Slashdot Mirror

There was also a very interesting endurance test done on extremesystems.org. Very impressive stuff. I don't yet own an SSD, but I'll continue to consider buying one! Maybe next Black Friday. Just waiting for the right deal.

The other problem is the write cycle limit on the SSDs, that may or may not be an issue depending upon how you use your computer. But, for those of us that regularly recompile the OS and kernel, an SSD isn't going to stand up to that for very long.

You'd be surprised at the endurance of a lot of SSDs. Especially name brand ones like Intel and Samsung (forget the rest).

Some tests by users who do nothing but write to the drives 24/7:
http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm

Like the old Intel 320s ,a 40GB one reported no more life at 190TB written, died at 685TB.
The Intel X25 80GB - worn out at 147TB, died at 883TB.

It's a challenge because they are damn hard to wear out so the newest and latest SSDs really take forever.

The SSD is likely going to be way too small for usefulness before one actually runs out of usable writes.

That SSDs go read-only as they wear down is a myth. I've never seen a single credible report of it happening. If you read real wear-down tests, what actually happens is that the drives stop retaining data when powered off as they get very old. Not a single drive tested there failed gracefully at the end.

How many effective READ/WRITE cycle can the chip in SSD perform, before they start degrading ?

There are user-done studies on such matters and some of them are quite impressive - to the point where you'll scrap the computer first before encountering failure.

The main reason why SSDs fail prematurely is their tables get corrupt. An SSD uses a FTL (flash translation layer) that translates the externally visible sector address to the internal flash array address. FTLs are heavily patented algorithms and there are many of them. The FTL also handles stuff like wear levelling, bad blocks and garbage collection. It's no surprise that the FTL has an impact on SSD performance.

All this is typically stored in some area set aside for it because scanning the entire array is much too slow (takes many seconds, slowing down the power on boot). In addition, the use of stronger ECC algorithms uses up the available spare area of the flash leaving none for the FTL to manage.

So on power up, the SSD controller gets the tables from flash and starts operating extremely quickly. Of course, the problem is what happens when the tables are corrupt - they're usually held in DDR SDRAM for speed, but they need to be synced to the media periodically, and that's where problems happen (because you don't want to wear out the table blocks, as well... so you need the FTL to manage *that*...).

So if you manage to pull power to the SSD at the wrong time, there's a strong chance it'll corrupt the tables because it's in the midst of syncing them to media.

Good SSDs avoid this by structuring their writes, or more commonly, using onboard capacitance to provide emergency power which can flush out the memory buffers to storage quickly. On some SSD teardowns, you can see a huge row of 10+ tantalum capacitors just for this purpose.

You mean this test where almost all the drives are used at very low amounts filled, and the drives that are used with large static data fail in extremely short periods of time?

Found it: http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm

It's a bit out of date, but basically: Stay the hell away from OCZ and certain Intel drives, and you'll be fine in nearly all cases.

The lifespan issue with SSDs has three main factors.

1: Type of flash memory (SLC, MLC, TLC, in order of decreasing durability)
2: Size of the flash drive (larger drives have more room for wear leveling algorithms to work with, thus staving off flash cell burnouts due to exceeding maximum number of writes).
3: The amount of throughput on the flash drive. An expected heavy load is roughly 10GB/day. Doubling the load halves the lifetime of the drive. Quadrupling the load quarters it.

Granted, the cache on a Hybrid is being used a bit differently than how you would use a straight SSD. But, with such a small cache drive, you ARE going to wind up cooking it after a relatively brief period of time.

Which for most users and usage scenarios, is basically forever. There's been a volunteer-run test of longevity which stresses an SSD until it fails by writing data to it continually. And the SMART data typically gives you plenty of advance warning - the Media Wear Indicator (MWI) tells you how many cycles are left in the array - once it hits zero, it means the number of write-erase cycles has hit the guaranteed limit and you're running in unknown territory (though there are usually still spare blocks and most will still have plenty of life). If you want guarantees, once the MWI hits zero, it's time to back up and get a new SSD. The tests run until the drive itself dies which tell you how long you have left. So you generally have a LONG indication of media wear out.

However, the biggest problem SSDs face is actually sudden loss and corruption of the FTL tables (the ones that map logical sectors to actual flash sectors). If you hear of SSDs dying prematurely, it's almost always because of table corruption. These tables contain things like sector translation, sector wear, dirty/clean bits, trim status, etc.

In the past, you could regenerate the tables from the spare area data (typically 16 bytes per 512 byte data area), but use of enhanced ECC algorithms consume that space up to accommodate better error handling. Plus it also meant way longer mount times as the controller had to scan the entire media for the information (many seconds long).

These days, controllers come with 512MB or more of RAM to hold the tables in memory for quick access. The problem is the tables are often written out lazily to storage, which means if you yank the power suddenly, the SSD might not be able to write the dirty data to stable media, or worse yet, it'll be in the middle of the write operation which leaves data in an unknown state.

Good SSDs often have piles of capacitors to serve as emergency power which can keep the array powered for a couple of seconds - more than enough time to flush the tables to storage and protect your data. Of course, this costs a lot more money and is usually present only in the top tier drives and enterprise class SSDs. If an SSD dies suddenly, it's usually because of this.

Hard drives use the back EMF produced by the spinning platters to perform emergency shutdown procedures, including retracting the heads.

The failure mode of flash memory is to become read only. If you couldn't access it after a week it is NOT the flash memory that failed. Most likely it's the controller

Is it? Not one of the drives being tested in the Xtremesystems endurance test had a failure mode of "went read-only". Drives have failed to retain data, they've corrupted data, they've failed to be detected by the OS, but they've never entered read-only mode.

100000 writes? 1M writes?

What the fuck is this submitter smoking?

Newer NAND flash can sustain maybe 3000 writes per cell, and if it's TLC NAND, maybe 500 to 1000 writes.

Actually, NAND flash doesn't "die" when you try to do the N+1 erase-write cycle (it's cycles, not writes. A cycle consists of flipping bits from 1 to 0 (aka write), and then from 0 to 1 (aka erase)). In practically all controllers, you do partial writes. With SLC NAND, it's fairly easy - you can write a page at a time, or even half pages. MLC lets you do page at a time as well - given typical MLC "big block" NAND of 32 4k pages, a block can be written 32 times before it's erased (once per page - you cannot do less than a page at a time).

And... other dirty little secret - the quoted cycle life is guaranteed. It means your part will be able to be written and erased 3000 times. Most typically, they're an order of magnitude more conservative - so a 3000 cycle flash can really get you 30,000 with proper care and tolerance.

Of course, a really big problem with cheap SSDs is lame firmware because what you need is a good flash translation later (FTL) which does wear levelling, sector translations, etc. These things are VERY proprietary and HEAVILY patented. A dirt cheap crappy controller you might find on low end thumbdrives and memory cards may not even DO translation or wear levelling. The other problem is the flash translation table must be stored somewhere so the device can find your data (because of wear levelling, where your data is actually stored versus where your PC thinks it is different - again, the FTL handles this). For some things, it's possible to just scan the entire array and generate the table live, but generally it's impractical at the large scale because it requires time to perform the scan. So usually the table is stored in flash as well, which of course is not protected by the FTL. Depending on how things go, this part could corrupt itself easily leading to an unmountable device or basically, a dead SSD.

For some REAL analysis, some brave souls have been stressing cheap SSDs to their limits until failure - http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm

Some of those SSDs are actually still going strong.

The best bet is to buy from people who know what they're doing - the likes of Samsung (VERY popular with the OEM crowd - Dell, Lenovo, Apple, etc.), Toshiba, and Intel - who all make NAND memory and thus actually do have experience on how to best balance speed and reliability. Everyone else is just using the datasheet and just assembling them together like they would any other PC part.

This chart is almost 2 years old now, but it is a fun read and has some good testing information:
  http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm

I have read that some of the newer SSD have only 500-1000 P/E cycles (eg. Kingston V300, Samsung 840), but I don't have proof. It is well documented that most of the current MLC drives have 3000 to 5000 P/E cycles while may of the SLC units are 100000 (eg. Intel X25-E, SuperSSpeed SLC S301).

Here is another good article about TLC SSD:
  http://www.anandtech.com/show/6459/samsung-ssd-840-testing-the-endurance-of-tlc-nand

You have to buy and use the correct type of SSD for your application. A new TLC SSD should not be used in any write intensive application (eg. ZFS ZIL) but it may be great for that new fast laptop that can use the speed and does not do a lot of writes to disk. For most standard uses a good SSD will outlast the laptop/desktop where it is installed. The key for good SSD use is detection of pre-failure (SMART is a good start). The SSD is now a consumable part, just like the battery or brakes on a car. We all know drives fail, but standard hard drives don't have the same fixed life expectancy as an SSD.

Don't forget about Write Amplification. It can help kill a drive faster than total bytes written:
  http://en.wikipedia.org/wiki/Write_amplification

There's a lot of conjecture and theorising in this thread so far. Not surprisingly some enterprising geeks have been busy testing SSDs to destruction, and they have some great stats. This thread with over 5000 posts has a ton of info about exactly what happens and some good hard numbers.

There are a variety of different ongoing tests to look at how long drives actually last. Looking at a fairly standard older Intel 320 40GB drive, it went 190TB written before the MWI threshold was reached, and continued on until 685TB. That means it completely rewrote the drive 17500+ times.

Yeah, but the problem is they keep reducing the size of the flash cells. It's getting to the point where they're actually having to expend write cycles to prevent the data evaporating, and with smaller gates you get significantly less write cycles to begin with, and once you start going from MLC to TLC, it starts to look kind of ugly.

That and I am kind of worried about data retention in these things. I have some EPROMs from 1988 which were rated for about 10 years, though the contents were still readable (and the machine still worked) in 2010 when I made a backup. I'd be impressed if 20nm TLC flash can retain data for 5.

What also is not addressed in the article is the reliability of the SSDs. Flash ram is not a permanent solution and will die due to the limited number of writes. If you use mysql or MS access or run low on space and use XP that thing will be dead in a matter of months. It can only handle so much paging and writes before it dies. Tricks in the firmware move the write bits to random places in memory to prevent this but as it fills up the paging needs to keep to keep hiting the same memory addresses.

There are a variety of different ongoing tests to look at how long drives actually last. Looking at a fairly standard older Intel 320 40GB drive, it went 190TB written before the MWI threshold was reached, and continued on until 685TB. That means it completely rewrote the drive 17500+ times.

No, it won't last forever. And it's not ideally suited for every single industry and use. But for the typical user, they are more likely to need a larger drive or otherwise upgrade then wear out the drive.

Common flash has nothing like a 2M write cycle limit. More like 50K for the older stuff, as little as 3K for the newer. This guy is talking about hand-selecting the chips, building a custom drive system from those raw chips, and writing his own drivers for a very undemanding and fault-tolerant application that he puffs up to seem like it's something fabulous.

Actual writes before failure have been tested to be about 200-1000TB for a 64GB 25nm disk, which is about 3000-15000 complete writes, or a few weeks of maximum use. (and bigger disks are not necessarily better.) Yes, it's not a limitation for most people, but the drives sometimes fail early and become totally inaccessible and unrecoverable.

Why have it go with the furniture when it can be the furniture?

Its actually quite simple, Intel got away with rigging the markets. First they bribed the OEMs to take the P4 and kill the Athlon which is why when the Athlon was stomping the P4 you couldn't find hardly a single Athlon but you could find Duron/Sempron because part of their bribes were based on how many you sold of the competitor's chips and they were given a higher quota of the weak chips. Now how anybody, when one of the former CEOs called Intel kickbacks "like cocaine' and during the MHz wars there were several quarters that dell wouldn't have shown a profit without Intel kickbacks, how anybody can say they shouldn't get busted for that is beyond me, this is even worse than what MSFT was doing with the OEMs.

Secondly to this very day you can take a Via CPU (the only chip that lets you change CPUID) and any of the major benchmarks, change the CPUID to Intel and watch the chip score 30% or higher than the exact same CPU which is because Intel rigs their compiler. this is also well known and documented and goes back years, for those that haven't read it here is a link to get you started. BTW note that even though they were supposed to remove the "cripple AMD" function instead all they did was documented it, and certainly not in any easy way to find, at least not on the compiler website last i checked. To compare this would be like rigging Windows so that when FOSS code is run it hangs and drags down the FOSS programs to make them inferior to MSFT's own programs. this is still affecting reviews to this day such as this quote I remember from a review of netbooks "The benchmarks we ran all say the Atom 525 beats the E350 by a very large margin but for some reason the real world tests don't seem to bare this out" sorry i can't remember offhand where, i believe netbookreviews was the name of the blog.

Both of these frankly should have gotten Intel seriously busted for antitrust but instead they were able to slip a check for 1.25 billion to AMD, which frankly was far less than they made by crippling their competitor and forcing AMD to sell its fabs just to stay afloat, and all the problems at least in the US went bye bye just like TFA. I was someone that bought Intel exclusively since the 386Dx but i also believe in having a fair market and I simply can't support outright market rigging of this level. Frankly Intel should have been busted just like MSFT and been watched like a hawk for a decade to make sure they couldn't pull this kind of crap. i only hope the EU busts the hell out of them because its obvious in this "corporation yay!" climate we have in the government now there is no way Intel has to worry about the USA saying anything, no matter what they do. hell if the MSFT antitrust came up in today's climate not only would they not have been busted, they probably would have been rewarded with more tax breaks!

You'd think with one party supposedly championing the free market as the solution to everything they'd care about someone subverting it but i guess the only free market they care about is the one where they can sell their services to the highest bidder.

The Japanese page cited does not actually have any further details. It does have some general information about Pi and describes the man's earlier 5 trillion digit record, which was done with software called the y-cruncher. It is extremely likely that the new 10 trillion digit record is computed with y-cruncher as well, which seems to be corroborated by the fact that the author posted on the xtremesystems.org forums about it.

While Shigeru Kondo (ja0hxv) is a programmer, the author of the pi calculating software is actually Alexander Yee.

Your right....

http://www.xtremesystems.org/forums/showthread.php?217797-Microcenter-does-a-price-match-on-newegg-prices-(and-also-for-many-other-retailers)!

That makes micro center an exception, might even have to go there myself thanks to this thread, online shopping = shipping wait,and I do have a car and shit.

It's a shame it's not more advertised, perhaps it's hard for them to afford to do it on a large scale...

Newegg and other online keep their prices lower by not stocking up 5000 computer boxes in a multi mllion dollar warehouse. They use a 6 figure warehouse, and stock 50 computer boxes expecting to sell 4-5 a day or so.

By price matching micro center is eating the cost of their inventory storage, not sure how well that works for them as I'm not familiar with their markup %s.

This might be of interest to you :

SSD Write Endurance 25nm Vs 34nm

AMD has chosen an architectural roadmap that makes the GPU and CPU part of the same APU. SYSmark does not measure 3-D graphics performance. At all. So while AMD is pursuing a path that will give its APUs greater overall performance than the CPUs they contain, they are actually hamstringing themselves in the CPU-only testing arena, because the CPU portion of thier APUs will seem relatively lower in performance at the same price point.

AMD's proper course of action should have been to promote an APU-specific benchmark. Instead, it tried to change SYSmark to do something it doesn't do.

It was denied the right to twist the benchmark in its favor. Rather than coming up with the obvious solution of spinning off a new benchmark consortium to develop an APU-specific test, it started crying and ran to its room shouting, "I hate you! I hate you! I hate you!"

AMD is, really, behind a major 8-ball right here. It has, again, put all of its eggs into a rather hopeful basket, and come up with fewer than expected. At least this time, unlike with the Barcelona debacle, it isn't doing it while roller-skating blindfolded through a car-wash. That time it cost them their fabs. They don't have much left to sell.

It's little wonder that it's not having an easy time of finding a new CEO.

To be fair, I don't know of a better way to test, and I'd love to see a discussion of better utilities. If I tried this I'd probably do mprime and keep an eye out for MCE's in the system logs, but don't delude yourself into thinking that core is error free because you ran prime95.

There are quite a few tools, mainly found in the overclocking communities. OCCT, Linx and Intel Overburn just to name some.

I was there, too. The coolest it got was approximately -242 degrees C; the warmest was approximately -218 degreesC, at least while I was watching.

The party was the XtremeSystems.org party at its LV headquarters, and it was sponsored primarily by AMD, DFI, Gigabyte, Cooler Master, and Thermaltake. It seems to me that Commodore had a presence there, too.

See ThinkComputers' blog for some more pictures (disclosure: my article).

http://www.xtremesystems.org/Forums/showthread.php?t=201723&page=7

My experience with one of these boxes shows performance numbers that are alot different from those shown. The RAID0 configuration stripe size makes a huge difference. This should have been discussed. Hard drives typically have 512 byte sector sizes. But with RAM, all requests are in 4k blocks. I'm not sure how the Acard hardware works to make it all work out, but 16k stripe sizes seem optimum. Who knows if the ANS-9010 interleaves the data or how they make it all work out.

The benchmarks provided at xtremesystems.org are significantly faster than what Techreport shows.

Look at the benchmarks provided for different stripe sizes. I have to wonder what stripe size they used for their benchmarks at Techreport.

Yes, those benchmarks provided there are from me. That's why I'm posting anonymously.

Nice trick doing this to a laptop mobo, but LN2 cooling is hardly novel. Xtreme Systems has had a forum dedicated to LN2 cooling for years, and guys there are doing it everyday.

Interestingly, the site owner did an overclocking demo for Intel. I guess Intel can't say they are anti-overclocking anymore.

That forum will answer all of your questions. Frankly, LN2 suicide runs are fun, but I am more impressed with phase change systems that can run 24/7 like mine. :D

Nice trick doing this to a laptop mobo, but LN2 cooling is hardly novel. Xtreme Systems has had a forum dedicated to LN2 cooling for years, and guys there are doing it everyday.

Interestingly, the site owner did an overclocking demo for Intel. I guess Intel can't say they are anti-overclocking anymore.

That forum will answer all of your questions. Frankly, LN2 suicide runs are fun, but I am more impressed with phase change systems that can run 24/7 like mine. :D

So how would this CPU compare to a Core2Duo E8400 overclocked to 5GHz?

And will be obsolete in a year. Honestly, who spends thousands of dollars every year for the most advanced stuff?

to answer your question, the people at Overclockers Forums, Xtremesystems Forums, and any other person who tweaks their computer for fun. It's a cheaper hobby than boating, for example.

Phase change and LN2 cooled quads are already running at over 5GHZ. This is just intel themselves overclocking as is mentioned in the article. Is this even news? It's not like Intel is actually going to be selling a phase change or LN2 cooler to go along with their new platform. And even if they were this doesn't sound like any sort of advance in silicon as is implied by the article summary.

Phase change and LN2 cooled quads are already running at over 5GHZ. This is just intel themselves overclocking as is mentioned in the article. Is this even news? It's not like Intel is actually going to be selling a phase change or LN2 cooler to go along with their new platform. And even if they were this doesn't sound like any sort of advance in silicon as is implied by the article summary.

This is equivelant to the modern "ghetto v-droop" mods -- the article addressess much more complex applications :P

To save thoughs who just want to see the setup pictures

Aside from a 2400 node cluster for the department of defense and university research, or a living room media center, I guess not much other than its cool?

"Only a matter of time before some nutter with liquid nitrogen hits 4-5GHz"

Already been done quite a few times by myself and colleagues, not to mention the huge crowd over at XS.org ;-)

Not terribly suprising for a new board, considering that existing boards can already support early kentsfield chips. Including a kentsfield benchmark run on a MSI 975X Platinum 2.b

did you install all the dual core related fixes/patches?

http://www.xtremesystems.org/forums/showthread.php ?t=81429

Let me just point you all to this little forum post:

http://www.xtremesystems.org/forums/showthread.php ?t=107092

Who knows how these guys get ahold of these parts but from the looks of it the Kentsfield trips are blazing. They are basically 2x Core 2 Duo (Conroe E6700) one one die.

At ~3.5 GHz this guy got almost 13000 marks in 3d mark 06.. WOW JUST CHECK IT OUT ALREADY

doesn't anyone else here frequent overclocking forums? intel's 4-core kentsfield chip has already been put through its paces by several respected overclockers, and its performance is anything but lacking.

http://www.xtremesystems.org/forums/showthread.php ?t=104773
http://www.xtremesystems.org/forums/showthread.php ?t=103982
http://www.xtremesystems.org/forums/showthread.php ?t=105355

multi-threaded benchmarks like cinebench and dvd recompression are scaling linearly - it's quite hard to starve these cpus for bandwidth, unless you're one of those synthetic junkies who just wants to read from memory without actually performing any work on it. the single-chip kentsfield has been outperforming dual-dual-core opteron configurations (i.e. 4 cores on 1 motherboard). so much for bandwidth concerns.

doesn't anyone else here frequent overclocking forums? intel's 4-core kentsfield chip has already been put through its paces by several respected overclockers, and its performance is anything but lacking.

http://www.xtremesystems.org/forums/showthread.php ?t=104773
http://www.xtremesystems.org/forums/showthread.php ?t=103982
http://www.xtremesystems.org/forums/showthread.php ?t=105355

multi-threaded benchmarks like cinebench and dvd recompression are scaling linearly - it's quite hard to starve these cpus for bandwidth, unless you're one of those synthetic junkies who just wants to read from memory without actually performing any work on it. the single-chip kentsfield has been outperforming dual-dual-core opteron configurations (i.e. 4 cores on 1 motherboard). so much for bandwidth concerns.

doesn't anyone else here frequent overclocking forums? intel's 4-core kentsfield chip has already been put through its paces by several respected overclockers, and its performance is anything but lacking.

http://www.xtremesystems.org/forums/showthread.php ?t=104773
http://www.xtremesystems.org/forums/showthread.php ?t=103982
http://www.xtremesystems.org/forums/showthread.php ?t=105355

multi-threaded benchmarks like cinebench and dvd recompression are scaling linearly - it's quite hard to starve these cpus for bandwidth, unless you're one of those synthetic junkies who just wants to read from memory without actually performing any work on it. the single-chip kentsfield has been outperforming dual-dual-core opteron configurations (i.e. 4 cores on 1 motherboard). so much for bandwidth concerns.

"Also, no one outside of select reviewers running Intel-selected benchmarks have seen this chip. The Intel chip was supposed to ship on Monday but it was only a "paper launch". Intel is only taking orders at this point."

No, the DDR2 is not faster than the DDR because it runs at much higher timings! Learn something before you post. Conroe destroys FX. Read some freaking enthusiast sites like xtremesystems.

Check out xtremesystems.org and tell me that FX' aren't dead as Elvis. Not even close. Sell AMD short.

I strongly suggest you check out xtremesystems.org forums, where the benchmarks have even AMD fanboys like yourself are ready to switch to Intel. These conroes have already run SuperPi 1M @ 10.xx seconds. Out of the box they will do 1M 15 seconds. An FX couldn't do that on LN2.

A $316 Conroe can blast a $900 FX, and Conroes have been overclocking to 3.8+GHz on air.

AMD is doomed, unfortunately, which means Intel will rest on its laurels again.

Everyone here is constantly saying "Oh its an Intel system, built by an Intel team, vs. an AMD system, built by an Intel team... I'll trust the reviews when independant people get them."

If you looked a little you would see, that there are already lots of people with the Conroe in their hands. And it has shattered every PI, 3DMark, world record there is. We are talking about 10s 1M SuperPi runs, and if you know anything about that benchmark you will know that is absolutely crazy. Why not read some forums, like XtremeSystems or more specifically some benchmarking threads where the world record was broken on air w/ Conroe, but now its under LN2 for some other people (including coolaler) and is holding the world record.

Everyone here is constantly saying "Oh its an Intel system, built by an Intel team, vs. an AMD system, built by an Intel team... I'll trust the reviews when independant people get them."

If you looked a little you would see, that there are already lots of people with the Conroe in their hands. And it has shattered every PI, 3DMark, world record there is. We are talking about 10s 1M SuperPi runs, and if you know anything about that benchmark you will know that is absolutely crazy. Why not read some forums, like XtremeSystems or more specifically some benchmarking threads where the world record was broken on air w/ Conroe, but now its under LN2 for some other people (including coolaler) and is holding the world record.

Why don't you AMD fanboys shut the fuck up and look at some inofficial numbers?

http://www.xtremesystems.org/forums/forumdisplay.p hp?f=59

There's a start where you could actually look for yourself instead of covering your ears and shouting "LALALA, Intel must be faking it". Stupid moron.

How about this then? http://www.xtremesystems.org/forums/forumdisplay.p hp?f=59

I trust enthusiastic overclockers with engineering samples to be pretty genuine. Conroe is breaking records left and right right now.

Unless you're using something with a lower boiling point. (perhaps flourinert fc-72, at 56 deg C? http://www.xtremesystems.org/forums/showthread.php ?t=42878)
And, my laptop CPU can get close to water's boiling point, if the thermometer chip is to be believed.

I'm happy it's al least processor upgradable. See here. Upgrading the processor would speed the internal graphics as well. Not a bad little bugger.

The Mini was definitely the most important part of the announcement. I guess Mac-heads aren't used to thinking like this, but remember - Intel releases new chips all the time, and sells them directly to consumers. The Mini and iMac are both socketed, so whether you think a Core Solo or 1.67Ghz Core Duo is a big deal or not in a Mini, XtremeSystems has already upgraded theirs to 2.16Ghz. An upgradeable CPU that doesn't require some third-party solution a year or two later? Not a big deal to PC users, but for the Mac, its the first in a hopefully long line of greater flexibility in Mac hardware.