poor SATA performance on X79 chipsets compared to P67

[adamantinepiggy]

Has anyone noticed a markedly poorer SATA3 performance on the Intel SATA3 ports using motherboards with the x79 chipset (LGA2011) versus the p67 chipset (LGA1155)? In testing SSD's with the latest intel drivers, the x79 chipset is a good 10-15% behind in overall performance numbers compared to a standard P67 chipset test platform. The x79 motherboards in particular are the Asus p9x79 Pro and the Gigabyte GA-X79-UD5. Both of these with i7-3960x processors do not match the (Intel 6G SATA3 port) SATA performance numbers of a MSI P67-GD65 (p67 chipset) with a 2600K cpu. RAM on all these is 8GB, Win7-64.

The numbers between the x79 boards are comparable, but both are behind the P67 board. While it doesn't show so much in fast throughput tests like ASSSD or CDM (although it does show lower numbers, just not as big of margins), longer more-intensive tests like Vantage or IOmeter show a 10-15% consistently lower performance numbers with same SSD's.

These SSD's are all low-level formatted before testing so they start in the exact same state. All SATA drivers are the latest (can't remember the version at the moment) on the x79's, while the p67 is using RST 10.8. All power saving settings are turned off, write caching enabled, and write-cache buffer flushing disabled, for all tests.

While I assume some may be blamed on driver immaturity on the x79 chipset, 10-15% difference seems a bit much, and I'm wondering if it's the particular MB's or perhaps a BIOS or OS setting I'm missing on the newer x79 boards.

 

[groberts101]

well known by now that driver immaturity is the main cause of this. Playing the.. "a new driver is coming!".. waiting game is about all you can do. Well.. that and trying every beta driver that comes along. lol

never good to low level format any SSD(since it writes HDD type data) and some will even penalize you with speed losses. Secure erasure(via ATA SE command) is the best way to assure that drives are factory fresh although some controllers can be trimmed back to near fresh speeds.

PS. some of the newer platforms have had weird speed losses associated with C-states(some quite severe) and you might try disabling that in bios, as well. Good luck with it.

 

[adamantinepiggy]

never good to low level format any SSD(since it writes HDD type data) and some will even penalize you with speed losses. Secure erasure(via ATA SE command) is the best way to assure that drives are factory fresh although some controllers can be trimmed back to near fresh speeds.

PS. some of the newer platforms have had weird speed losses associated with C-states(some quite severe) and you might try disabling that in bios, as well. Good luck with it.

Well, the low-level formats we do in the R&D labs are not standard. Ours are scripted to include returns to blank factory state with a given test firmware. Longevity is not an issue when we produce the drives

Guess we'll just wait for different/newer drivers..

 

[Diogenes2]

My M4 128 actually shows improved performance on X79 vs Z68..

The M4 has had a firmware update since the Z68 test was run, so that might account for some of it..

 

[adamantinepiggy]

My M4 128 actually shows improved performance on X79 vs Z68..

The M4 has had a firmware update since the Z68 test was run, so that might account for some of it..

Well, since we make that SSD, I know it's not a firmware issue causing this issue. The z68 showed no statistical difference in performance from the p67 for us, even just after it was released (they use pretty much the same HBA drivers). It's the x79 that is being a dog, even with the fastest CPU. The current x79 driver appears to be a Enterprise/Server type driver as compared to the RST drivers for earlier chipsets. It actually shows a miniport SCSI driver under Storage Controllers rather than a driver under the IDE ATA/ATAPI controllers section.

 

[Diogenes2]

What issue ?

You said:

Has anyone noticed a markedly poorer SATA3 performance on the Intel SATA3 ports using motherboards with the x79 chipset (LGA2011) versus the p67 chipset (LGA1155)?

So no, I haven't noticed a markedly poorer SATA3 performance on the Intel SATA3 ports using motherboards with the x79 chipset (LGA2011) versus the p67 chipset (LGA1155)...


My X79 performance is better than Z68 , which should be essentially the same as P67 ...

 

[adamantinepiggy]

What issue ?

The issue where we can get 75+K marks in PCMark Vantage HDD Suite on a p67 chipset versus around 63K on an x79 chipset with the same SSD, same firmware, same OS, approximate same CPU clock speed, same amount of RAM, etc, etc. The assumption is that even though a new chipset, performance should not be markedly different of the SATA controller. As I mentioned, I'm guessing a immature driver issue, but 15% is quite a big margin of difference with the raw numbers.

My X79 performance is better than Z68 , which should be essentially the same as P67 ...

[/QUOTE]

What x79 motherboard and what Intel driver version are you using? And when you say the performance is better, what percentage better and in what test (roughly)?

 

[Diogenes2]

What x79 motherboard and what Intel driver version are you using? And when you say the performance is better, what percentage better and in what test (roughly)?

You might want to take another look at Post # 4, which you have quoted once ..

http://forums.anandtech.com/showpost.php?p=33030574&postcount=4

 

[adamantinepiggy]

Ahh, sorry, didn't see the pictures of the benches because I was at work and stupid IS blocks everything related to pictures because god forbid some boobies might show up accidentally . I can see the pics now that I'm at home. Thanks!

 

[groberts101]

Well, since we make that SSD, I know it's not a firmware issue causing this issue. The z68 showed no statistical difference in performance from the p67 for us, even just after it was released (they use pretty much the same HBA drivers). It's the x79 that is being a dog, even with the fastest CPU. The current x79 driver appears to be a Enterprise/Server type driver as compared to the RST drivers for earlier chipsets. It actually shows a miniport SCSI driver under Storage Controllers rather than a driver under the IDE ATA/ATAPI controllers section.

you answered your own question there, bud. :thumbsup:

If you take the enterprise drivers and use them on the 6 series chips?.. you will see similar losses. Might/probably won't be quite as severe as the same driver version being used on the X79 platform.. but you can certainly see where some of those losses are coming from.

Again we wait for Intel to get their drivers sorted out.

 

[Edrick]

I thought the new RST v11 drivers Intel just released fixes the issues on x79?

 

[Diogenes2]

I'm still wondering where the OP is getting their information.

Here are some tests by Anand, showing P67, Z68 and others compared to X79..

The Z68 & etc. was with a Vertex 3 which is faster than the Real SSD C300 used for the X79 testing..

http://www.anandtech.com/show/5434/asrock-x79-extreme9-review-price-for-performance/5

http://www.anandtech.com/show/4493/gigabyte-z68xud3hb3-review/5

http://www.anandtech.com/bench/Product/352?vs=419

 

[groberts101]

the problem with these platforms is that they often respond differently to various SSD controller/bios/driver implementations. Is/was the same on the previous platforms really.. just that we're starting over again, is all.

http://www.xtremesystems.org/forums...ntel-C600-series-chipset-SATA-RAID-Controller

http://forum.crucial.com/t5/Solid-S...-SSD-BENCHMARK-SCORES-HERE/td-p/58420/page/28

same sh**.. different platform

 

[adamantinepiggy]

Looks like I figured it out.. Something about the motherboard's OEM disk's 9.2.3 inf updater and the latest downloadable RSTe3 driver. Redid the system OS but used the downloadable 9.3 inf updater (skipped the 9.2.3 version) from Intel then directly installed the downloadable RSTe3 driver driver, bypassing the OEM driver disk completely. So basically something about the 9.2.3 inf updater and the new RSTe3 driver didn't like each other, or the inf updater on the OEM disk wasn't installing/corrupted.. who knows..

The speeds actually slightly surpassed (by less than 2-4%), but faster is faster and that is what we were looking for a new standard benchmarking machine.

FYI: The Asus MB died on me :-( so I didn't get to test whether the same procedure would work on it too..

Thanks all for the help!

As to Edrick who asked about RST 11, the latest version officially listed on Intel's site is 10.8 in the download center, and even then, if you run the executable, it tells you "unsupported platform" (or similar). Yes you can force the update using the inf files alone, but this machine has to be replicable to our Singapore manufacturing site so I can't do anything odd to it.

 

[Edrick]

As to Edrick who asked about RST 11, the latest version officially listed on Intel's site is 10.8 in the download center, and even then, if you run the executable, it tells you "unsupported platform" (or similar). Yes you can force the update using the inf files alone, but this machine has to be replicable to our Singapore manufacturing site so I can't do anything odd to it.

Strange, I downloaded v11 yesterday and installed it directly from Intel.

http://downloadcenter.intel.com/Det...&DwnldID=20868&keyword=rapid+storage&lang=eng

 

[Diogenes2]

X79 requires the ' Enterprise ' version of RSTE ..

 

[adamantinepiggy]

Strange, I downloaded v11 yesterday and installed it directly from Intel.

http://downloadcenter.intel.com/Det...&DwnldID=20868&keyword=rapid+storage&lang=eng

Hm Intel must be falling behind on their search engine. i went to the download center, put in the chipset, OS, driver menu requester and got 10.8 drivers as the latest according to the search.... oh well go figure..

I'll try tomorrow at work and see if the v11 still tells me "not compatible" when I exectte it on the x79 motherboard. The 10.8's executable certainly did.

 

[Diogenes2]

Those are still not the Enterprise version required for X79..

They are not easy to find ...

http://downloadcenter.intel.com/Detail_Desc.aspx?DwnldID=20507

They released version 3.0.1.7016 back in January but it has disappeared from Intel as well as the MOBO makers ..

 

[slow_poke]

you can load the 11 series driver thru the device manager and it gives a big boost to benchmarks on the x79. No doubt about it, they are not even close in performance on either of my x79 mobo's.

 

[Edrick]

you can load the 11 series driver thru the device manager and it gives a big boost to benchmarks on the x79. No doubt about it, they are not even close in performance on either of my x79 mobo's.

I just got my X79 boards and I have not been able to do that. I am stuck on the RSTe 3.0 driver and my SSD performance is not as good as it was on my P67 system

 

[bonidinimon]

Hi all

I'm having the same probleme with the X79 on my ASUS Rampage Formula IV.

The writing speed in the 4k-area is too slow:

http://h9.abload.de/img/dddddddd5oyqb.jpg

BUT when I disable the Windows write Cache on my SSD, the write speed for the 4k's is normal:

http://h11.abload.de/img/ssd06kto.jpg

Is this a bug in the Chipset or in the Driver?!

 

[Diogenes2]

Why would you want to cache an SSD ?

 

[groberts101]

there are various layers of windows based caching involved and I'll take any and all I can get on my systems. Especially since that's what ram caching was designed for in the first place. So, when SSD becomes as fast as ram?.. I'll probably revisit whether or not ram caching even helps at all. lol

As to the question above?.. that surely looks to be a driver related issue to me. Have you tried disabling c-states in the bios to see if that helps?

 

[Dkcode]

Does anyone know if this issue has been resolved?

 

[ryansc375]

I have an Asus Rampage IV Formula with a Sandisk Extreme 240GB and the latest Intel chipset and RSTe drivers and the SSD speed positively stinks. I previously had a Samsung 830 256GB and thought it was the drive that was the problem. These RSTe drivers are TERRIBLE (as of 6/7/12). My boot-up time (not counting time to post) is about 25 seconds (takes six to eight seconds on older systems using this same SSD) and my ATTO scores are way below what they should be. And yes, all the bios and software tweaks have been make to ensure max performance. I tried switching to the ASMEDIA SATA ports and they were even worse. Intel builds this high end X79 chipset only to leave its users with poor SSD performance because we can't use the "normal" non-RSTe drivers. If anyone figures out a solution short of future driver updates, please post!