If the good guy gets the girl, it's rated PG; if the bad guy gets the girl, it's rated R; and if everybody gets the girl, it's rated X - Kirk Douglas
INTEL HAS DENIED that its solid state drives (SSD) start getting slower as they get older.
A review, with the catchy headline 'Long-term performance analysis of Intel Mainstream SSDs' over at PC Perspectives claimed that the Intel X25-M drive degrades in performance as a result of "internal fragmentation" and that "a 'used' X25-M will always perform worse than a 'new' one".
However Chipzilla has denied the alleagation saying that its labs have been unable to duplicate the results.
It thinks the synthetic workloads they used to stress the drive were not reflective of what happens in real life. After all nothing gets slower as it gets older. µ
L'Inq
CNET
Maybe someone should also ask them about using DDR3 memory with XMP profiles on Core i7 platform. Some kits (for example Kingston HiperX 1866) increase VTT/QPI voltage (basicly uncore/memory controller voltage) to very high levels - from standard 1,2V to 1,5 or even 1,7V.
Considering the fact that their own recommendation for VTT/QPI max voltage is 1,35V, i really wonder how long will those Core i7 rigs kited with high-perf XMP DDR3 memory last.
Even OCZ states that modern synthetic testing methods are not applicable to Solid State Drives. As well, due to the necessity to ALLOW fragmentation across the entire 'disk' in order to properly meet design life expectations.
You want a proper test? Create an app that can counter act the built-in "fragmenter" and write the data a few thousand times to a single location and continue to do that repeatedly. THEN you would have a benchmark that would be applicable for measuring the solid state storage chip performance over time.
Beren, the CPUs you are wondering about are spec'd to work with 800 and 1066 DDR3; if you are plugging in 1866, then you are on your own.
You have a very good question... maybe someone should ask that question to the people that are SELLING THE DDR3 PART as they are the ones you are manufacturing the part. If you plug in a high performance part in your engine that might screw up the engine - do you expect the engine manufacturer to do that testing or the people who make the high performance part? Do you expect the engine manufacturer to test all high performance parts even if they are outside the spec, or do you expect that testing to be done by the people making the high performance part?
While I realize people like to bash Intel... there is no god given right to be able to plug in any part or operate a chip outside of the spec and expect the CPU manufacturer to do all the testing for them. They sell a part - they spec an operating window. Go outside that window... and you get what you get.
I'll be sticking with tried-and-true hard drives, you know - the things we've been using for the last .. oh 53 years
Did I mention they're cheaper too and have much more capacity?
It will always be possible to design a stress test that will exceed design capacity of things like SSDs. The right thing is to measure against a real workload or model thereof, with peaks and averages in work rate. The SSD has to offer a balance of competitive performance and perf/$$ with a good expected warranteed lifetime. With devices like flash - where some "ageing" of the media is (today) unavoidable, life extending techniques are used. These can be temporarily triggered by synthetic workloads reaching many, many times the demand of a real workload. The moral: use benchmarks that are "real world" meaningful.
The file system's software organization of the drive (into 512 byte sectors) doesn't match the physical hardware organization which may be 256K chunks and even worse is not aligned. So when a drive is new and you write to it, the drive can combine the writes and puts them in one physical sector or at most span hardware sectors once. But what happens when you erase part of it and then write more data back? Fragmentation again. It's not a huge issue on reads because the seek time is so low but it's a total mess on writes.
Here's a brand new article that covers the details and how Vista & Linux can work around it partly.
http://thunk.org/tytso/blog/2009/02/20/aligning-filesystems-to-an-ssds-erase-block-size/
ssd rock. stop the fud. i've been hammering a handful this last week, 10+ gig a day write/erase on "cheap" drives. they rock, speeds aren't diminished, you get whiplash opening ie and the like. i'm having a blast. didn't fujitsu say they were going ssd? scary stuff...
Well i bought one of these crappy rip off drives and I did test for myslef and after each HDtach run i found it went from 220mb/s to 180mbs. damn crap wasnt much faster than the WD 74GB veloci. it showed much better benches, but it did not load apps quicker and this is a joke for the price. i was expecting so much from what ive seen and read.
intel, i know you bullhit and cant wait for someone who can proove this in a proper way.