P4 2.2GHz Overclocked to 3.5GHz

GraveD sent linkage to a site explaining how a homemade nitrogen cooling system overclocked a P4 from 2.2Ghz to an incredible 3.5ghz. There's plenty of stuff to poke at over there. Update: 01/17 20:42 GMT by T : boaworm writes: "According to this paper, the Finnish geeks have successfully oveclocked a Pentium 4 to 3675 Mhz. They claim it is a new World Record, and it sure looks like they beaten another O/C'd Pentium 4 submitted earlier today on slashdot. (Summary in English in the end)."

Comments

Mirror:

[Saint Aardvark]

http://saintaardvarkthecarpeted.com/oracle

Re:I know, let's make the story something it isn't

[leuk_he]

We use them in some extremely large and significant datawarehousing situations and have probably managed to kill the server once in three years

Then you certainly have not tried the following thing:
-Install forms 4.5 now. Forms 4.5 is year 2000 complient, the installer crashes (!)(there is a patch but.....)
-use plsql records in a 7.3.4. DB. Bye bye instance.
-use designer 2000 1.2.

and the list goes on and on.

2 cents

[f00zbll]

Ok, I read the article and here's my thoughts as worthless as they are:

revealed that a common programming error -- a buffer overflow -- was present in Oracle's application server, potentially allowing hackers to gain remote access to the system over the Internet.

If the researcher is referring to Oracle 9i application server, it's really Orion Server. Since Orion is pure Java implementation, the threat is pretty low. Reguardless, the Orion developers will fix it. They're pretty quick about bug fixes.

We can actually interject ourselves in between that communications process and run commands as SYSTEM on Windows NT or 2000. If it's running on a Unix system, we can run commands as the Oracle user remotely

I'm not sure what this bug is referring to specifically, but it most likely is related to Oracle's GUI administration tool. If the user can run Unix commands, that doesn't necessarily mean a person can erase all the data. The suggested installation is to have the server run under the Oracle user. If ownership is root and the priv. is execute only, an instance would only be vulnerable to "kill -9". To erase the actual data, the cracker would have to login to the instance and delete the data.

I've done some crazy tests with sql server 6 and oracle 8i on low end hardware and I have to say oracle out performs sql server hands down. This is no excuse for Oracle though. They still need to back up that slogan with real blood.

Re:Larry Ellison is The Rock

[lunenburg]

It doesn't MATTER who The Rock is, jabroni! :-)

Re:Buffer Overflows Myth

[Vulture_]

Better a denial of service attack than all of your customers' credit card numbers getting posted on USENET. Oh, and that's what firewalls, IDSes, and sysadmins are for. Denial of service attacks are easy to notice, easy to block, and easy to perpetrate.

Regardless, that code was intended to run in its own thread. If you don't have such a configuration, you'd better use timeouts or async I/O or something. Otherwise you will be only able to service one client at a time... Java 1.4 (currently in beta) adds many new facilities for asynchronous I/O and a select() type thing.

Re:Buffer Overflows Myth

[Tom7]

Well, first of all, buffer overflows are far far more serious than denial-of-service attacks, since buffer overflows can lead to a compromise of your system, and DOS attacks do not. Typically, the attacker needs to continually expend resources in order to carry out a DOS attack, as well.

Second, being free from buffer overflows (and other perils of C programming like manual memory management and core dumps) gives you more time to spend looking over your program for other kinds of security problems. This is good, of course.

Java is not the greatest language, for sure, but safe high-level languages ARE the future, whatever they are named.

Re:Operating systems

[sql*kitten]

Other are just ports.

Well, yes and no. Oracle is developed in two layers, VOS or "Virtual Operating System" abstracts all the primitives like threads, pipes, file handling etc from the underlying OS, and Oracle itself, which is written to VOS APIs. So the core Oracle engineering team code for pure functionality, and the VOS teams keep their APIs in sync with each other on different platforms. If Oracle want to target a new OS or platform, they simply develop a VOS for it.

I believe the Oracle engineers work on Suns, but they are targetting VOS, not Solaris directly.

That's why you have to start the service before you can start the instance on NT. Win32 is sufficiently different from Unix-like systems to need an environment in place before starting Oracle, whereas Unix-like systems can just link the VOS into the main binary. It needs to work like this because Oracle is Oracle, on any platform, once you log into SQL*Plus, it's exactly the same. Oracle is more complex than many operating systems, it provides its own scheduling, resource quotas (storage and CPU), IPC mechanisms (AQ, DBMS_PIPE, DBMS_ALERT, etc), programming languages (PL/SQL and Java) and a whole lot more. It is a platform in its own right.

Those crazy Finns

[milkmandan9]

Are at it too.

Here you can see they've got it to boot at 3.674GHz. The page is in Finnish (I assume), but there's some English text at the bottom too.

link to the most important part

[carlcmc]

the picture of the results that ISNT IN JAPANESE.

Re:Neat, now how about my box...?

[Anonymous Coward]

The Commodore 64 actually had a 6510 ... :-)

Re:Neat, now how about my box...?

[hattig]

You can overclock the G4's in the Power Macs if you know how. The multiplier is encoded by some resistors near the CPU on the CPU card, and if you know the layout, you can overclock your 800MHz G4 reliably to 933MHz or even 1GHz. I don't know if you can do the same with the new iMac2, I reckon there is a good chance of it once someone finds out where these resistors are on the motherboard.

Amigans have been overclocking their 68k series processors for years. Witness the 28MHz 68000 for the A500, or the 75MHz 68060s (instead of 50MHz), a 50% overclock easy when decent coolers are added to the equation.

It is harder to overclock the 8-bits, as the rest of the system messes up in many cases, and the video output and audio go haywire. But it has been done (Enterprise 64 in one example, upping the 5MHz Z80 by a MHz or two, or replacing it with ones that do 10's of MHz I believe. Dunno about the C64 or Atari 8-bits though.

Jumpers?!

[Anonymous Coward]

Dood you are living in the past, it's all soft-menu now. Most overclockers out there wouldn't know a jumper if they saw one.

I'm amazed

[k98sven]

That the thing still functions at 77 Kelvin.
Incredible that the motherboard doesn't break, at that low
temperature, the resin should undergo a phase transition and become very, very brittle.

(Some notes for all those D.I.Y.ers out there:
Liquid nitrogen is cheaper than milk.
Short-circuits can't occur, N2 doesn't conduct.)

Although why he used nitrogen and not dry ice, which is cheaper, easier to handle, and probably
better for these purposes, beats me.

Re:Compare it to an Athlon

[kawaichan]

Do you want to konw why P4 is slower than Athlon? Probably one of the main reason is not the subsystem in your case, since RC5 only utilizes your CPU, P4 has a longer pipeline than your Athlon, making the CPU doing less calculation per clock.

Long Pipeline does have an advantage however, longer the pipeline usually mean higher Mhz.

G3 overclocking - all the rage.

[Elazro]

Not only can you overclock PowerPCs, but when G3 upgrade cards came out for older PowerMacs, some (notably PowerLogix) had thumbwheels on them to control bus speed/multiplier settings. No yanking of boards, no dropping jumpers into the dark recesses of the case. Shutdown, rotate the dials, startup.

Overclocking for the rest of us.

-matt

Re:Neat, now how about my box...?

[larien]

Of course, the problem with overclocking something like a speccy or C64 is that you're likely to speed up the gameplay of anything you're running! These systems didn't have the same kind of clock as modern PCs so timing was handled by running NOPs (or whatever). Instead of increased frame rates (or possibly as well as), you have a game running twice the speed! Sometimes you might want that, but you probably don't.

As an aside, I bought a game ages ago that must have been written for a 386/486 and ran it on my P233 (as it was at the time). The game was unplayable because of the speed. I dread to think how it would run on my Athlon 1800+XP... *shudder*

Re:Other overclocking

[Omnibus]

There is a Palm app called "afterburner" that is used to overclock palm based pda's.

The Mhz Myth

[nihilist_1137]

Here is a Apple Video of the megahetz myth. It basically explains why a bigger Mhz doesnt always mean more perforance when it comes to things like pipeline length and recursive instructions.

Re:Compare it to an Athlon

[rschwa]

P4 lacks a certain shift instruction in hardware, it's emulated, and it's the biggest part of the rc5 algorithm.

Re:Overclocking with super-cooling systems?

[irecleas]

The disparity is caused by the length of the P4 pipeline in relation to the AthlonXP. So it's a ratio. 2000/1600=1.25 or 1600/2000=0.8
So a 3.5GHz P4 is the equivalent of a (3.5*0.8) 2.8GHz Athlon XP

PowerPC overclocking

[Paladeen]

You can actually overclock quite a lot of Mac systems, way back to the 68k's.

F.e. you can OC the original iMacs (don't know about the new ones, but I had one running on 300Mhz, up from 233), the G4 Sawtooths and quite a bit of the older machines and clones.

However, this often requires soldering on or removing transistors on the motherboard, as is the case with todays G4s.

One notable exception to this are the PowerMacs based on the Yosemite motherboard (Blue & White G3 and the Yikes! PowerMac G4, which had a modified Yosemite). They have transistors on the motherboard and its remarkably easy to change the bus speed and clock speed.

For a good source on Mac overclocking, check out www.xlr8yourmac.com.

Re:Neat, now how about my box...?

[jeffb]

The Powerbooks work this way too, except that the resistors involved are smd type instead of the easier to fiddle with ones on the desktop. In fact, on the desktops, you can clock them with a Circuit Works pen and an X-Acto, if you'd prefer not to solder. I've been running my 1st generation iBook (300MHz) at 400MHz for almost 2 years now, and it has worked well since the day I 'clocked it. (processor temp went up an average of only 6F, which is good, since the iBook also doesn't have a fan). The chart reviewing the various combinations of processor speeds and ratios available on the iBooks and Powerbooks is available at The Mystic Room, if you're curious. (or just want to see a 666MHz iBook in the Apple System Profiler, if only for a sec.)

:jeffb Apple Certified Tech

Re:their edge was being in Finland in January

[Anonymous Coward]

If only it were so. You see, we're having a heat wave here. (In Finland that means it's above zero outside) I have to keep the window constantly open or else all the electronic gadgets I have will keep the room temperature at a sweaty sub-tropical 26 degrees.

Re:Overclocking with super-cooling systems?

[Ryan Amos]

Not necessarily. There are a large number of factors that go into the "mHz disparity," including cache, memory, fabrication process, voltage, and heat dissipation. If we were to base the "ratios" solely off pipeline stepping, how would that place a PowerPC G4, with a 7 step pipeline, against a P4, with a 20+ step pipeline? Is a new iMac as fast as a P4 2400? I know they're different architectures, but in actuality the Athlon has a LOT in common with the design of the PowerPC.

The "ratios" are a good guess, and will be reasonably accurate, but as a chip heats up it actually gets slower (i.e. it takes more time for an electron to move through the circuit) and the AthlonXP gets a lot hotter than a P4 a lot quicker. My AthlonXP 1700 practically needs its own air conditioning unit (and why most athlon heatsinks weigh more than the reccomended 300 grams) while, from what I hear anyway, the P4s aren't quite as bad (though not exactly frigid.)

Re:Compare it to an Athlon

[rschwa]

I'm not sure. From The Dnet FAQ

Integral to the mathematics of the RC5 algorithm are 32-bit rotate operations. For whatever reason, the designers of the IA32 (32bit Intel x86) and the PowerPC architectures decided to implement the rotate function as a hardware instruction.

Many other CPUs do not have built-in hardware rotate instructions and must emulate the operation by (at the very least) two shifts and a logical OR. This handicap is why many non-32bit-Intel [1] and non-PowerPC computers run RC5 slower than one might expect based on real-world benchmarks. It is also the main reason why the RC5 client is a poor benchmark to use in determining the speed or performance of a particular CPU.

[1] The IA32 architecture is that used by the Intel 80386, 80486, Pentium, Pentium Pro, Pentium II, Pentium III and Pentium 4 processors. The Pentium 4 does not however have a hardware rotate instruction.

(emphasis mine)

Re:What about underclocking?

[collar]

Underclocking is actually usefull. I used to work for a company that used embedded PC hardware, we used to routinely underclock chips when we knew that they would be going into very hot climates or in places where the ventilation would be poor.