http://img690.imageshack.us/img690/3708/capturezkw.jpg
Can't say I am all that impressed w/ it so far.
Core i7 @ 4.2GHz
6GB @ 1200MHz
X58A-UD7
9800GX2 :p:
Printable View
http://img690.imageshack.us/img690/3708/capturezkw.jpg
Can't say I am all that impressed w/ it so far.
Core i7 @ 4.2GHz
6GB @ 1200MHz
X58A-UD7
9800GX2 :p:
dang...
hey lowfat lets see some other number, like a 4k sequential run. also lets see some other file size randoms and sequentials. you shouldn't get too much of a pinhole view on 4k. in gamin usage 64k is a more important file size for instance.
Low could you run PCMark05: Windows Start Up, HDD General Use and Anti Virus Scan tests?
http://service.futuremark.com/compare?pcm05=2161534Quote:
Originally Posted by Jor3lBR
I just have the free version so here is a compare link.
Quote:
Originally Posted by lowfat
Thanks man! Did you install PCMark05 on to the Fusion IO?
4k sequential read, 64 queue depth.Quote:
Originally Posted by Computurd
http://img704.imageshack.us/img704/6349/captureja.jpg
YesQuote:
Originally Posted by Jor3lBR
Well that's unfortunate. Especially on the gaming benchies.Quote:
Originally Posted by lowfat
4k sequential write, 64 queue depth
http://img690.imageshack.us/img690/7572/captureui.jpg
Agreed.Quote:
Originally Posted by One_Hertz
I guess that to acquire more PCMark05 points I will need to get the outdated Acards :shakes:
God I was so excited even ordered one but now have to return it.
Probably it's still too soon and lowfat will need more time to put in results, kudos for him for being so kind with us :clap:
I wonder why your result was so different from these on Vantage and PCMark HDD Suits:
http://hothardware.com/articleimages...e-vantage1.png
http://hothardware.com/articleimages...e-vantage2.png
I'll give Vantage a go now :shrug:Quote:
Originally Posted by Jor3lBR
Windows Defender 210.34 MB/s
gaming 155.19 MB/s
importing pictures to Windows Photo Gallery 160.4 MB/s
Windows Vista startup 199.39 MB/s
video editing using Windows Movie Maker 111.6 MB/s
Windows Media Center 122.6 MB/s
adding music to Windows Media Player 122.36 MB/s
application loading 145.38 MB/s
http://smiliesftw.com/x/augentreher.gif
http://img69.imageshack.us/img69/342/capturevki.jpg
EDIT: Something is wrong here. I have PCMark installed to the ioXtreme but all the benches are from the X25-M. I just ran the Vantage test on my X25-M and they were damn near identical.
vantage will only run on your boot drive.
Then I wonder how hothardware was able to do it w/ their ioXtreme.Quote:
Originally Posted by Computurd
http://hothardware.com/Articles/Fusi...Review/?page=7
You can select which drive to use for the HDD tests from Options in Vantage.
I've used that option a lot of times.
Now, what about the CrystalDiskMark 3.0 beta run :)
I guess that would have been the smart place to look :p:Quote:
Originally Posted by Anvil
I'll do another run and the CDM 3.0 soon.
I remember reading a comment David Flynn left on an Intel forum discussing the 1million IOPS setup they had.
His comment went like this:
Our drive does use some system memory and CPU cycles in order to streamline the I/Os and allows programs to access the physical data locations instead of LBA's by keeping a record of the LBA-to-physical mapping in system memory. This allows us to deliver higher I/O rates at lower latency overhead. This method adds only a few ĩs latency overhead to the physical NAND latency, and allow our MLC products to deliver accesstime lower than competitors SLC producs, and our SLC products unmatched.
This would also help explain why RAM usage is reverse proportional to Block Size like you quote. It's also true they use MFT for random writes to deliver better random r/w ratios than most competitors.
I expect next generation ioMemory (the architecture, not boards) to have an onboard decent sized non-volatile write buffer wich performs MFT-like actions of combining full erase-blocks in memory, thereby delivering RAM-like random write accesstimes and random write throughput close to sequential. This may allow RAM-readback, non-blocking writes, and if used with the same system memory LBA-to-physical mapping still deliver the same random read rates.
I can post IOmeter setups if you want, so you only have to edit the QD between each run. I thought you guys knew enough about IOmeter to make the config from the settings I mentioned, but i can make a custom one tailored for ioXtreme where you only have to change the 1 parameter.
BTW, that crystal 3.0 bench was unimpressive. Still, you used the 100MB lenght despite everybody yelling 1000MB or 2000MB lenght.
One single x25-M can do 160MB/s 4KB random QD32. Anvil has gotten >450MB/s with 3 x25-M from ICH10R. (103K 4KB random read IOPS in IOmeter)
Can you do some IOMeter runs that are identical to the HotHardware review? Just to see if your numbers match theirs or not. It seems to be too slow...
Try to run on a different memory divider too. Higher mem speeds can result on a boost in performance :shrug:
I just tried there 8k, 80% read, 80% random test. At the beginning of the test results were the same, but by the end of the 2 minutes the results had dropped in half. :rolleyes:Quote:
Originally Posted by One_Hertz
oh so RAM is boosting the figures near the start and the real numbers are 50% slower? If that is true, then it is a pretty big fail on the part of the reviewers of hothardware to report this. Have you ran vantage properly yet?Quote:
Originally Posted by lowfat
http://service.futuremark.com/result...&resultType=18Quote:
Originally Posted by One_Hertz
Still consistently slower than hothardware.
This is just upsetting. I wonder what the difference comes from and how they got double your result in some categories.
edit: I wonder if you are in a used state from all the IOMeter runs. Is there some sort of secure erase option?
I'm not sure that is the problem. Here is a 4k write. 100% random result. 64 queue depth. Stayed consistent throughout the entire test.Quote:
Originally Posted by One_Hertz
http://img263.imageshack.us/img263/7...andomwrite.jpg
Thought about this. I did a 'Low Level Format' with the Fusion-io utility it comes with.Quote:
Originally Posted by One_Hertz
I have written over 1.2TB to the drive in the past 2 days though. :p:
EDIT:
Another weird thing. How it is possible that it can do over twice as many 4k 100% random write IOPS than read.
This is looking better - with 1231ML-2G 8xacardR0 best I can get at Q64 is approx 48K @ 4KB 100@ random 100% write.Quote:
Originally Posted by lowfat
lowfat! how fast does L4D loadup ?
I don't have L4D :shrug:Quote:
Originally Posted by NapalmV5
Get it "for testing purposes only" :D
come on lowfat i didnt have it either.. i acquired it/loaded it up/posted results
what about other games ? whats your os setup on ?
^^ He did post some game results here Napalm.Quote:
Originally Posted by lowfat
Windows 7.Quote:
Originally Posted by NapalmV5
I don't have a lot of games.
Crysis, Cryis: Warhead, UT3, HL2, SupCom, SoaSE, The Witcher, Far Cry 2 :shrug:
I don't 'aquire' games.
Quote:
Originally Posted by Jor3lBR
apart from the 3gb copy/paste the other results seem to be loaded from x25mQuote:
Originally Posted by lowfat
come on guys x25m is fast but not ioxtreme fast
so when you buy you dont acquire ? :shrug:
did u check for any updates like firm and drivers??
The games are installed to the ioXtreme. I can't tell you why they aren't any faster than the X25-M :shrug:Quote:
Originally Posted by NapalmV5
ioxtreme operating @ x4 pcie ?
look in other words you should be faster then hothardwares results given your system/win7/etc
Don't think so. The CDM and Iometer resuts are pretty much where they should be.Quote:
Originally Posted by NapalmV5
http://kb.fusionio.com/KB/a33/driver...ns-lineup.aspx
Article ID 33
Created On 15/9/2009
Modified 4/12/2009
Firmware versions 1.2.6 and 1.2.7
36867 (recommended)
24160
17350
Tuning Techniques for Writes
Fusion-io is the only company that enables you to format solid state storage for optimal write performance. This section discusses performance tuning techniques, as well as considerations for writes.
Increased Steady-State Write Performance with fio-format
Under sustained, random writes, which are written randomly across the entire ioDrive, write performance will decrease. This decreased write performance is referred to as steady-state write performance.
There are very few real-world storage use cases that create this sort of write workload, the most common being RDBMS systems. However, the most common activity that triggers this behavior is running micro benchmarks rather than real-world applications.
Unlike drive form factor devices, the ioDrive can be tuned to achieve a higher steady-state write performance than what it is shipped with from the factory. The first line of tuning is to provide the ioDrive with additional working capacity, or reserve (storage capacity that is not visible but is usable by the ioDrive for internal optimization). Changing this reserve capacity can be done using the fio-format utility. Running fio-format changes the reserve size and affects the overall long term performance of the ioDrive. Repeated fio-formats are not required to get proper performance, unless the initial change in reserve size did not yield the needed steady-state write performance.
A recommended starting point is decreasing the drive's capacity by 25%. For example, if the drive is a 160 GB drive, you would fio-format it to 120GB
Warning: This will destroy all data on the drive!
Assuming the drive that you would like to format is /dev/fct0 (use fio-status to determine this), you would use the following commands:
$ fio-detach /dev/fct0
$ fio-format -s 120G /dev/fct0
$ fio-attach /dev/fct0
The following graph shows the effect of reserve on write performance under some circumstances. This graph shows that the larger the reserve, the better the steady-state performance achieved.
http://kb.fusionio.com/KB/a51/tuning...or-writes.aspx
That is for the ioDrive. Not the ioXtreme. The ioXtreme has no newer firmware/drivers available.Quote:
Originally Posted by Tiltevros
This is also for the ioDrive, not the ioXtreme. I don't have any choices when formatting the ioXtreme.Quote:
Originally Posted by Tiltevros
http://hothardware.com/Articles/Fusi...Review/?page=1
at page 12 is No Mercy or Dead Air levels
these are weird results lowfat, i would suggest that you keep plugging away, keep testing different configurations, make all changes that you can to find that sweet spot. you are in uncharted waters here pretty much but perseverance is the key! i see the upside down nature of the read/write benches and that really makes me feel that something is going on here....you will find it! if it writes that fast dammit it should read like a mother :banana::banana::banana::banana::banana::banana:!
I get 53k (208MB/s.)Quote:
Originally Posted by SteveRo
But why are 4k-random-read resultes so bad? (only 35k)
Increasing PCIe increase maximum bandwidth?
http://smiliesftw.com/x/eek3run.gif
PCIe bus @ 100MHz
http://img710.imageshack.us/img710/976/capturenk.jpg
PCIe bus @ 110MHz
http://img412.imageshack.us/img412/8968/capturese.jpg
The CPU is underclocked (2.4GHz) currently as I am trying to find my max bclk. But the sequential reads are considerably higher.
you might check to see if your motherboard allows the pcie packet sizes to be modified as well. the higher the packet setting the better. i have achieved great results with that one:up:
@ 115MHz for the PCIe bus I am now getting 876MB/s reads and 315MB/s writes http://smiliesftw.com/x/eek3.gif
However my sound card isn't working. :down:
How come the 4k scores took a dump when you clocked PCIe up?
@ low
Would you buy this card again?
I think the answer is obvious...Quote:
Originally Posted by F.E.A.R.
Well for the price of this card you can get 2 x Intel x25-m g2 + Acard 1231ml + 2Gb DDR2 for cache.
I ordered one of these thru Amazon but canceled the order after I saw results, but I guess this will change once Fusion hear some feedback and decide to release some improved firmware.
Well I don't know what is holding back actual real world speeds. Bandwidth and random IOPs are both pretty damn high. It may be that we are at the point that we need a dedicated processor to handle all storage devices. I don't think the chipset that the ioXtreme does all that much. If you take a look at my CPU usage in my Iometer runs you'll see it is at like 40%.Quote:
Originally Posted by F.E.A.R.
Hopefully I'll see more of an increase in performance once I am back on phase and up to 4.6-4.8GHz.
Now to answer your question. With the current results, no way would I have bought it. I would have stuck w/ my X25-M.
Thanks a lot for getting it before anyone else and testing it. I personally appreciate it very much.Quote:
Originally Posted by lowfat
that is strange that it takes such a large percentage of the cpu cycles, forty percent is outrageous. using the host bus adapter i have been playing with the highest i have seen is 9 percent, and that is using it as a dynamic disk with eight devices in software raid, so you think that the overhead would be even higher than that of the i/o extreme. strange. i hope that further firmwares etc will fix that. or maybe it was rushed to market? also maybe you have a borked unit? i would at least give a shot out to their tech support to touch base and see if that is in line with the expected performance of the unit. it does suck to test a new device, especially when it is expensive, and have it crap out on ya. i got totally bonered with the highpoint 4320 by god thats the biggest peice of :banana::banana::banana::banana: i have ever seen. i would say thanks for testing it, but i have a feeling you aren't done yet, something isnt adding up here, the upside down read and write issues along with the high cpu usage gives me this feeling that something odd is going on here! your efforts are appreciated though! i look forward to comparing its performance with other devices, etc. good luck getting the bugs ironed out.:up:
Kind of a shot in the dark, but do you have another Mobo you can test it on?
I have a :banana::banana::banana::banana:ty Asrock X58 Extreme, but it will have to be on stock clocks as the board won't startup anytime you need to change something in the bios. :rofl:Quote:
Originally Posted by Griff805
Then do it ;) Or try another PCI-e slot.
I am not surprised that is uses higher CPU% than say X25s on a dedicated RAID card, since as noted several times in the thread it does need the CPU for MFT like operations. I didn't think it would need 2 Nehalem cores however :D Where does that leave the actual processing of data?:)
I just hope that the really high CPU usage will not prove to be a "feature" of these PCI-e SSD cards.
That would seriously suck.
Hopefully the answer will reveal itself to be some kind of incompatibility / weirdness....
I don't recall IoDrive tests, but that would tell - if it's not high on IoDrive then it has dedicated RAM and IOP to do what IoXtreme seems to do on CPU itself.
@ Low
Look 4k random r/w :eek:
http://www.micronblogs.com/
Thats a real ssd :D
haha
I want that IO-monster :D
Pretty sad that thing can beat the ioXtreme for random read IOPS. However writes are still considerably slower.Quote:
Originally Posted by F.E.A.R.
Nice link. Looks like the C300 could well be the next ssd I buy.Quote:
Originally Posted by F.E.A.R.
Not so sure about the Marvel host adaptor though. I wonder why they used that for pc mark benchmarks. :shrug:
.132 on access times for that ssd? you would think it would be much better than that.
Well, you'd think that eventually in even a lower end company they might somehow come out with at least 1 item that is above average. Never know- Maybe one of the engineer's spent a little overtime working on it?Quote:
Originally Posted by audienceofone
I'm guessing that is the marvel controller more than the SSD :shrug: Either way the drive looks promising considering it is at pre-release stage.Quote:
Originally Posted by Computurd
Anyone know how much this drive will retail for?
"AS SSD Benchmark" isnīt a exactly benchmark.Quote:
Originally Posted by Computurd
I prefer IOmeter. Look the video from 07/Dec.
http://www.tomshardware.com/reviews/...-ssd,2488.html
Another review. Takes a huge performance hit in a used state. Thankfully it is really easy to do a low-level format on this thing.
after reading that toms review it is pretty clear they are underwhwelmed by it as well.
A well-known problem ;)Quote:
Originally Posted by lowfat
Look at the date of this review... http://forum.ssdworld.ch/viewtopic.p...54308f9fdf77a4
But you can try to clean the card without low-level-format.
FreeSpaceCleaner cleans a drive with "0" or "1" (FF) http://alex-is.de/PHP/fusion/downloa...download_id=10
That's a huge hit... i professional use you fill it up with writes in a day... what then, program a reformat on each reboot? stupid.
do you still have the device? i would be curious to run some benchmarks on it to measure the in game performance. that would probably be superb with this device.
I am using a 9800GX2 which really isn't the videocard I should be using for a test like this.Quote:
Originally Posted by Computurd
Did you try to format using the Maximum Write Performance setting? What does it actually do in terms of capacity?
If you did try it, did you try getting the IoXtreme in "surely used" state? (that is, writing several times its REAL capacity over, otherwise, the used state comes at a time when there is still free flash area).
Another test I would be very curious to see would be the performance differences between the X58 and P55 chipsets. Now that the PCIe clock is on the CPU for the P55, I heard the latencies are lower.
From this...I wonder if this would have any difference in terms of performance at least synthetic benchmark wise? The only problem with the P55 chipset would be the limited available lanes (16).
i think that if you puyt a raid card on a p55 you will be suprised with how well it doesnt perform. not sure but with less lanes available i would have a hard time seeing it work well. is the device x8? i know that there have been conclusive tests showing that the x8 restriction on the p55 does in fact hamper performance on video cards, even if not in sli.
This does not exist on the ioXtreme. You can only choose to do this w/ the SLC versions of the Fusion-io's.Quote:
Originally Posted by alfaunits
I was referring to his ioxtreme...which is PCIe 4x. That should be no problem. As for a raid card, you may be right, but I haven't tried it. A P55 should allow one PCIe 2.0 16x or two PCIe 2.0 8x. I doubt there is a raid card out there than saturate that.Quote:
Originally Posted by Computurd
TH was able to make a format with different size on IoXtreme as well.Quote:
Originally Posted by lowfat
http://www.tomshardware.com/reviews/...sd,2488-9.html
Maybe they just used Fio-format to specify a smaller size instead (i.e. there's no predefined setting).
Anyway, I found on FusionIo forum. The Improved Write Performance capacity is 50% of the max. and Maximum Write Performane is 25% of the max.! (i.e. if you format a 80GB IoDrive/IoXtreme for Improved Write Performance it will only provide 40GB usable space!)
I'm curious what copy speeds you guys can achieve between RAID arrays on the same system? Or between a RAID array and IoXtreme.Quote:
Originally Posted by Spoiler
I currently have 1680 and 1280ML in the system - 1680 runs 3xX25-M and 1280 runs 4xX25-E. Obviously copying TO X25-M array won't go over 300MB/s because the drives are limited, but the copy from X25-Ms to X25-Es should yield something like.. 500MB/s - well I ain't getting far over 400MB/s either way.
So, please run a simple TC copy (using Big file copy option).
Problem is TRIM isnīt mature.Quote:
Originally Posted by alfaunits
TRIM with Indilinx-SSDs works fine (only with Beta-FWs)
There's no TRIM. You prolly mean Garbage collector.
From THGQuote:
Originally Posted by alfaunits
Quote:
I/O performance is one of the most important factors when purchasing an SSD. Obviously, the SLC-based ioDrive remains by far the fastest solution, but the new ioXtreme still beats Intel’s latest X25-M G2 with the TRIM feature enabled. The Improved Write Performance setting offers the highest peak I/O performance, while the Maximum Write Performance run secures a higher minimum performance level.
They did that w/ the ioXtreme Pro, not the ioXtreme.Quote:
Originally Posted by alfaunits
Ye, X25-M had the TRIM, IoDrive didn't (and does not support it on Windows).Quote:
Originally Posted by F.E.A.R.
Can anybody run PC Mark 05 HDD tests?:D;);)
Well I think my ioXtreme died today. Reboot my machine and it failed to mount. Tried a different PCIe slot same thing.
Getting an 'Attach Failed (internal failure) :wtc:
I guess I am about to see how well Fusion-io's support is.
:(:( have u tried another system?
I am about to. Just transferring a few TB of data at the time. Just waiting for it to finish.Quote:
Originally Posted by NapalmV5
you got it like 3/3+ months ago ?
sob! 3 month/1 year warranty ??
I got it 2-3 months ago, but it does have three years of warranty. Well I was able to reformat the drive on another machine and able to get it working. But I lost my Steam folder and a bunch of other stuff. :down: Guess I'll be keeping a backup of the drive for now on.Quote:
Originally Posted by NapalmV5
have you noticed any change in the amount of available storage space? that might be an indicator of some type of nand failure. i would rma it just for the hell of it. just to be safe. dunno if they would allow you to rma though considering you have gotten it to work.
Nope. The ioXtreme also does integrity checks every time the computer isn't shut down properly. So I know all the cells are in working order.Quote:
Originally Posted by Computurd
Sorry to hear that but at least it is covered by warranty. Do you think it might have had anything to do with uping the voltage?
I never upped the voltage. Just the PCIe bus. And the data corruption happened when it was @ 100Mhz.Quote:
Originally Posted by audienceofone
Hey Lowfat how is the IoXtreme behaving now? I was just looking at the WEI scores and the IoXtreme wipes the floor with random 16 reads. Has there been an update on boot ability?
Card has been working fine. Besides that one issue of the drive becoming inaccessible, no issues. Still nothing on being able to boot, although I seriously doubt this will ever get fixed.