you couldn't maintain system stability for more than a few minutes, and even if you did the cooling and power requirements are well beyond reasonable.

Re:One core, two threads?

[Jeng]

This is not for practical use, it's just to one-up the last guy.

Re:One core, two threads?

[Anonymous Coward]

When you absolutely, positively have to have your calculation done on time...invest in lots of liquid cooling to do something for no purpose!

Re:One core, two threads?

[h4rr4r]

You know what I meant. Just as you would not call an i7 binned part that was crippled to be sold as an i3 this should not be called one either.

Re:One core, two threads?

[K. S. Kyosuke]

So... what's the advantage to having one core running at 7GHz versus four at standard clock speeds, assuming that whatever you're running takes full advantage of all the threads?

Probably this tiny little inner satisfaction of having one's feeling of insufficient manliness adequately compensated by non-anatomical means.

Re:One core, two threads?

[Anonymous Coward]

Actually, and i3 has 2 cores with 4 threads.
Only the very lowest end celerons and atoms have a single active core.

Re:One core, two threads?

[tripleevenfall]

That's what sports cars are for.

Re:One core, two threads?

[Anonymous Coward]

The biggest feature of the i7 that the Core 2 line didn't have is the ability to clock one core to crazy-fast speeds when the rest are disabled.

Re:One core, two threads?

[ZeroSumHappiness]

Thank you for confirming my suspicion. Obviously I have more interest in using rather than record breaking so I'd have liked to also see what it can do on all four cores under liquid cooling.

Re:One core, two threads?

[Luckyo]

... do it on a dedicated server farm. Cheaper and more reliable.

Yes, I got the joke.

Re:One core, two threads?

[Anonymous Coward]

Or maybe high-frequency trading

Re:One core, two threads?

[Jeng]

No, these with these types of overclocks you can normally only run the system long enough to run CPU-Z and get a screenshot. The majority of windows services are not running, only one CPU core is running, and you are cooling it with liquid nitrogen that tends to boil away rather quickly.

Re:One core, two threads?

[Anonymous Coward]

No. It's Two Threads One Core.

Re:One core, two threads?

[Anonymous Coward]

It's kind of stretch to call that "the biggest feature" when it only appeals to about 0.0001% of their customer base. I would say the onboard memory controller, improved IO and huge cache increase were probably a bit more relevant to the vast majority of consumers.

Re:One core, two threads?

[BlackSnake112]

And hot women

Re:One core, two threads?

[Anonymous Coward]

It's the same performance as a 1.6Ghz quadcore with just air cooling, what you're seeing is how fast one core can go, given that all 4 running would quadruple the thermal requirements.

Generally the purpose of doing this is for shits and giggles, you would not be able to actually use the computer configured that way since the cooling solution would run out pretty quickly. The most you can do for a functional system is to immerse in mineral oil, with SSD drives and not submerge the PSU.This is generally a non-viable solution for most people. Even water cooling is non-viable for most people. This is why you see heatpipes in aftermarket coolers, this is about as complicated as you can get in a system that is used by someone competent. Everyone else prefers quiet over speed/complexity.

Re:One core, two threads?

[Anonymous Coward]

*your non-paralellalizalable calculations

Re:One core, two threads?

[geekoid]

Yes, keep telling yourself you can't get hot women because you are too confident and have a huge cock.

Re:One core, two threads?

[Penguinisto]

So... what's the advantage to having one core running at 7GHz versus four at standard clock speeds, assuming that whatever you're running takes full advantage of all the threads?

To play Quake at ~9x its rated speed, of course.

Why else would you want to do it?

Re:One core, two threads?

[LtGordon]

It's the same performance as a 1.6Ghz quadcore with just air cooling

Except that it's not. For some theoretical computations that could be made perfectly parallel, this might be nearly true. However, in most cases (presently), the limiting factor in computation speed is the clock speed of an individual core.

Re:One core, two threads?

[s.petry]

Oh come on now, he is not doing any calculations. He's not even going to be able to play Warcraft faster than the guy living in the basement next door!

Useless epeen waving..

Re:One core, two threads?

[fnj]

Actually, and i3 has 2 cores with 4 threads.

Actually, these days even an Atom has 2 cores, 4 threads, and 64 bit code capability now. And all in 10 watts. Blew my mind.

Re:One core, two threads?

[smash]

Nothing. but it does prove that with adequate cooling you can get to 7ghz on current generation silicon.

Re:One core, two threads?

[smash]

Lol, if you are running anything financial related on a machine overclocked, you're an idiot, let alone oc'd to 7ghz.

Re:One core, two threads?

[jedidiah]

The main culprit there is Flash and a normal clocked single core is more than adequate for that.

Re:One core, two threads?

[Anonymous Coward]

Maybe you do things to compensate for lack of dick, but don't push your own inadequacies upon others.

Re:One core, two threads?

[Anonymous Coward]

HFT houses run overclocked CPUs, certainly. They hire kernel engineers to shave off every nanosecond, they have infiniband vendors give them custom firmware and access to parameters nobody else has, they get the CPUs before anybody else does, and they certainly overclock them.

They measure microseconds in $millions/hour.

Re:One core, two threads?

[rubycodez]

said the small-dicked AC

Re:One core, two threads?

[Jeff DeMaagd]

I thought it was a joke.

Re:One core, two threads?

[Jeff DeMaagd]

Did anyone run the small penis jokes on the explorers that climbed mountains? That's something else that had little practical value, but I don't think climbers were heaped with ridicule for doing useless things. I'm not saying that overclocking is comparable, but it's less ridiculous as the risk of dying is lower.

Re:One core, two threads?

[Anonymous Coward]

Better than a huge clock (as in clock rate), I guess.

Re:One core, two threads?

[Anonymous Coward]

If he drags his setup up to the summit of Mound Everest, we promise not to make those jokes.

Unless it's too hard to resist.

Re:One core, two threads?

[thegarbz]

I thought it was a joke.

The GP's comment or the financial trading system?

Re:One core, two threads?

[Anonymous Coward]

Exactly, re-read TFA with this in mind. "not for the faint of heart" ? wtf ?

Re:One core, two threads?

[RivenAleem]

And to run Crysis.

Re:One core, two threads?

[Anonymous Coward]

Except that isn't the full story, its really about single core instructions per second, or IPS. With two processors of the same architecture and pipeline, you'll get more IPS at higher clock speeds, yeah, totally. But when you compare Intel to AMD, for instance, its totally apples to oranges. AMD used to get way more bang for your "buck", getting equivalent IPS to higher clocked Intel processors. Now i'm not sure, if anything it might be the other way around.

Re:One core, two threads?

[mhajicek]

Main culprit here is CADCAM software. Some functions can be multithreaded, others can't. I spend hours per day waiting for single threaded operations.

Re:One core, two threads?

[Anonymous Coward]

Hot* women aren't objects of compensation. They're human beings. Be a bit mature.

*What's "hot" is also subjective.

Only 1 core, 2 threads, clocked at 7.03 GHz

[Anonymous Coward]

The results are pretty impressive, but if you look at the screen capture in the article, only 1 core was enabled.

Re:Only 1 core, 2 threads, clocked at 7.03 GHz

[Genda]

This is essentially the only way to run this experiment, if you run all the cores at this speed, fusion is initiated, a black hole forms and time runs backwards!

Re:Only 1 core, 2 threads, clocked at 7.03 GHz

[Jeng]

Woah! So if time runs backwards, but you still measure it as going forward, does the cpu end up running at infinite hertz?

Re:Only 1 core, 2 threads, clocked at 7.03 GHz

[FrootLoops]

Woah! So if time runs backwards, but you still measure it as going forward, does the cpu end up running at infinite hertz?

You made my brain hertz.

Re:Only 1 core, 2 threads, clocked at 7.03 GHz

[geekoid]

It would run infinitely fast at 0 hertz

Re:Only 1 core, 2 threads, clocked at 7.03 GHz

[Jesus_666]

This is essentially the only way to run this experiment, if you run all the cores at this speed, fusion is initiated [...]

So if you overclock all of the Intel's cores at once it turns into an AMD?

Re:Only 1 core, 2 threads, clocked at 7.03 GHz

[AngryDeuce]

The results are pretty impressive

I honestly don't understand why. These ridiculous liquid nitrogen overclocks have absolutely no real world implications whatsoever. They completely trash the hardware, and for what? A big number? What the hell good is that?

It's a shame, because the people that should be getting the hype and recognition are the ones that are overclocking their systems while still having a modicum of stability with real-world applications and reasonable up-time, because at least that's useful to enthusiasts and pushes a real envelope as opposed to a bullshit fake one that only a very, very select few can duplicate and even fewer would even bother.

Want to impress me? Crank out stable 5+ GHz on air cooling across all the cores in an always-on machine. Playing games with liquid Nitrogen is not impressive at all. These guys are the ricers of the computer world.

Re:Only 1 core, 2 threads, clocked at 7.03 GHz

[asliarun]

The results are pretty impressive

I honestly don't understand why. These ridiculous liquid nitrogen overclocks have absolutely no real world implications whatsoever. They completely trash the hardware, and for what? A big number? What the hell good is that?

It's a shame, because the people that should be getting the hype and recognition are the ones that are overclocking their systems while still having a modicum of stability with real-world applications and reasonable up-time, because at least that's useful to enthusiasts and pushes a real envelope as opposed to a bullshit fake one that only a very, very select few can duplicate and even fewer would even bother.

Want to impress me? Crank out stable 5+ GHz on air cooling across all the cores in an always-on machine. Playing games with liquid Nitrogen is not impressive at all. These guys are the ricers of the computer world.

Actually, you are wrong. I'm not speaking for overclockers and in fact, I'm not even one. However, extreme overclocking is very valuable. It tells normal overclockers how much headroom they can expect (at least relative to another chip), it gives an indication of how robust the chip design and the process technology is.

Your car analogy is completely wrong as well. A ricer analogy would be someone who uses a fancy case but does nothing to improve the internals. The analogy would be someone who takes a stock engine and tries to rev it to the maximum possible rpm by using any means. I imagine that many people would find this a valuable metric especially when they are comparing various engines, especially for specialized needs such as drag racing.

Re:Only 1 core, 2 threads, clocked at 7.03 GHz

[beelsebob]

No it doesn't, because it tells you nothing about the shape of the curve. For example, Sandy Bridge never hit huge liquid nitrogen over clocks, but it did overclock extremely well for enthusiasts (it can generally hit >4.5Ghz on air, 5Ghz on water). Meanwhile Bulldozer hit a record 8.4Ghz, but in practice, it doesn't overclock that well (4-4.5 is about the most you can get out of it even on water).

Re:Only 1 core, 2 threads, clocked at 7.03 GHz

[AngryDeuce]

It tells normal overclockers how much headroom they can expect

No it doesn't, because it's being demonstrated in completely unrealistic conditions with no real world component at all. Your car analogy failed as well the instant you said "stock", because this chip wasn't stock, nor was the means to o/c it. They disabled all but one core. Who is going to do that? Nobody but an idiot overclocker looking to get a big number. They used liquid nitrogen and completely trashed all of the hardware that went into this o/c. Who is going to do that? Nobody but an idiot overclocker with either deep pockets or a corporate sponsor.

What the hell good is pumping jet fuel through an engine on a stand to see how fast you can rev it until it explodes? How useful is that measurement? How many people are going to push their engines that far? How many people are going to look at that measurement and try and apply it to a real world situation? Nobody, just other idiots doing the same meaningless things, too. Oh, he got 20k, I'll get 21k, that's moving the technology forward! Except not, 99.9% of people aren't buying engines to blow them the fuck up.

I bet even drag racers aren't pushing their engines to the point where they are completely trashed at the end of every single race. Maybe the Tony Stark billionaire playboys (there's so many of them, am I right?) are doing that because they don't give a shit about exploding an engine at the end of every quarter-mile, but most drag racers are trying to find the fastest speed they can coax out of an engine without totally trashing it. Hell, how many drivers would even want to be behind the wheel of a car that was probably going to explode at the end of the race?

That's why these o/c events need to be tied to the real world to impress me. No disabling cores, no fucking liquid nitrogen. You o/c with off the shelf components, you put the machine under load, and you show me your o/c running, under the same load, 24 hours from now. That is an impressive o/c. This is nothing more than wanking off for bragging rights.

Re:Only 1 core, 2 threads, clocked at 7.03 GHz

[Nemba]

Well, yes, that would be impressive. Because if it could run and stay stable at that speed with off the shelf components, surely it would just be sold at that speed for a nice $50 premium or whatever? Not that I'm saying I think it's impressive or even worth their time. But I think it's probably the only kind of overclocking which is really possible (except for really small amounts, and then still a bit of loss in stability over time). I'm no insider in the industry but I can only assume these chips are tested pretty extensively...

Re:Only 1 core, 2 threads, clocked at 7.03 GHz

[tzot]

OIC. Deep Thought was overclocked, then. That's why we know the answer, but we still haven't run the computation.

Is that a joke?

[Anonymous Coward]

Couldn't POWER CPUs do >5 GHz as their normal speed already a loong time ago? (Apart from being a much better architecture to begin with.)
And didn't many people do 7GHz overclockings, using liquid nitrogen, over five years ago?
How meaningless is a overclocking speed? It's like saying: Your Smart will go 400km/h... if only we run it as ten bazillion RPM. It's still a Smart!! And you will never get this in real life!
This is damn close to fraud, to spread such bullshit so people get a false feeling of it being so fast.

Re:Is that a joke?

[smelch]

The point is that the CPU and the motherboard are all manufactured in such a way that the processor can work at a clock cycle so damn fast. That is precision. Not everything is about work/time. To your analogy, it's more like saying "We can have the engine do ten bazillion RPM! That's a quality engine! However, if you try to drive at that RPM, any flaw in the system can cause a huge amount of damage, so don't expect to drive it at that speed."

Re:Is that a joke?

[Jeng]

How meaningless is a overclocking speed? It's like saying: Your Smart will go 400km/h... if only we run it as ten bazillion RPM. It's still a Smart!! And you will never get this in real life!
This is damn close to fraud, to spread such bullshit so people get a false feeling of it being so fast.

So please, tell me your thoughts in regards to auto-racing.

Re:Is that a joke?

[wmbetts]

They have an irrational fear of right turns?

Re:Is that a joke?

[Jeng]

Odd, I was going to type NASCAR, but I ended up typing auto-racing.

How did you respond to my first draft?

Re:Is that a joke?

[Rogerborg]

You said "auto" == you are a Yank == your "auto racers" go: while(!assploded) { go(left); go(straight); }

Re:Is that a joke?

[Jeng]

I tend to watch F-1 and the Rolex series. The only interesting part of watching NASCAR are the crashes.

Re:Is that a joke?

[Anonymous Coward]

As a beer drinker, I find your statement offensive.

Re:Is that a joke?

[VortexCortex]

> you are a Yank
That makes you a Wank.

Re:Is that a joke?

[MightyYar]

The part of the country that is into NASCAR would not appreciate being called a Yankee... entirely different connotation down there - something about the invention of "total war" occurring on their land.

NASCAR started with liquor running during Prohibition, so speed was more important than handling. Given that, I can kind of excuse the initial use of an oval track. But once the speeds got too high, they started restricting the cars and so the oval track was sort of an anachronism... very hard for me to get into oval track racing. That said, I'm sure it is fun as hell to drive a car around an oval track (I've only done an autocross course, which is terrifying in a good way).

Re:Is that a joke?

[AngryDeuce]

NASCAR is fucking retarded. God, what I would give for some real Rally racing here in the States. It's so much more interesting than watching cars do hundreds of laps around a big fucking oval.

As with realistic overclocking, Rally racing has some relevance to the real world. They're driving on real roads, in varying weather conditions, unlike the stupid NASCAR bullshit.

Re:Is that a joke?

[amiga3D]

I lost interest in Nascar when they no longer represented real cars. I'd like to see them go to the showroom, pick out a car and race the thing. Some safety upgrades maybe but it must be "stock" at least the fucking shell should be anyway.

Re:Is that a joke?

[amiga3D]

Maybe we could get the Canonball run going again?

Re:Is that a joke?

[AngryDeuce]

I agree, that would also be pretty damn cool. I watched the show Bullrun when it was on the Speed Channel a few years ago, but it was annoying with all the added drama nonsense that reality TV does these days, and the challenges were pretty retarded, too.

Just a bunch of teams, drive whatever you want, Point A to Point B. Good stuff.

Re:Is that a joke?

[Anonymous Coward]

Couldn't agree more with you. Posting AC so you get my mod point.

Re:Is that a joke?

[wmbetts]

I hacked the gibson.

Was it actually doing anything?

[Anonymous Coward]

While this is an achievement, what bothers me about these records is that most of the time, they aren't really doing anything at that speed.

Re:Was it actually doing anything?

[Jeng]

Not only do they not do anything at these speeds, they cannot do anything at these speeds except run CPU-Z long enough to get a screen-shot.

Contests to see who can run Superpi the fastest are more interesting.

http://forums.guru3d.com/showthread.php?t=185163

Re:Was it actually doing anything?

[h4rr4r]

Then how do they call it stable?
To me it seems like you would want to test it doing some sort of calculation you can check later.

Re:Was it actually doing anything?

[Jeng]

With the Superpi contests they do, your checksum has to be validated for your score to count.

These max gigahertz records such as tfa, although interesting, are pretty useless. I do have to give them credit for what they do though, it ain't easy.

Does this mean....

[OldGunner]

...the system clock ran 63 times faster than normal? If so, does an hour-long test take slightly less than a minute?

Re:Does this mean....

[dgatwood]

No, it means the CPU clock was 63x the bus speed, which means that the bus speed was a dog-slow 111 MHz. The CPU would not be particularly useful running at this speed because of the slow bus speed.

Re:Does this mean....

[Korin43]

The multiplier on my processor (2 year old i5) was set around 20x by default. I suspect the i7 starts out higher. I think this is less than twice as fast as the non-overclocked version.

Re:Does this mean....

[Korin43]

Oh, and the CPU multiplier is just how much faster the CPU is than the system bus. It's been a loooong time since any CPU's have run at 1x.

Re:Does this mean....

[Cid Highwind]

Standard base clock for Sandy/Ivy Bridge is 100MHz...

Re:Does this mean....

[dgatwood]

Ah. So Intel is somewhat abusing the traditional meaning of a bus multiplier, and we don't actually know anything about the memory bus speed. *sigh*

However, we do know that they're running that base clock more than 10% faster than normal, which probably means that either their RAM is faster than the spec requires or they are running at a slower bus speed than the maximum. No idea which.

Re:Does this mean....

[Cid Highwind]

Yes, with Intel putting all the traditional northbridge functions on the CPU die now, that 100MHz only drives the PCIe bus directly. There's no FSB to speak of anymore.

Pfft.

[busyqth]

That's nothing. My signal generator will do 15Ghz easy... maybe even 20.

Re:Pfft.

[Anonymous Coward]

Can it do math?

Re:Pfft.

[Anonymous Coward]

Yep. Plug whatever you want to multiply into the MOD port, and you get sum and difference!

Re:Pfft.

[Anonymous Coward]

Yeah, but the output is always 10

From the article

[Anonymous Coward]

"These new world records highlight our belief that top notch quality and design deliver truly world-class, record breaking performance," commented Tim Handley, Deputy Director of Motherboard Marketing at Gigabyte.

Moments later, Deputy Director of Motherboard Engineering at Gigabyte commented, "Yeah, um...It of course could be overclocked but we don't recommend it. Essentially, you are only running one core at this new speed and the amount of time and energy required to do this doesn't make any sense. It's quite useless actually."

7 GHz boot

[value]

7 GHz processor speed and it still takes 10 seconds to load Windows. . .

Re:7 GHz boot

[Anonymous Coward]

It seemed like the limiting factor was the speed of the logo animation

Re:7 GHz boot

[Anonymous Coward]

I wonder how many vacuum flasks of liquid nitrogen it would take to play a game of COD?

Re:7 GHz boot

[Anonymous Coward]

My i2500k clocked at 4.6 GHz boots windows 7 in 4.5 seconds. You do need to configure "No GUI boot" with msconfig though, and be using an SSD.

Re:7 GHz boot

[ideaz]

No overclocking configuration could get over sleep(10)

Re:7 GHz boot

[jones_supa]

I have sometimes thought whether games that show some information during loading will run on future CPUs too fast for it to be observed.

computing power

[Anonymous Coward]

and he used all this computing power to ... let me guess - send a tweet upstairs to his mother telling her what he wanted for dinner.

Standard speed == 3.5 GHz

[cpu6502]

Not too impressive. My 10-year-old Pentium 4 is almost as fast (3.2 GHz).

Re:Standard speed == 3.5 GHz

[Anonymous Coward]

Except your not even close to as fast.....

Re:Standard speed == 3.5 GHz

[Henriok]

Yeah because Hz == Hz? Hmm.. welcome to the MHz Myth. You probably haven't heard of it, but you should. It's like this: Hz a measure of steps per time, like counting number of steps mer minute taken by someone who runs. Imagine a short person and a tall person that take an equal number of steps. Who runs the farthest? The tall man or the short man? Pentium 4 or a Ivy Bridge Core i7? It's not your 10 year old Pentium4.. In case you cot lost in the lesson somewhere.

Re:Standard speed == 3.5 GHz

[Anonymous Coward]

Can I have a cot too? I really could use a late afternoon nap...

Re:Standard speed == 3.5 GHz

[cpu6502]

Yes.
I was being tongue-in-cheek.
I know my P4 isn't as fast..... I know that every time I try to play an HD-quality movie, and it runs at quarter speed. The CPU is fast in cycles but slow in calculations.

Re:Standard speed == 3.5 GHz

[Anonymous Coward]

Hmm.. welcome to the MHz Myth.

Welcome to the whoosh.

Re:Standard speed == 3.5 GHz

[geekoid]

Half = almost now.

Good to know.

Re:Standard speed == 3.5 GHz

[Anonymous Coward]

In the future, just use the hammer/nail analogy. Much easier to understand than that train-wreck you ventured in to.

I can hammer a nail into a block of wood using many light swings or one big crushing blow.

Clockspeed is how many times you hit the nail. Performance per clock is how hard you hit the nail.

You may "vibrate" the hammer against the nail, which technically is hitting the nail over 10 times a second. Not much work gets done though as the nail hardly moves.

This record is essentially vibrating the hammer against the nail. You've reduced the amount of work so much that almost nothing gets done so that you can achieve these speeds.

amd did better with 2 cores on a bulldozer

[Anonymous Coward]

AMD did better with 2 cores, now I would like to see a side by side comparison of actual tasks.

link

Re:amd did better with 2 cores on a bulldozer

[beelsebob]

Except that AMD's 2 cores aren't actually 2 cores. They're 1 instruction fetch unit, 1 instruction decode unit, 1 floating point unit, and 2 integer units, so this is a very similar result, it's one CPU core that is able to do a bit of stuff at the same time as it does some other stuff, not really 2 cores.

AMD hit their record with all cores enabled

[Anonymous Coward]

AMD hit their record with all cores enabled.... intel requires one core disabled or else you will brick the CPU regardless if it's LN2 cooling.

Also, Gigabyte gave them this "special" motherboard.

Re:AMD hit their record with all cores enabled

[Anonymous Coward]

AMD hit their record with all cores enabled.... intel requires one core disabled or else you will brick the CPU regardless if it's LN2 cooling.

Also, Gigabyte gave them this "special" motherboard.

Actual AMD had to disable all but one core to hit the speed's as well.

Re:AMD hit their record with all cores enabled

[TeknoHog]

Also, Gigabyte gave them this "special" motherboard.

So it was like winning in "special" olympics.

Re:Cookie Crunch

[ThePeices]

That's a hell of a cookie crunch

Oh that was pretty lame. Try harder, dont be lazy. Effort spent is wittiness gained.

And your 10 year old P4 can do 256-bit vector math

[Sycraft-fu]

Oh wait, no it can't. It also isn't quad core, and does about 20% of the work per clock.

There are limits to GHz scaling. It isn't a situation of "Oh just make it faster," particularly if you want to hit a power budget. What has happened is that CPUs have gotten much more parallel, much more efficient per clock, and have gotten much better at vector math. My Sandy Bridge processor pulls like 80 Gflops on Linpack using AVX. Try that on a P4, let me know how it goes.

CPU companies aren't interested in optimizing for high GHz at the expense of thermal and computational efficiency. They want CPUs that do more, and do it on a power budget.

Re:And your 10 year old P4 can do 256-bit vector m

[geekoid]

Things for you to do:
1) Go outside
2) walk to the nearest store
3) by a humor detector.

OMG your so funny

[Anonymous Coward]

seriously you have to be the funniest motherfucker on the planet. 10 seconds is still a pretty good boot time. Ubuntu takes about as long as windows 7 for me, and windows 7 does not default to 640x480 resolution when my tv is connected as a second monitor like linux does.

Re:OMG your so funny

[buck-yar]

Reminds me of the 1996 usenet thread "Whoa! Win95 boots in only 3 seconds!" crossposted to about a thousand newsgroups and started the biggest flame war in history to that point.

Re:OMG your so funny

[buck-yar]

Here's the thread (though won't show the original that started the war) Whoa! Win95 Boots in only 3 seconds! From google groups usenet archive

Where is his "so funny"?

[Anonymous Coward]

Just wondering...

Pointless

[Anonymous Coward]

I remember when modding was interesting and expensive.
I now find it pointless. I just want my computer to work, I don't care how fast it is anymore. Sorry :(

Re:Pointless

[jones_supa]

Maybe you simply don't need more speed and thus don't get excited about tuning performance? I haven't lately found much of use for anything improved than a basic Core 2 Duo.

My middling, mostly working UD5H.

[game kid]

I own the mentioned UD5H motherboard used for the record memory speed; I bought it to replace a very old P5B Deluxe. I am in no way jealous or unappreciative of HiCookie's feat, and the board definitely looks like something that can handle such a thing, but my experience with the board has been middling.

I haven't had the freezes that people have mentioned on its Newegg page (thank the gods!) and things generally work, but Windows 7 64-bit simply refuses to hybrid sleep or hibernate, and after a non-hybrid standby to RAM, things subtly fuck up (no audio, and other devices I forget at the moment mess up), which means I have to fully reboot (really fun when waiting for big programs like Catalyst) or leave the rig on at FULL POWAH through the night or whatever. Arch Linux was working well at first (RAM standby and even disk hibernate if properly configured and I choose to boot from the Linux drive after the suspend), but updates seem to have made it less compatible with my audio (audio out works except through the standard green line-out...odd) and TV tuner (not detected), for whatever reason. (I left a few more details on a review on the Newegg page, minus the less-compatible part.) The P5B had no such problems: its audio had lots of RF interference through headphones (the UD5H has beautifully clear onboard audio when it works) but it suspended, automatically resumed from the suspended drive, and otherwise worked nicely.

For me, a "middling" board is worse than a "horrible" one, because at least a horrible would be bad enough for me to undo all the cable connections and screw placements and attachments and all that to trade for something better (a very old backup PC I had started getting POST errors as I built the new one so combined with other factors it made referring to the internet kinda impossible...that was fun). With a middling one I simply tolerate the few problems because it mostly works. *shrugs*

Sorry if that came off as a dumb ramble; just my experience with it.

Heat dissipation

[Shag]

I don't have any liquid nitrogen, but I really, really like french toast. On a MacBook Air with an Ivy Ridge CPU, how many slices would I need to cook simultaneously to match the effects of using LN2?

Re:Heat dissipation

[rubycodez]

For maintaining proper cooling, I'd only recommend thawing pre-frozen french toast on your macbook. Do the actual frying on an Itanium2 or something.

He still won't get laid.

[Anonymous Coward]

Fucking nerd bullshit.

Re:He still won't get laid.

[FranktehReaver]

You are on the wrong side of town man.

You people are embarrassing.

[PwnzerDragoon]

You people are embarrassing. A guy manages to overclock a processor to 7GHz and all you can do is bitch about how you can't do anything with it. Do you make fun of people who climb mountains or build with legos because there's no practical purpose to them?

Nobody is saying this is useful. It's just some guy saying, "Hey look, I got my processor all the way up to 7GHz!", stop taking things so seriously.

Re:You people are embarrassing.

[Anonymous Coward]

Except 7Ghz isn't that fast...

Others are at 9Ghz.

http://flyingsuicide.net/news/9-ghz-barrier-falls-hard-amd-hits-milestone-with-9062-mhz-validated/

Still 7Ghz is fastish..

Re:You people are embarrassing.

[Antarius]

Well, actually I find this rather boring and unimpressive.

I can remember the days when every Mhz made a difference and you had to turn off Turbo for some programmes.

Back then, reaching 100Mhz was a huge milestone; But then the GigaHertz wars began.

It was actually interesting to see how well someone overclocked their Celeron 300A or their Athlons. As the Mhz crept up to 500Mhz, 600Mhz, 700Mhz, we were all getting excited, waiting for the day that one of the x86 giants would hit the Holy Grail of 1,000Mhz!

Then Intel threw their hands up in the air and went "Fuck it!" and pushed the P4 bullshit; clock speed became meaningless as the Athlons ran rings around the P4 despite the significantly higher clock speeds of the P4s.

Since then, it's largely a non-event. The architectures are so different that MHz/GHz ratings aren't useful to compare between brands - not like when you could know that an AMD 80486 DX40 was going to be faster than the (more expensive) Intel 80486 DX33. (Anyone remember the disdain and contempt the brands and magazines pushed on the Bus multipliers back then? "Don't buy a 486 DX2 50! It's clock multiplied! Buy a real 486 DX 50 instead!")

So, yeah... Compared to the days where someone would brag about a stable overclock of x MHz more than stock, these e-peen measuring contents for the highest GHz is nothing more than being a total tosser. Show it actually running stably and applications, or I couldn't care less.

POWER6 was at 6Hz in 2005

[Anonymous Coward]

Pretty impressive, although IBM has been shipping 5GHz POWER6's for years, and it has been verified at up to 6GHz.

False

[thegarbz]

Actually AMD had only 2 cores out of 8 running to hit their record speeds.

That said I fail to see the excitement for this news. This is only a record for Ivy Bridge chips and AMD's attempt managed to beat it by more than 1GHz.

Re:False

[shaunbr]

Now if AMD could just get their processors to be competitive with Intel at normal speeds. By the time AMD gets to Ivy Bridge level performance, Intel will be 2 or 3 generations further ahead.

Re:And your 10 year old P4 can do 256-bit vector m

[Anonymous Coward]

Given the source (cpu6502), it's really not that surprising that people thought he was serious.

Doesn't mean much practically

[Anonymous Coward]

The problem with this is that every minimal geometry IC today is on the hairy edge of HCI, BTI, EM and other end of life reliability failure stresses even at normal voltages. The typical targets at the moment are 10-15 year operating lifetimes at spec'ed Vdd values. The problem is that lifetime accelerating stresses follow square-law, negative reciprocal and exponential relations to voltage, current and heat. That means that operating so high (yes 1.956V over spec is insanely high in this context with today's device dimensions) that the likely operating lifetime is hours to weeks at most. At that point the failure is not an infant failure nor is it a random midlife failure, but rather it is end-of-life terminal wear-out failure. See Reliability Bathtub Curve

The primary reason over-clocking is "dead" as a practical hobby is that the "headroom" afforded by reliability physics for over-voltage has monotonically decreased with device dimensions. It's now only about half a volt at best for reasonable operating lifetimes (~5 years) and far less above that much over-voltage.

Re:Doesn't mean much practically

[Anonymous Coward]

Should also point out that simply operating at LN2 temperatures does NOT remove accelerating stresses because voltage and current still operate as stresses regardless of temperature.