PCMark Vanatage WR is now 29272

[Computurd]

holy crap tilt that is nice

and no can say its a cache run, there is no cache!

[GullLars]

Wow, nice numbers tilt.

I just have to ask, are you running it as a integrated raid-0 or as pass-through software RAID?
I see according to CDM 3.0 you get 120K IOPS read, but AS SSD gives you 175K.

CDM 3.0 4KB QD32 test is single-thread, while AS SSD 4KB QD64 is 64 threads. You may have run into a CPU limit at 120K IOPS in CDM that AS SSD doesn't have.
8x25-M at 30-40K IOPS each should get you 240-320K IOPS, but you maxed out at 175K.
On the other hand, 175K IOPS @ QD 64 = 2734 IOPS pr QD.

Would you run a couple of quick IOmeter setup if i provide you with the config? (or you can make it yourself)
First will be 1 worker, 4KB random read, 4KB alligned, 1GB testfile (since you don't have cache), queue depth at exponential stepping 1-256 with 2^n stepping, 1 sec ramp time, 10-15 sec run time (since you don't have cache). It will total 9 runs and take ca 2-3 minutes.
Second will be #CPU cores (logic) workers, same access spec and test size/lenght, but QD exponential stepping 1 - {256/#workers} (say 8 "cores" makes 8 workers, each with QD 1-32, giving 6-7 runs).
Results should be fairly accurate and paint a picture of IOPS scaling. The reason for QD 1-256 is to give all SSDs a QD of 32 at the end (wich is max supported by SATA spec).

[Tiltevros]

Wow, nice numbers tilt.

I just have to ask, are you running it as a integrated raid-0 or as pass-through software RAID?
I see according to CDM 3.0 you get 120K IOPS read, but AS SSD gives you 175K.

CDM 3.0 4KB QD32 test is single-thread, while AS SSD 4KB QD64 is 64 threads. You may have run into a CPU limit at 120K IOPS in CDM that AS SSD doesn't have.
8x25-M at 30-40K IOPS each should get you 240-320K IOPS, but you maxed out at 175K.
On the other hand, 175K IOPS @ QD 64 = 2734 IOPS pr QD.

Would you run a couple of quick IOmeter setup if i provide you with the config? (or you can make it yourself)
First will be 1 worker, 4KB random read, 4KB alligned, 1GB testfile (since you don't have cache), queue depth at exponential stepping 1-256 with 2^n stepping, 1 sec ramp time, 10-15 sec run time (since you don't have cache). It will total 9 runs and take ca 2-3 minutes.
Second will be #CPU cores (logic) workers, same access spec and test size/lenght, but QD exponential stepping 1 - {256/#workers} (say 8 "cores" makes 8 workers, each with QD 1-32, giving 6-7 runs).
Results should be fairly accurate and paint a picture of IOPS scaling. The reason for QD 1-256 is to give all SSDs a QD of 32 at the end (wich is max supported by SATA spec).

make the file and then give it to me to run it and i will post the results and then fix them to see the specs.
Now im running on IR raid 0 64 stipe size i cant change it
im using only 1 worker.
im waiting for ur test file to run it if u promise me that u going to fix the xls file and post the results here

[GullLars]

Sure, i'll make an xls file from your raw data (.csv), and make a graph if you want.
Here's the IOmeter configs. One with 1 worker, and one with 8 workers.

BTW, i'm impressed, i thought the max IOPS from integrated RAID on the 9200-series was around 80-90K, but you've shown 175K in AS SSD. Still, that's only 22K IOPS pr x25-M, and a single x25-M can do 35-40K IOPS (4KB random read).

[Tiltevros]

opened and run nothing else touched :/

[GullLars]

So you get max 110K IOPS with 1 worker, but with 8 workers you get 198.931 IOPS, sweet.
I'll make the graphs later today, just skimmed the sheets now. I've got stuff i need to tend to :P

BTW, while i'm gone, you could try just cloning the 8 workers to 16 workers whitout changing anything else and see if you pass 200K IOPS, that would be nice

What score do you get in PCmark vantage HDD with that setup? Do you beat Steve's 90K from his new WR today?

[dj_ski69]

As far as I'm concerned, ne1 who hides their info does not deserve the spot they hold... Its called being a PUNK.... Futuremark should remove his score and listing for hiding is score. It just ain't right to claim world #1 and then hide the goods.... Thats like claiming that you have the fastest car in the world then telling folks when asked that oh yeah "It was so fast you couldn't even see it" Its ing invisible..... Gimme a ing break. Be a man and show your system details. Quit hiding for whatever reason your doing so. Your no better than Corsair if you choose to continue hiding.

[mike]

As far as I'm concerned, ne1 who hides their info does not deserve the spot they hold... Its called being a PUNK.... Futuremark should remove his score and listing for hiding is score. It just ain't right to claim world #1 and then hide the goods.... Thats like claiming that you have the fastest car in the world then telling folks when asked that oh yeah "It was so fast you couldn't even see it" Its ing invisible..... Gimme a ing break. Be a man and show your system details. Quit hiding for whatever reason your doing so. Your no better than Corsair if you choose to continue hiding.

LOL... chill buddy chill. Stevero has asked me several times what my config is, and I have answered him whenever I got the chance to log in and respond to my inbox...

My answer to him a couple of days ago - was the same as I have answered several other people who have been asking me, still running on my almost 4 year old RETAIL Areca 1231ML ( first GEN, not 2nd GEN like Stevro) with 2GB of cache and ONLY 2 ( TWO) ANS9010. And those 2 drives have been somewhat properly configured....Don't beat me up for spending the time to learn my controler and make sure I get the most out of it. And NO - not sponsored, paid for by MYSELF...

it's obvious that Stevo has been gunning for the top score for a while now. WTF is wrong with me wathcing him silently. YOU EXPECT ME TO CONFIGURE HIS CLEARLY SUPERIOR RIG SO HE CAN BEAT ME?????
WTF?
He's been getting plenty of info to figure out what he has been doing wrong. DO I REALLY NEED TO DO THAT???

Also NOT MY fault that FM can not properly detect my HW.

GET A LIFE OR OFF

[mike]

But Mike is not saying what storage he is using Classic

You have no idea how many times I have been asked privately in a "friendly" manner of what hardware I use, and how many times that "friendly question" turned out to be used against me to copy my configuration and beat me....I enjoy being able to figure out myself what the best possible configuration to build the fastest PC overall. I understand tho that some people prefer to just copy a proven solution and it can be helpful information.

[mike]

Mike giving a rundown of his would be amazing clearly he knows how to tweak!

Either of you not wanting to share your secrets is understandable too Just thought I would ask.

Give me some time on that - I have been working on a guide for the pcmarks, not quiet ready yet

[SteveRo]

Give me some time on that - I have been working on a guide for the pcmarks, not quiet ready yet

That will be a bookmark for sure!

Also - much thanks to Mike and to everyone here for all that I have learned in the past year.

Mike, Planet, and most other folks on this forum have much more oc experience than I, it is a pleasure to learn from you all.

[mbreslin]

To the person who had a go at mike for not giving his specs/secrets: If we knew *exactly* how to get the most out of pcmv, then it's just who can throw more money at it. I think 'whoever has the most money to burn wins' is horribly boring. I've enjoyed learning about my hardware and any tips for pcmv would be great. Just being handed the answer would make the whole process uninteresting for me and I'd just stop.

Edit: 'whoever has the most money to burn wins' or 'whoever has room for a SS and a wife who doesn't care' *cough* steve *cough*

[Expat GriZ]

Edit: or 'whoever has room for a SS and a wife who doesn't care' *cough* steve *cough*

I'm pretty lucky as well. My missus has let me keep a Giant PURPLE Cascade in the living room for now....

[Yellowbeard]

I'm pretty lucky as well. My missus has let me keep a Giant PURPLE Cascade in the living room for now....

I bet that is an interesting conversation piece when friends or family visit

[Expat GriZ]

Yes...it HAS come up in conversation. But the garage is stuffed to the rafters, so until I get my shed build (this summer...please!!) while I'm at work it's hiding behind 2 palm trees in a corner. She has the patience of a SAINT!!

[One_Hertz]

You are still above him in my books. 138k is very obviously non legit/cached/whatever (non sustainable), especially considering it can not be ran on dynamic raid.

The 1GB++/s bandwidth (if legit) means the only possibility is an LSI card and none of them are capable of that score.

So I was right.

Congrats on figuring out how to run the vantage bench in cache I guess.

[mbreslin]

I never did like Expat GriZ.

[Computurd]

maybe sell me a few of those c300 for half price, to me of course, and get enough for phase mbreslin?

[SteveRo]

@Expat - that is a beautifull machine!
@mbreslin - yes - my wife is a saint also.
By the way - my SS was much less than you might think - contact xtremesystems member - sdumper for details.

[mike]

So I was right.

Congrats on figuring out how to run the vantage bench in cache I guess.

What's the point that you are trying to make here? That you are right?



When I had spoken to a friend of mine ( who is a very active member in the storage forum ) before posting results, I had asked, should I beat Stevero in the manner he is running, or should I go all out and cache the FIRST subtest ontop of it?
His response was - go all out, and show how it is done.

Sorry if I displeased you by doing so.

From you negative undertone in your comments, it appears that you disagree with me using my RAID controler to it's max capabilities. Why?

My primary use form my Areca is video editing. I need to use all the system memory that is available to me. I cannot afford to create a cache drive based on system memory. A raid controler with onboard cache, is a solution with real world benefits and advantages. Want me to cripple my rig even more as it already is???

People have been using software based cache drives in PCMarks for a while now. Some intentionally, others accidently. I can clearly see the real world benefits from running software based cache drives on your system memory, but for me it is not an option, and I prefer to use a DEDICATED hardware solution for my and this purpose.

Do you also complain when people are using Intel's onboard raid controler, or some of Adaptec's controler where the driver smartly creates a cache drive using your system memory? OR some clowns who ontop create multiple small arrays, software stripe these again, thus creating a BIG software based ramdrive with the sole purpose of gettting a top score in PCMarks?



If you don't like what I am doing just ignore me. But PLEASE don't be-little me.

Afaik I still have the highest HDD suite overall score without any RAID controler caching involved, to appease even folks with your viewpoint.

[SteveRo]

Week before last I figured out how to cache the first subtest - it is in my latest results also

[One_Hertz]

snip

If you like to benchmark the speed of your cache then go ahead. Those scores don't have much to do with the actual speed of the array, but if that is fine with you then congrats. To each their own. Will take your advice and ignore you from now on.

And yes, I would say the same to anyone a using software ramdisk or even to the ICH10R results where it shows 1200mb/s for one of the tests when caching is turned on (media test I think).

[mike]

If you like to benchmark the speed of your cache then go ahead. Those scores don't have much to do with the actual speed of the array..

It doesn't???? Really? Sorry had no idea

http://www.xtremesystems.org/forums/...7&postcount=69

SO all the snappy performance that I thought I had all those years while using my Areca 1231ML's array in real world useage where just in my mind - and not REAL???

Damn I must be easy to fool

congrats. To each their own.

Save you snippy grats - not interested in them...

Agreed to each their own.

[NapalmV5]

sorry mike but your 1231 is total crap!!

If you insist on spamming one application, try making a "multi-thread" batch setup, make one batch launching 10 other, and then have those 10 launch N instances of the program each, like 30 for a total of 300. Should be interresting to see if the QD skyrockets.
I stilll suspect the ML1231 reads every FF instance afte the first 2-10 from cache, and thereby get cache speeds instead of disk speed, so you'd basically be benching the RAID controllers caching performance.

and according to this statement its no better @ real apps

mike your 1231 is just a stinkin pos!!

[mbreslin]

Makes me want to get a 1231 :x

Edit: Also, since mike and steve both cheated, and planet worked his ln2 magic for corsair, I'm claiming first place.. (nobody tell nizzen)!