asrock z97 pro4 multicore enhancement

I recently completed two new builds using the Asrock z97 pro4 motherboard.  One build has the i5-4590 and the other has the i5-4690k.  Everything is working as expected at stock settings and I'm quite happy with the performance of both systems.

Edited by fdisker - 01 Sep 2015 at 8:39pm

Sorry, just noticed your thread!

I think Multi-Core enhancement should work with your i5-4590. My ASRock Z97 board does not have an option called Multi-Core Enhancement (MCE), but does have an option called CPU OC Fixed Mode.

It does the same thing as MCE, sets all cores to the maximum Turbo multiplier of the CPU. I have a Haswell Xeon E3-1276 v3 in the Z97 board, and  CPU OC Fixed Mode sets the multiplier to the maximum Turbo value (40) of that CPU. This Xeon is NOT an unlocked processor, as some Xeons are. I am quite sure the E3-1276 v3 is not an unlocked processor.

So why does your 4590 not respond to MCE apparently? Could be caused by other settings in the UEFI/BIOS that you've set. Since you mentioned you were an "old school" over clocker, I'll mention a few important things.

Do NOT disable Turbo Boost! That was popular back in the Core 2 and Nehalem processor days when over clocking. Not any more. Turbo MUST be enabled or you won't even move beyond the base clock (non-Turbo) speed.

Disabling Speed Step can affect Turbo. Haswell processor OC is different than the previous generation Intel processors. Forget much of what used to work to OC, it does not apply anymore.

Try loading UEFI defaults and then enable MCE on the 4590 PC. Don't change any of the Turbo or CPU power saving options, except to Enable them if necessary. Have the CPU multiplier option set to Auto, not Per Core or All Core.

MCE and CPU OC Fixed Mode may be just smoke and mirrors on non-K processors. With OC Fixed Mode enabled on my Xeon, when I start a CPU stress test, the core speeds change from 4.0GHz to 3.8GHz.

Must sleep now, but I'll chat about your SSDs later, my favorite PC topic.

fdisker wrote: Hello parsec, thanks so much for getting back to me.  I spent a couple of hours testing the 4590 and no matter what I did I couldn't get multi core enhancement to work.  Guess what .... I don't really care.  I spend the majority of my time messing with the 4690k and testing various SSD drives.  I can't explain why I'm so fascinated with them ... I just am. The EVO 850 is an excellent drive.  I really like it because it's fast and the form factor is super cool.  Having just spent several hours messing around on my alternate rig (4590 with Intel 730 SSD) I can tell you it's difficult to tell the speed difference between the two drives.  Looking back at your message I can see you like SSDs as well.  That good because I have some questions for you! I'm going to make a couple of assumptions and I'd like your thoughts on them. 1. The 850 EVO is about as good as it gets on the SATA interface. 2. Read/Write performance increases require a move to PCIe drives such as the Plextor M6e, Kingston HyperX Predator or Samsung sm951 drives. 3. Perceived speed doesn't really change until next generation NVMe drives show up in the next few years Point 3 is the big one.  The EVO 850 is a big leap forward from a spinning disk but a small step compared to my last generation Intel 730.  What I'm really interested in is that next big step forward in performance that I can feel in day to day use.  Sure, I could upgrade my EVO 850 to a sm951 and double or triple my benchmark score but would I really notice a difference in day to day performance?  My current machine boots from power off to desktop in 15 seconds without any tweaking whatsoever!  I almost pulled the trigger on the sm951 until I started looking at the benchmarks. SM951 4k read iops = 11,970 EVO850 4k read iops = 11,942 SM951 4k write iops = 37,565 EVO850 4k write iops = 33,979 Results for the SM951 come from THESSDREVIEW.COM and the EVO850 results are mine (both using ANVIL).  I'm under the impression that 4k read/write speed is THE most important aspect regarding storage in day to day use.  Doesn't make much sense to upgrade when they're so close.  I'm hoping the NMVe drives are the real game changers. One more thing.  I want to reiterate how happy I am with the two ASROCK z97 pro4 boards I own.  I've spent way more time fiddling, changing, messing and tweaking with them than I should have and I haven't had a single crash or blue screen with them.  I give them two thumbs up.

The 840 EVO and 850 EVO (250GB+ capacity) are the only two SATA SSDs that significantly surpass 10,000 IOPs in 4K random read. The SM951 can do a bit better than both of them, but not by much, showing us how difficult it is to reach 10,000+ 4K random IOPs.

SATA III is at its speed/performance limits now, PCIe and other techniques are necessary to move beyond where we are now with SATA III. SATA Express which we still don't have is really SATA ports in parallel, but the equivalent can be done with PCIe, and already has been done.

I have one or more Samsung SM951 (AHCI), 840 EVO, 840 Pro, SanDisk Extreme II and Extreme Pro, Intel 520, 530, and 730, and others. I used a Plextor M6e for a short time. I've used RAID 0 arrays of 840 Pros, SanDisk Extreme IIs, Intel 730s, and others.

In my experience, any of the top performing SSDs are almost indistinguishable from each other in actual use. That includes the SM951 and Plextor M6e, and the other Samsung and SanDisk SSDs I listed. The Intel SSDs are also very close if not identical when booting an OS, for example.

RAID 0 arrays do not improve booting speed, they slow it down somewhat actually. Their 4K random IOPs are lower than a single SSD of the same model. All benchmark tests reveal that. So they boot slower than single SSDs, and the RAID software takes a second to get going.

I time "boot time" as the time from the single POST Ok beep, to the desktop being displayed, no login screen. Three seconds seems to be the limit I cannot improve upon, with the SM951 or any other SSD I've used. My SATA SSDs can do three seconds, so no improvement using the SM951. That may be a Windows "limitation", or just how long at best it takes to do everything it must do.

I can say that the SM951 can install software faster in some cases. Virus scans using it seem slower to me than some SATA SSDs, but that might be me expecting it to be faster. A Windows installation from a USB 3.0 flash drive in a USB 3.0 port was about five minutes, from the "Preparing files" to the first restart. The shutdown time is the fastest I've ever seen with any SSD, the screen goes blank almost immediately, and the PC is off two seconds later.

NVMe SSDs are so new and so few it's almost impossible to comment on them reasonably. The Intel 750 is known to be tuned for best performance in 8K and larger files. It also boots slower than SATA SSDs for most users, but I suspect that is because their Windows installations are done in Legacy mode (CSM Enabled) so aren't using the EFI boot loader. Note that Intel supplies their own NVMe driver for use with the 750, and only the very latest versions of the Intel IRST driver when used with the Z170 chipset recognize PCIe SSDs, including NVMe.

The Samsung NVMe SM951 SSD is IMO more problematic than the Intel 750. That is caused by the Windows NVMe driver (apparently), which causes poor write performance. I mean worse than a SATA SSD, given some of the benchmark tests I've seen. That is caused apparently by an incompatibility between the Samsung firmware and the MSoft NVMe driver. Samsung does not have their own NVMe driver, and the Intel NVMe driver will only work with an Intel 750. Personally, I would not be jumping on an NVMe SM951 quite yet. If anyone must use Windows 7, the MSoft NVMe driver is only included with Win 8.1 and 10.

Of the three major advantages of NVMe SSDs over AHCI SSDs, generally lower latency, reduced 4K random read overhead time (faster), and vastly greater command queue processing capabilities, the first two only matter to PC users. Real world usage queue depths on PCs with standard SATA SSDs rarely even reach four, since they are so fast now. What good would multiple queues that can contain thousands of IO requests be to PC users?

What we need is a redesigned file system for flash storage devices. Since HDDs are far from extinct yet, we may not be seeing that for a while.

Another thing that is apparently on the horizon, is Intel teased a new type of flash storage technique and chip, different than the current NAND. Supposedly 100 times faster than NAND, we might see it as early as sometime in 2016. Adding that to NVMe could create a whole new world of SSDs.

Bottom line for you, I could not recommend anything that will be night and day over your 850 EVO, unless you do a lot of processing and copying of large files, meaning hundreds of megabytes or gigabytes.

Don't let me stop you from trying some new SSDs, their prices are at an all time low now. That is, except for the NVMe models and both SM951 types.

I sure did see the new 950 Pro NVMe PCIe SSD... of course just days after getting my second Samsung SM951, both the AHCI version.

While the SM951 and XP941 are in a sense "gray market" SSDs, the 950 Pro is cheaper, has full Magician software support (not that important IMO, but good), and has a full Samsung warranty.

One question I have about it is related to the NVMe version of the SM951. An NVMe driver is required for NVMe drives, which Windows 8 and 10 has now. But that driver had write performance issues in some of the reviews of the NVMe SM951. I don't know if MSoft updated that driver or not, and Samsung has not provided a NVMe driver for it yet on their download pages. Intel provides their own NVMe driver for their NVMe 750 SSD, which does not seem to works with other NVMe SSDs.

Looking at the specs of the 950 Pro in comparison to the SM951, if we ignore the high queue depth specs, and look only at sequential large file and random 4K small file performance, I feel much better about my SM951s. The benchmark results of my SM951's are equal, somewhat better, or a somewhat less than the 950 Pro specs, depending on the benchmark test used. As always, the specs are "Up to" or maximum speeds, and NOT guaranteed as some SSD users believe.

If we look at the high queue depth specs compared to my test results, the 950 Pro can be much faster. That is, if there are 32 IO requests waiting to be processed by the SSD. The Samsung 950 Pro specs are still referencing the standard AHCI queue depth max (32) which is FAR surpassed by NVMe. Odd they do that IMO, unless they believe we won't understand NVMe based specs. I can see that, but they could include some of both.

The potential (more likely actual) drawback of NVMe performance applied to the PC environment is, we rarely if ever are in an IO situation where the queue depth is greater than 4. That includes loading an OS. In other words, the new increased performance capability will never be used by 98% of PC users, and only seen in benchmark results. We are already there for the most part with current SSDs IMO. Unless you are constantly reading and writing multi-Gigabyte single files, the real world user experience will apparently be small to none.

I experience that now with my SM951. The fastest OS load/boot (which is from the POST OK single beep to the desktop display, without a login screen) is three seconds. That is the same with an 840 Pro (and others) and an SM951. Shutdown is faster with the SM951, about half of a SATA SSD. It is not a night and day difference using an SM951 IMO.

So wait and see what the reviews say, but I don't consider my SATA SSDs obsolete because of the existence of the new PCIe SSDs.