ASUS RAMPAGE III EXTREME BIOS - V1601 - Updated OROM MOD

[pir8man]

Got the PCIe SSD but seems to be capped at around 1000MB/s , not sure why. EDIT: Got 1600MB/s in another program but seems maybe limitation of PCIe 2.0, this thing is supposed to do 2700, and also 270k IOPS but it's doing 90k.
Also, it's impossible to boot from it which is really REALLY annoying as I wanted to use it as the boot drive. NVMe need the UEFI BIOS which r3e isn't.. /grumble
How did you get it to overclock so well? I have a 980x which is basically the same thing as the W3680 and it's not happy at 4.6ghz although 4 is great
HOW Did you get it to boot from a PCIE drive? My BIOS won't see it? is it some special card with controller/option rom etc?
(Crazy I have 7 hard drives in my computer now testing various things.. lol)

Check out the attached benchmarks.. the OCZ SSD's aren't even SATA3
I want to know how my super old SSDs are outdoing the new Samsung 950 at the lower block sizes even though they are in raid0 should be no contest

Which slot are you using for your PCIe SSD card?  the 4x slot is PCIe 1.0 and limited to 2500Gb/s.  the x16 slots are alll PCIe 2.0, so a 4x card gets 5000Gb/s.  And use a queue depth of 10 for test, NCQ is where SSD shines.

On the UEFI BIOS for R3E obstacle, i was just reading today about how to trick that...   http://www.rodsbooks.com/bios2uefi/
It appears to be a boot partition that loads a UEFI environment that in turn boots UEFI mode.
EDIT: I also just found this: http://www.sevenforums.com/installation-...-duet.html
(both appear to need a BIOS bootable device that can then point to any device to boot from)

My card is OCZ RevoDrive3 it has it's own firmware/bios which makes it recognized by R3E BIOS as bootable device.

I am still fine-tuning my OC. I will post my BIOS template once I finalize it.

[pir8man]

I was just reading on win-raid that INTEL RAID OROM 11.2 is the tested best firmware match for our ICH10R. It gets the best benchmarks.

so 10.1 is if you need to fit in memory constraint... 11.2 is fastest... 13.x is latest.

[pir8man]

I have completed my overclock except for memory timings.  Although I do not want to hijack my own thread with OCing, I will post my findings for others to review...  Your mileage WILL VARY!

I have overclocked BloomField (C0 and D0 stepping) and now GulfTown (B1).  I have done extreme air and water cooling solutions.  My current setup is a full motherboard waterblock and cpu waterblock with custom radiator/res/pump setup. 

BloomField C0 are horrible overclockers:  Air coolers may peak somewhere around 3.7GHz.  Water coolers may peak about 4GHz.
BloomFiled D0 and Gulftown B1 are good overclockers:  Air coolers should peak around 4GHz. Water coolers hope for +4.4GHz.

In my testing, I found that the motherboard does NOT need any better cooling than the stock cooler.  I would only recommend motherboard waterblock when runninng with an LN2 phase-change unit for the CPU.  Only with that setup would anyone really push this motherboard to it's thermal limits.

My Findings:
There are 2 combatants for us overclockers jitter/noise in the internal signaling of our cpu/mobo components, and heat.  We tend to throw voltage at any overclocking issue which actually creates more jitter AND heat.  On this motherboard, we have three helpful BIOS options that directly reduce jitter/noise.  These are CPU Clock SKEW, IOH Clock SKEW, and CPU PLL Voltage. Adding clock SKEW does more to stop jitter/noise than adding voltage.  A little goes a long way here, thankfully it's also hard to overdo it.  I ran mine all the way up to 1000ps in testing and worked down to 500ps for both CPU/IOH when it crashed.  This is where I learned they like to be at different offsets 500/600ps passed on down to the current 100/200ps delays. Every motherboard and Processor is different in how much skew it will want for any overclock, so start high and work your way down here once you reach your overclock goals.  If reaching higher from a stable overclock, add a few hundred to each to assist with keeping voltage needs down.

HEAT is the first wall i hit with any of my CPU's on this motherboard for both UnCore AND CPU.  First the BloomField's 2x multiplier to memory speed was creating way too much heat.  D0 stepping perform better here as they can run same mhz with less voltage (thus heat).  Gulftown has 1.5x uncore speed requirement which allows for high ram speeds with lower uncore speed and voltage.

CPU PLL Voltage feeds the clock generator that clocks up the CPU and UnCore.  Any noise here is multiplied.  So the lower the CPU PLL voltage, the less noise passed on to the CPU and UnCore.  Once I found a decent overclock at 1.8v. I then lowered this as much as possible.  Once lowered, it produced less heat and noise, allowing me to lower my CPU Voltage and QPI/DRAM Voltage.

Finally, make sure you enabled QPI LL Calibration.  (the jumber just left of ATX24 power connector)
_______________________

I gotta say though my Xeon W3680 has a memory controller rated for DDR3-1333.  The 980X is still rated for DDR3-1066.  So maybe my memory controller & un-core are better.  Thus allowing me such low QPI/DRAM voltage.

IF anyone has any questions, please PM me or start a new thread somewhere.  I will not respond to OverClock questions in this thread.
_______________________
New Template has many changes/updates.  Too many to list as it was still a work in progress. pretty [censored] complete now tho.

[f00z]

[quote pid='108749' dateline='1455521457']
Which slot are you using for your PCIe SSD card?  the 4x slot is PCIe 1.0 and limited to 2500Gb/s.  the x16 slots are alll PCIe 2.0, so a 4x card gets 5000Gb/s.  And use a queue depth of 10 for test, NCQ is where SSD shines.

On the UEFI BIOS for R3E obstacle, i was just reading today about how to trick that...   http://www.rodsbooks.com/bios2uefi/
It appears to be a boot partition that loads a UEFI environment that in turn boots UEFI mode.
EDIT: I also just found this: http://www.sevenforums.com/installation-...-duet.html
(both appear to need a BIOS bootable device that can then point to any device to boot from)

My card is OCZ RevoDrive3 it has it's own firmware/bios which makes it recognized by R3E BIOS as bootable device.

I am still fine-tuning my OC. I will post my BIOS template once I finalize it.
[/quote]

The whole board is pcie 2.0, which means each lane is (5Gbps with like 20% overhead) so probably 500MB/s max each lane.  So the Samsung ssd should max out around 2000 although its not.. I did get a 1700 in one benchmark here but I got less in another, it's strange.

Still no way to boot from it because it requires NVMe driver so I have to have some BIOS I can actually load drivers in; I saw the bios2uefi but that looks pretty hackish

[pir8man]

f00z: if your board is pcie2.0 and you plug it into one of the 16x pcie slots. you have theoretical speeds of 5Gbps data path to your drives controller chip and the memory behind it.

This gives you a theoretical drive speed of 2500MB/s. Your Samsung 950 PRO M2 card is equivalent of PCIe3.0. so it's benchmarks would be reproduceable only on PCIe3.0. Our board with PCIe2.0 would never keep up with what the ssd drive is capable of. I assume you got the 512MB card as the 256 is only capable of 1500MB/s speed. In a perfect world, we would be thinking I have a controller with a data path of 2500MB/s and a drive capable of same... it should work right? BUT, the x58 chipset struggles to get 2500MB/s available and some always gets lost in translation. If you can reproduce that benchmark of 1700MB/s, I would consider yourself lucky there. Honestly, to expect anything faster than 1000MB/s off this chipset is unlikely (IMHO).

[TechyMike]

Any chance of having the following:

*** Intel ICH10R SATA RAID Controller: v11.2.0.1527 (TRIM mod)***
Marvell 91xx SATA 6G Controller - Bios : v1.0.0.1038*
Marvell 91xx SATA 6G Controller - Firmware: v2.2.0.1125*
Marvell 91xx SATA 6G Controller - Bootloader: v1.0.1.0002b*
JMicron JMB36X Controller: v1.08.01
Intel 82567V-2 Gigabit Network: v1.5.43
SLIC 2.1: ASUS (SSV3/MMtool method)

I am using Intel Rapid Storage Technlogy (RST) Drivers v11.2.0.1006 and I assume that the version above for the ICH10R controller is the best one to have?

[pir8man]

Any chance of having the following:

*** Intel ICH10R SATA RAID Controller: v11.2.0.1527 (TRIM mod)***
Marvell 91xx SATA 6G Controller - Bios : v1.0.0.1038*
Marvell 91xx SATA 6G Controller - Firmware: v2.2.0.1125*
Marvell 91xx SATA 6G Controller - Bootloader: v1.0.1.0002b*
JMicron JMB36X Controller: v1.08.01
Intel 82567V-2 Gigabit Network: v1.5.43
SLIC 2.1: ASUS (SSV3/MMtool method)

I am using Intel Rapid Storage Technlogy (RST) Drivers v11.2.0.1006 and I assume that the version above for the ICH10R controller is the best one to have?

Check post # 3.  that is exactly what you requested

[TechyMike]

Thanks.

[javitux]

Hello im a new user of bios mod. I actually instaled zios mod 1601_S17. Im planing to install yours S18. Which version shall to download? I got a SSD 480g SATA3b, 1tg HHD SATA 2.
Thanks for your time, im noob.

[pir8man]

welcome javitux. glad to help out.

If you only have one of each of those drives, you do not need any of the TRIM modified versions. As you did not mention any other drives, I will assume you do not use the PATA controller. And for those particular drives, I would not use the Marvell controller either. Disabling both of those in BIOS will speed up boot time and ensure enough BIOS memory space for updated Network Card firmware.
That leaves you with 1601_s18_slic.rar from post #2.. AND 1601_S18i_slic.rar from post #3.

As far as which is better for your particular hardware... Only testing both and running benchmarks will tell the truth. Fortunately this is another reason why our mainboard has two BIOS'.