How much host writes does Windows 8 Fast Start feature use?

Windows 8 has a cool feature called Fast Start. When you enable it from Power Options dialog, it uses the hiberfile.sys to save part of the running state when you shutdown. So when you boot into Windows, it loads that part it saved to the hiberfile.sys directly to memory, instead of doing the full blown loading process. Like you expected, shutting down takes some time because it has to save whatever it needs to the disk before switching the power off.

At first, I thought this Fast Start feature saved only a small amount of information to the disk. But it turns out, it saves a considerable amount of information to the disk. I checked this by monitoring the amount of host writes to my SSD.

Previously I monitored the amount of host writes as at 27th of March to be 0.8GB. Then today (7th of May) I measured it to be 1TB. That’s roughly 200GB (it could be more or less depending on how Samsung Magician software rounds it off) in a span of 40 days. During that time, I only installed Octave and VMware Player to the SSD. It would have totaled to about 500MB. Maybe give another 10GB for temp files, browser cache, mail cache, Windows updates, driver updates and few other not-so-important things. Surely, it all wouldn’t take up 10GB. That’s all. The rest of it has to be those writes related to the Fast Start feature. So it’s roughly 5GB/day.

And I don’t even turn off my PC that often. On weekdays, I would do it twice. On weekends, only once. Restarts don’t count because then it does not save the state. So in total, it would come to about 70 shutdowns. OK, call it 100. Doesn’t matter how much it saves per shutdown. What matters is how much it saves per day. In total, that’s about 5GB per day. That’s a lot!!! That’s a lot of SSD wear and tear.

So, what I’m going to do is, I’m going to disable Fast Start and monitor for month and see how things progress. Since I haven’t done a properly controlled test, you can always argue these results are not valid. But it’s for your SSD’s own safety. Disable Fast Start if you value your SSD. Heck, you are only saving a couple of seconds of boot time. You can always use Standby. :)

To upgrade to Haswell from Sandybridge or not! That’s the question.

With Intel’s Haswell release being right around the corner, the upgrade itch is crawling back into me. I currently use a Intel Core i7 2600K CPU and ASRock Z68 Extreme4 motherboard (i.e. Sandybridge) in my PC and they are quite capable even in the demanding tasks. The CPU is overclocked to 4.5GHz, which makes it way faster than the current (3rd) generation flagship CPU in every task. Probably, it would remain the same when Haswell CPUs come out as well, because the rate of increase in performance with each new generation of CPUs has dropped to single digit percentage numbers in the last few years. If the rumors are true, Haswell won’t be any different (this, this and this). If you checked that last link, you will see that most of the performance improvements come from the GPU side. I don’t care about that because I would always be using a discreet graphics card. (But knowing that Handbrake would get QuickSync support in the future, makes things a bit more interesting.) All in all, my CPU at 4.5GHz would still own the flagship member from Haswell at stock – at least in most benchmarks.

But the question remains, why am I even considering upgrading to Haswell? That’s because Haswell itself will overclock and once overclocked, it would leave my current overclocked CPU in the dust. If the rumors are true, you might be able to hit 6GHz with proper cooling with a Haswell CPU. That would make it about 50% faster than my current CPU. No concrete information is available about the overclockability of Haswell CPUs so I cannot really make a decision until the official reviews are out. But I am hopeful. We’ll find all about it in a month.

Haswell is not only about performance. It is also about power efficiency. Haswell CPUs will be able to do more work with less power. And the idle power usage seems to have dropped by 10 times. This is not anymore a rumor, but it seems that most of the crappy CPUs won’t be able to deliver the low currents required for Haswell CPUs when at idle. We always looked at how the PSU copes with high power demands and never at how the PSU copes with low power demands. Seems that crappy PSUs cannot deliver low power “stably”. That would probably give BSODs. I checked on Intel’s website and seems like my PSU is supported after all (Corsair 75-001311 aka 850TX). So I don’t need to change the PSU. But even if I had to, I wouldn’t be to displeased about it because I get to buy a PSU with modular cables and also something more efficient that the one I have.

But, an important question remains: how important is power efficiency improvements in a desktop environment? Not so much, right, because there are so many other things that would be using up the power. The video card is the biggest culprit. Then there is RAM, hard drives, audio card and fan.

Then there is this USB issue with the early chips. With the buggy chips, when your PC goes into standby and you have a storage device plugged into USB3.0 ports, you will have to disconnect and reconnect the device to make it work after you resume from standby. It’s a nasty bug, but it is fixed in the latest chips. But the bug fixed chips won’t make it to the first motherboards. Rumor is that it would take about a month before the bug fixed versions of the motherboards hit the market. I would definitely wait for that.

When Haswell CPUs go on sale in Japan, the second hand market value of my current CPU and the motherboard would drop. I don’t know by how much, but it will definitely drop. I will have to evaluate the situation against those figures. I don’t want to receive only 1/3 the price of what I’m paying for the Haswell upgrade when I sell my current CPU and motherboard. That just doesn’t seem right.

All of these have to align, if I am to go ahead with the upgrade. The CPU has to overclock well; the bug fixed motherboards have to hit the market; I need to get something valuable from selling off my old stuff. I’m mostly concerned about the latter. Because even right now, I’m getting only about JPY 22,000 if I sell my CPU and mobo. :(

One important thing I forgot to mention. I currently use the Zalman CNPS9900MAX-B air cooler, and it is not the best out there, especially because it is so loud when the fans are running at full speed. I might upgrade this cooler to something decent with the Haswell upgrade. All depends on the overclockability. Too bad this case doesn’t support most of the hot (not literally) coolers out there. :(

Finally, let me list up the reasons for the upgrade.

• 50% performance improvement over my current gear – when overclocked
• Reduced overall power consumption
• Better Windows 8 compatible hardware (motherboard site perhaps)
• I only have to upgrade the CPU and the motherboard. (unless speed of RAM makes a difference; I only have ones running at 1600MHz)
• Native USB 3.0 support (IvyBridge brought this to the table, but I don’t have an IvyBridge rig, do I?)
• PCI-E 3.0 support. (Again, IvyBridge brought this to the table, but I don’t have an IvyBridge rig, do I? Besides, I already have a PCI-E 3.0 graphics card.)

Pixel Clock overclocking without a GeForce Titan?

NVidia’s GeForce Titan brought the single GPU performance to a new level. So did the single GPU price. But they were not the only things it brought to the table. There were some fancy features, and one of them was Pixel Clock Overclocking.

We all know that most displays these days run at a native resolution of 60Hz. While 60Hz is enough for most things, you can actually get some benefit out of displays running at a refresh rate higher than 60Hz in some specific use cases. One such use case is gaming. If you have a graphics card with enough juice to give more than 60FPS, your frame animations will be more fluid on a display that runs at a higher refresh rate.

120Hz displays are becoming more and more common place – among gamers at least – because of this reason. I haven’t actually used a 120Hz display, but most people speak so highly of the improvement after the upgrade. Sure, you need a GPU capable of pushing those frames past 60FPS to get any advantage. These displays were primarily made for 3D because the actual refresh rate halves when you use 3D. But you don't have to use 3D to get the benefits of a 120Hz display.

By overclocking your display’s Pixel Clock, you can raise the refresh rate beyond the stock refresh rate. This not only applies to 60Hz displays, but 120Hz displays as well. Just with standard CPU and GPU overclocking, how much you can increase the refresh depends on luck. You won’t get the same results between two identical displays. You have to find out what your display is capable of, on your own.

While Pixel Clock Overclocking is a feature that’s officially supported only on the Titan, you can use EVGA’s PrecisionX utility to enable it with any GeForce 600 series graphics card.

Here's how you do it.

PrecisionX Main Window

PrecisionX Settings Page (first tab)

Pixel Clock OC settings