3Ware 9690SA-8i with BBU - Revisted (PICS)

[Mr.Guvernment]

Well if the flash doesnt do anything and they arent willing to replace this card with a card that performs as they advertised, i am going to the BBB and everyother organization and going to make life hell for 3ware, i am tired of companies promising X performance on a product, but once you get it, it doesnt come anywhere near what they claim

 

[SLORider]

Slorider... Turn on the writeback cache and watch it scream!

Are you saying that you've gotten better performance out of the 3Ware card? I've already tried turning on WB cache and switching to "Performance" mode in the 3Ware web management utility.

Well if the flash doesnt do anything and they arent willing to replace this card with a card that performs as they advertised, i am going to the BBB and everyother organization and going to make life hell for 3ware, i am tired of companies promising X performance on a product, but once you get it, it doesnt come anywhere near what they claim

Not sure where 3Ware advertised any specific claim of "X" performance. It states PCIe 8x, but that isn't a claim of guaranteed throughput. The product is solid, reliable and works. It just doesn't perform well.

I think you'll be hard-pressed to nail them on the basis of bad faith advertising. I'm just going to eBay it and take my lumps for not doing better research.

I've purchased a lot of 3Ware cards in the past and was happy, but I never put them to a hard performance test. This will be the end of my 3Ware purchases---HOWEVER, maybe since LSI bought them they will have some future products that will be much better.

 

[Joeteck]

Are you saying that you've gotten better performance out of the 3Ware card? I've already tried turning on WB cache and switching to "Performance" mode in the 3Ware web management utility.



Not sure where 3Ware advertised any specific claim of "X" performance. It states PCIe 8x, but that isn't a claim of guaranteed throughput. The product is solid, reliable and works. It just doesn't perform well.

I think you'll be hard-pressed to nail them on the basis of bad faith advertising. I'm just going to eBay it and take my lumps for not doing better research.

I've purchased a lot of 3Ware cards in the past and was happy, but I never put them to a hard performance test. This will be the end of my 3Ware purchases---HOWEVER, maybe since LSI bought them they will have some future products that will be much better.

no, write back cache with the matrix raid.

Mr Gov, I'm with you 110%. Lets do it... The last online tech support conversation put me over the edge!!!

 

[SLORider]

no, write back cache with the matrix raid.

Oh, right...

Intel X25-E SSDs (Matrix RAID 1) without/with write-back cache:


Somehow, I think this might be sort of a "lab" result and not exactly representative of reality--although the numbers are astoundingly different. I'd like to run some more "real world" comparisions.

 

[Mr.Guvernment]

well tomorrow hopefully i can flash my seagate drives and can post back any new tests i can run.

 

[ElFi]

I had the same issue with my 9690SA 4I4E. I write more tomorrow.

 

[Joeteck]

Oh, right...

Intel X25-E SSDs (Matrix RAID 1) without/with write-back cache:
View attachment 73548 View attachment 73549

Somehow, I think this might be sort of a "lab" result and not exactly representative of reality--although the numbers are astoundingly different. I'd like to run some more "real world" comparisions.

HOLY CRAP!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

 

[ElFi]

Hard drives are Seagate ST3500418AS 7200.12 500GB x2 in RAID0 mode.
Controllers are: 3ware 9690SA 4I4E, Intel SRCSAS8BBI and Adaptec 5805.
3ware and Intel work with Windows XP pro 32-bit, Adaptec with Windows Vista 64-bit.
Here is few tests:
​

 

[ElFi]

3ware ReadCache function is not work correctly? I'am not sure

When disable read cache function in Adaptec controller results are same that 3ware controller:

 

[ElFi]

But Everest test in Adaptec when ReadCache is disabled show 222MB/sec (Buffered read) instaed 3ware test show 308MB/sec - 86MB/sec higher than Adaptec (with ReadCache mode off).

Any suggestion?
3ware ReadCache function not work correctly or HDtune read test on 3ware controller is not work correctly?

This picture is Everest test on Adaptec 5805 with ReadCache function is disabled:

 

[Joeteck]

well tomorrow hopefully i can flash my seagate drives and can post back any new tests i can run.

Hows that class action suit coming??

 

[Mr.Guvernment]

ahaha, no where yet

okay finally flashed the drives today, speed did go up about 20MB average, not sure maybe i am expecting too much for a raid 50 array? for drives that can do around 95-100MB avg on their own, 6 of them..

Unfortuantly i cant just rip this array apart and test raid 0 or something since it is a production machine, but may need to see if i should talk to our CEO about lawyers talking to lawyers

 

[Mr.Guvernment]

just saw this

You Can this card be used in a PCI-e x16 video slot?
Sam no it won't work with the video card slot

a PCIe slot is a PCIe slot is a PCIe slot..lol

they are all the same!!!!

besides i tried this in a dell SC440 that has 8x PCIe slots and same performance, so BS to you 3ware...

So right now in my Asus K8N-DL it is in the only PCIe slot, 16x and it is working....

 

[Mr.Guvernment]

So i emailed them back saying performance still didnt seem great and they sent me this link

http://www.3ware.com/KB/article.aspx?id=15228

numbers are kind of all over the place.

 

[solstice]

I just want to clarify a few things about PCI-e 16x slots which seem to lead a lot of confusions here. A PCIe x16 slot is often not a true PCIe x16 if it can be used in a share mode. So the most simplest example are the triple SLI. The First slot and Third slot if occupy provides x16 lane each slot. The second SLI slot is only x1 when using any card other than SLI card. When using SLI mode, the First and the second share the bandwidth. This basically creates an x8 to each lane. So you get, x16, x8, and x8 mode. In this mode, you are not true independent operations. This all came from my ASUS motherboard manual.

Some of you may recall the first PCIe SLI requires a card to be flipped to enter SLI mode. Well, now this may done electronically now with either the PCIe chipset or external chip. In either way, my feeling is that this is not a true configurations as a server environment would have. Also, in order to have SLI, the PCIe chipset act differently to enable performance.

Now, this is nothing to do with 3ware card performance. I just want to make an note that it is important to read your motherboard manual. My ASUS actually told me that my second PCIe slot is really only x1 and should only be used under triple SLI condition.

I will be getting a 3ware 9690SA card tomorrow or the day after. Since I already have SLI, I am stuck with x1. I am surprised that the performance seem to be limited at 150MB/sec. I would have expected closer to 200MB/sec since PCIe raw data is basically 250MB/sec. With overhead, I would expect things to be around 200MB/sec. In any case, I will try to benchmark this with against ICP8085BL (Adaptec bought ICP and the card looks exactly like Adaptec 3805.)

I will also try the supposedly x16 slot by removing my beloved SLI card just for benchmarking.

 

[solstice]

I just want to clarify a few things about PCI-e 16x slots which seem to lead a lot of confusions here. A PCIe x16 slot is often not a true PCIe x16 if it can be used in a share mode. So the most simplest example are the triple SLI. The First slot and Third slot if occupy provides x16 lane each slot. The second SLI slot is only x1 when using any card other than SLI card. When using SLI mode, the First and the second share the bandwidth. This basically creates an x8 to each lane. So you get, x16, x8, and x8 mode. In this mode, you are not true independent operations. This all came from my ASUS motherboard manual.

Some of you may recall the first PCIe SLI requires a card to be flipped to enter SLI mode. Well, now this may done electronically now with either the PCIe chipset or external chip. In either way, my feeling is that this is not a true configurations as a server environment would have. Also, in order to have SLI, the PCIe chipset act differently to enable performance.

Now, this is nothing to do with 3ware card performance. I just want to make an note that it is important to read your motherboard manual. My ASUS actually told me that my second PCIe slot is really only x1 and should only be used under triple SLI condition.

I will be getting a 3ware 9690SA card tomorrow or the day after. Since I already have SLI, I am stuck with x1. I am surprised that the performance seem to be limited at 150MB/sec. I would have expected closer to 200MB/sec since PCIe raw data is basically 250MB/sec. With overhead, I would expect things to be around 200MB/sec. In any case, I will try to benchmark this with against ICP8085BL (Adaptec bought ICP and the card looks exactly like Adaptec 3805.)

I will also try the supposedly x16 slot by removing my beloved SLI card just for benchmarking.

What I found is that my x16 on the third slot seem to have become x1 transfer rate. I get better performance on x16 slot 2. The performance seem decent enough and I am okay with it since I only have 2 drive RAID 0.

1. ICP8085
Max was max 124 MB/sec, min 80.2 MB/drv, avg 107.7MB/sec.

burst 200MB/sec!

2 Raptor 74GB WD74G RAID0 with 64 kb strip

2. 3Ware 9690SA-8I
Max was 143MB/sec, Min was 94.4MB/sec, avg was 117.3MB/sec

burst 92.1MB/sec

Note: Read cache has been set to basic, store save to balance mode.

2 Hitachi 7200RPM 640GB RAID0 with 64 kb strip

I am happy with my RAID set up and will keep 3Ware. The ICP will go into another computer and will use Raptor in RAID0.

I still don't see x1 being the true limitations since 200MB/sec was reached with the burst mode. What I suspect is that the strip is set to be 256kb and that slows down sequential test. In any case, its an interesting excercise.

 

[solstice]

Quick Update:

I enabled Read Cache to intelligent. The HD Tune 2.55 jumps to 201MB/sec peak with the same RAID0 configuration as above. Now, I am limited by PCI-E x1.

Min 92.2MB/sec, Max 201.2MB/sec, Avg = 157.9MB/sec

I suspect this intelligence algorithm is really best for sequential operations. For me, that is perfectly fine since I use this for a desktop gaming system with video editing on the side. So, I am happy.

 

[Airbornederekc]

OK i think i have to jump in here and defend 3ware. You are running a SAS RAID card with 4 sata drive sin a RAID 0 on a celeron based PC and you are wondering why the drives arent performing well?

you need some more CPU horsepower to be able to handle the data throughput for one.

Have you gone into the 3dm2 settings and adjusted the settings as far as performance and caching are concerned? the battery allows you to have caching enabled and saves you from data loss in the event of power failure. it is not a performance enhancer. turing on the cache is and the battery is a saver.

When talking to SATA hard drives with a SAS controller there is a sligh overhead in the conversion of the SAS protocol to the SATA drives. it isnt much but the SATA cards perform better with SATA drives.

 

[solstice]

OK i think i have to jump in here and defend 3ware. You are running a SAS RAID card with 4 sata drive sin a RAID 0 on a celeron based PC and you are wondering why the drives arent performing well?

you need some more CPU horsepower to be able to handle the data throughput for one.

Have you gone into the 3dm2 settings and adjusted the settings as far as performance and caching are concerned? the battery allows you to have caching enabled and saves you from data loss in the event of power failure. it is not a performance enhancer. turing on the cache is and the battery is a saver.

When talking to SATA hard drives with a SAS controller there is a sligh overhead in the conversion of the SAS protocol to the SATA drives. it isnt much but the SATA cards perform better with SATA drives.

I was with you until you say there is a slight overhead with SATA drive on SAS controller? Where did you read that? Now I am curious about that? My understanding is that the SATA controller is basically just a variant of SAS. Is it not true?

Oh, the latest release of the firmware 9.5.2 and driver 9.5.2 added an updated intelligent Read Cache. It supposed to boost performance. For me, it was very significant boost. I highly recommend anybody with 3ware card to update the firmware.

The original post was before the new firmware was released.

 

[Joeteck]

OK i think i have to jump in here and defend 3ware. You are running a SAS RAID card with 4 sata drive sin a RAID 0 on a celeron based PC and you are wondering why the drives arent performing well?

you need some more CPU horsepower to be able to handle the data throughput for one.

Have you gone into the 3dm2 settings and adjusted the settings as far as performance and caching are concerned? the battery allows you to have caching enabled and saves you from data loss in the event of power failure. it is not a performance enhancer. turing on the cache is and the battery is a saver.

When talking to SATA hard drives with a SAS controller there is a sligh overhead in the conversion of the SAS protocol to the SATA drives. it isnt much but the SATA cards perform better with SATA drives.

Actually, This is untrue... I have a Q6600 CPU with this same RAID card with 8 SAS drives, and it still did not perform... I even added the BBU, which claimed to enable the caching function of the device... the 9690SA is just a POS...simple as that.

So many people with the same outcome.... something has to be wrong... and its not the Computers or hardware...

I was told by 3ware that is will not work in a PCI-e video card slot. It has to be a dedicated 8x slot not for graphics....