PDA

View Full Version : someone please help me interpret benchmarks?



Ton
05-05-2001, 03:18 AM
Ok, as said in another thread here on the forums, I'm seriously thinking adding two more Cheetha X15's to my 4 X 15 Raid 0. Cost around $900,- including GD internal cabling and active terminator. Anyone in for an educated guess what the gain (in %) will be?
I'm not willing to spent that kind of $$ for, let's say, an increase of 10% or so, it wouldn't be worth it. Also you would expect better performance with the new 533's mobo but in fact it's the opposite. Anyway I find it hard to predict what the gain will be. Thanks.


Ton's benchmarks 4x X15 Ultra Atto 3 Dual PM G4 533 MP

QuickBench¬ô 1.0.1 Test Results File
Test Volume name: RAID 0 72 GIG

Xfer Size Sequential Read Sequential Write Random Read Random Write

1 KByte 1.034 MB/sec 696.136 KB/sec 511.797 KB/sec 742.280 KB/sec
2 KBytes 11.238 MB/sec 1.456 MB/sec 2.158 MB/sec 1.401 MB/sec
4 KBytes 3.171 MB/sec 2.810 MB/sec 3.521 MB/sec 2.903 MB/sec
8 KBytes 39.477 MB/sec 5.596 MB/sec 8.893 MB/sec 5.550 MB/sec
16 KBytes 53.824 MB/sec 11.039 MB/sec 12.949 MB/sec 10.441 MB/sec
32 KBytes 11.940 MB/sec 19.521 MB/sec 38.618 MB/sec 19.421 MB/sec
64 KBytes 37.533 MB/sec 31.609 MB/sec 12.263 MB/sec 14.228 MB/sec
128 KBytes 60.015 MB/sec 50.848 MB/sec 35.384 MB/sec 36.727 MB/sec
256 KBytes 70.171 MB/sec 84.766 MB/sec 37.606 MB/sec 46.754 MB/sec
512 KBytes 100.733 MB/sec 124.511 MB/sec 60.957 MB/sec 71.182 MB/sec
1 MByte 120.925 MB/sec 142.268 MB/sec 72.746 MB/sec 90.017 MB/sec


Macgurus benchmarks: Atto 3 Dual, believed on a PM G4 450 Mhz (yikes?)

4 x X15
QuickBench¬ô 1.0.1 Test Results File
Test Volume name: 4xST318451LW_DCUL3D

Xfer Size Sequential Read Sequential Write Random Read Random Write

1 KByte 1009.693 KB/sec 597.836 KB/sec 350.079 KB/sec 637.186 KB/sec
2 KBytes 5.104 MB/sec 1.263 MB/sec 1.863 MB/sec 1.291 MB/sec
4 KBytes 3.124 MB/sec 2.486 MB/sec 1.886 MB/sec 2.480 MB/sec
8 KBytes 54.709 MB/sec 4.447 MB/sec 6.631 MB/sec 5.171 MB/sec
16 KBytes 13.612 MB/sec 8.537 MB/sec 22.166 MB/sec 8.889 MB/sec
32 KBytes 15.499 MB/sec 15.757 MB/sec 19.894 MB/sec 15.611 MB/sec
64 KBytes 45.551 MB/sec 30.281 MB/sec 17.241 MB/sec 19.500 MB/sec
128 KBytes 81.903 MB/sec 49.120 MB/sec 27.272 MB/sec 22.519 MB/sec
256 KBytes 87.985 MB/sec 80.290 MB/sec 48.075 MB/sec 61.508 MB/sec
512 KBytes 127.188 MB/sec 101.632 MB/sec 60.829 MB/sec 80.580 MB/sec
1 MByte 127.520 MB/sec 134.687 MB/sec 78.572 MB/sec 94.668 MB/sec


6x X15
QuickBench¬ô 1.0.1 Test Results File
Test Volume name: 6xST318451LW_DCUL3D

Xfer Size Sequential Read Sequential Write Random Read Random Write

1 KByte 416.563 KB/sec 560.978 KB/sec 389.864 KB/sec 532.481 KB/sec
2 KBytes 8.182 MB/sec 1.104 MB/sec 1.099 MB/sec 1.082 MB/sec
4 KBytes 5.410 MB/sec 2.244 MB/sec 2.627 MB/sec 2.184 MB/sec
8 KBytes 7.623 MB/sec 4.359 MB/sec 7.143 MB/sec 4.019 MB/sec
16 KBytes 11.992 MB/sec 8.421 MB/sec 6.027 MB/sec 7.815 MB/sec
32 KBytes 9.099 MB/sec 15.218 MB/sec 11.133 MB/sec 14.629 MB/sec
64 KBytes 27.597 MB/sec 27.520 MB/sec 23.153 MB/sec 25.656 MB/sec
128 KBytes 46.656 MB/sec 46.929 MB/sec 31.626 MB/sec 38.397 MB/sec
256 KBytes 90.803 MB/sec 74.341 MB/sec 44.405 MB/sec 61.674 MB/sec
512 KBytes 136.169 MB/sec 103.042 MB/sec 65.180 MB/sec 91.030 MB/sec
1 MByte 157.255 MB/sec 130.056 MB/sec 94.373 MB/sec 113.035 MB/sec

8x X15
QuickBench¬ô 1.0.1 Test Results File
Test Volume name: 8xST318451LW_DCUL3D

Xfer Size Sequential Read Sequential Write Random Read Random Write

1 KByte 437.407 KB/sec 566.283 KB/sec 408.764 KB/sec 584.659 KB/sec
2 KBytes 4.467 MB/sec 1.192 MB/sec 2.868 MB/sec 1.189 MB/sec
4 KBytes 6.189 MB/sec 2.392 MB/sec 5.738 MB/sec 2.451 MB/sec
8 KBytes 6.462 MB/sec 4.621 MB/sec 10.319 MB/sec 5.000 MB/sec
16 KBytes 9.344 MB/sec 8.963 MB/sec 6.283 MB/sec 8.132 MB/sec
32 KBytes 8.737 MB/sec 15.719 MB/sec 35.994 MB/sec 15.396 MB/sec
64 KBytes 12.946 MB/sec 26.791 MB/sec 27.375 MB/sec 13.935 MB/sec
128 KBytes 45.882 MB/sec 48.493 MB/sec 23.990 MB/sec 25.773 MB/sec
256 KBytes 96.399 MB/sec 74.551 MB/sec 43.124 MB/sec 51.179 MB/sec
512 KBytes 143.209 MB/sec 105.959 MB/sec 79.088 MB/sec 94.936 MB/sec
1 MByte 168.098 MB/sec 130.774 MB/sec 123.611 MB/sec 120.135 MB/sec



[This message has been edited by Ton (edited 05 May 2001).]

kaye
05-05-2001, 11:58 AM
Ton,

Yes I see what you mean when comparing your 4x with magician's 4x, except for Sequential Write. Some variables come to my mind. OS version (can make a pretty big difference), SoftRAID version or he may have used ATTO ExpressRAID, the ATTO UL3D firmware version (I remember magician noting way-way back when that the later version was slower, and that test may have been with the slower version), memory throughput (the capabilities of the RAM in both machines), and some Mode Page Parameters (which magician does not mess with). Without having either of the machines, I can't extrapolate what you would get from adding two X15s. k

weenies
05-05-2001, 04:09 PM
Ton,

Some more information: We tested your RAID using the same version of ATTO ExpressRAID and SoftRAID that we included with your shipment. We did change the Mode Page Parameters to either Prepress or Video. This optimized the read and write evaluations for our set-up. If you think that changing these settings will improve the speed on your set-up you can do this through ExpressRAID. You may have even tried this already. We were running OS 9.1. This might have changed something if you are running a different version. Also, I know that the ATTO firmware was the lastest version (.03) I don't know exactly what improvements would come from the upgrade you are talking about, but I know magician will help you with more specifics and information on speed upgrades.

-spud.

Ton
05-06-2001, 01:52 AM
Thanks Kaye & Spud,
My remark about the bench differerences was merely one on the sideline, not a big thing in my view, after all the differences are not that big. Anyway I resetted the mode pages through Atto express tools. I guess this overrides any change made in SR earlier?
Here's some results

Prepress settings:

Xfer Size Sequential Read Sequential Write Random Read Random Write

1 KByte 923.532 KB/sec 606.281 KB/sec 510.830 KB/sec 641.067 KB/sec
2 KBytes 3.551 MB/sec 1.246 MB/sec 1.064 MB/sec 1.250 MB/sec
4 KBytes 3.523 MB/sec 2.464 MB/sec 2.211 MB/sec 2.636 MB/sec
8 KBytes 38.695 MB/sec 4.807 MB/sec 4.352 MB/sec 4.806 MB/sec
16 KBytes 57.785 MB/sec 9.214 MB/sec 7.721 MB/sec 9.317 MB/sec
32 KBytes 9.592 MB/sec 16.810 MB/sec 13.582 MB/sec 17.857 MB/sec
64 KBytes 31.470 MB/sec 26.799 MB/sec 14.148 MB/sec 9.425 MB/sec
128 KBytes 49.690 MB/sec 48.566 MB/sec 27.365 MB/sec 25.914 MB/sec
256 KBytes 78.678 MB/sec 78.520 MB/sec 36.551 MB/sec 44.569 MB/sec
512 KBytes 110.093 MB/sec 118.506 MB/sec 59.861 MB/sec 77.057 MB/sec
1 MByte 126.987 MB/sec 137.140 MB/sec 68.315 MB/sec 95.087 MB/sec

Video settings

Xfer Size Sequential Read Sequential Write Random Read Random Write

1 KByte 822.301 KB/sec 679.763 KB/sec 522.958 KB/sec 716.025 KB/sec
2 KBytes 3.545 MB/sec 1.220 MB/sec 900.779 KB/sec 1.458 MB/sec
4 KBytes 2.862 MB/sec 2.838 MB/sec 2.221 MB/sec 2.847 MB/sec
8 KBytes 12.484 MB/sec 5.628 MB/sec 4.390 MB/sec 5.806 MB/sec
16 KBytes 19.400 MB/sec 10.976 MB/sec 7.766 MB/sec 10.658 MB/sec
32 KBytes 10.127 MB/sec 19.079 MB/sec 13.630 MB/sec 20.069 MB/sec
64 KBytes 30.099 MB/sec 30.915 MB/sec 12.450 MB/sec 22.364 MB/sec
128 KBytes 44.766 MB/sec 54.188 MB/sec 27.193 MB/sec 19.035 MB/sec
256 KBytes 92.227 MB/sec 82.500 MB/sec 33.330 MB/sec 50.650 MB/sec
512 KBytes 127.956 MB/sec 123.940 MB/sec 59.880 MB/sec 79.718 MB/sec
1 MByte 128.225 MB/sec 139.072 MB/sec 66.960 MB/sec 95.282 MB/sec

Btw Kaye, is your 6 x X15 raid finished already?
Hope you post some benchmarks when it is.
Ton

kaye
05-07-2001, 10:48 AM
Ton,

First, I recommend you update to QuickBench 1.5. It has a tab for testing up to 10 MByte size. And yes, ATTO ExpressPro Tools mode page changes do override the SoftRAID mode page parameters. You can view a limited number of the changes by looking at the SR parameters and then click Cancel.

My Burly with 6x X15s is finished. Have to take some pics for the ftp site is all. It is too much for 2x Miles2 in a vintage Mac. Results are the same as with 4x X15s, only MB5 Disk and Pub Disk are faster. I put a general result on the Burly RAID Topic. k

magician
05-08-2001, 02:08 AM
I'm thinking we need to put those pix on the FTP site onto pages, k....I'll throw it out to the tribal wolves on the tribal forum....but I'm also thinking that we need to have a series of pages....a section....where CUSTOMERS can post what they have done with their machines, if only for bragging rights.....that would be kinda cool....sort of like vicariously enabling others to drool over all our own beasts......

yeah. good idea.

ok. I need to act on that.

ok. Ton? The benchmarks on our site are for a single 500Mhz G4 Sawtooth. I forget the OS version, but am pretty sure it was 9.04, with ATTO firmware 1.6.2.

the benchmarks included by us on YOUR RAID are using OS9.1, SoftRAID 2.2.2, and ATTO firmware 1.6.3. I have literally NO idea which machine we used to benchmark your array. I know it ran variously on miss_bam, on banshee, and on berserker (we finally named jorge's old Yosemite)....it may also have been benchmarked on a dual-G4 450MHz Sawtooth we had in here around the same time.

FWIW, X15's just rose $20 in cost, contrary to the trend with all other Seagate drives this week. They are increasingly constrained. I am not sure if new X15 models are on the verge of release, or if this is just a momentary glitch in the market.

Ton
05-08-2001, 03:05 AM
So it appears I have to hurry.
For a performance gain of about 20% consider an order placed http://macgurus.com/infopop/emoticons/icon_biggrin.gif
(I still do not have an answer to my original question)
Ton

magician
05-08-2001, 04:46 PM
what is the original question?

Ton
05-09-2001, 01:34 AM
The original question was posted in another thread, short en simple: what would be the approximately performance gain adding two additional internal X15 cheetah's to an excisting 4x X15 cheetah raid 0 in a new mac.
Thanks Ton

drallen
05-09-2001, 10:54 PM
Ton,

Will your lights dim when you boot up those six X15s? Wow...

Are your drives all internal? What do you boot off of? What is your set up?

Anyways, I wish I could help you with SIX X15s in a new world Mac...

I'm getting the same benchmarks in my machine as yours...

Mine, G4 733, Adaptec dual channel card, Granite cables and terms, and FOUR internal X15s. OS 9.1, SoftRAID 2.2.2, 34.17 Gig striped RAID 0 (plus a boot partition, a mirrored RAID, and a storage partition).

QuickBench 1.5, RAID 0

1MB SR SW RR RW
140 156 78 105

8MB R W
135 152

ATTO 8 MB scores are about the same as the QuickBench.

Has anyone done a test on SIX+ X15s on file sizes larger than 1MB? I would like to see if the 20% increase is maintained...

Mark

kaye
05-10-2001, 10:26 AM
Mark,

QuickBench 1.5 has a tab called Extended that will test, in one MB increments, from 2MB up to 10MB. Or did you know this already? k

drallen
05-10-2001, 11:03 PM
Kaye,

Thats where the 8MB read/writes came from (QuickBench 1.5).

I was just asking if anybody has done a test on file sizes (with 6+ X15s) larger than 1MB...

You got the Burly...now where is the New World Mac?

I only know what you guys tell me...

Mark

Ton
05-11-2001, 02:08 AM
Now that's what I would like to know too...

Btw Mark, you were curious about the setup? ok here goes:
PM 533 MP/1,5 gig ram/4 x X15 cheetah's in Gurus burly with ultimate hd coolers and granite cabling and diagnostic terminators/Atta UL3D.
Besides pci cards, the inside of the machine is still virgin so I have room for two more X15 cheetah's. Just curious if it's really worth the extra dow.
T.

magician
05-11-2001, 06:23 PM
off-hand, no idea.

kaye
05-11-2001, 06:41 PM
Mark,

That New World Mac is just a gleam in my eye. Want to wait for the NYC show first. k

magician
05-12-2001, 12:47 AM
agreed. I am waiting, too.

I'm pretty sure I will not leap until dual-1GHz machines ship.

That's about what it will take to retire beast. In the meantime, I have the shop machines in the basement when I need to test. The horrible thing is...they do not feel faster than beast.

Still.