Everything You Need to Know about Intel's New Z68 Chipset

Posted 05/10/2011 at 11:16pm | by Gordon Mah Ung

25

Comments

Print

Intel’s new Z68 chipset brings SSD caching and switchable graphics

On the surface, it’s easy to shrug your shoulders and say “meh” at Intel’s new Z68 chipset.

It doesn’t, for example, add any more than the two SATA 6GB/s ports that the P67 had nor does it add native USB 3.0. The single x16 PCI-E 2.0 isn’t improved either (nor can it be because those are within the CPU). But that’s doesn't mean the Z68 isn't an important step forward. 

In fact, the improvements it brings to the table are actually uniquely compelling. The top new features are:

1) The ability to overclock the graphics core in the Sandy Bridge CPU.

2) An Identity Protection Technology feature that essentially integrates a hardware token in to the PC.

3) Optional support for switching graphics between a discrete GPU and the integrated graphics processor in Sandy Bridge.

4) SSD caching that greatly improves the responsiveness of a system.

We’ll discuss them in order as introduced:

While the Sandy Bridge graphics core is much improved, enthusiasts are unlikey to ever need to overclock it as no amount of overclocking will make it supersede a discrete graphics card for gaming.  So, yay, you can overclock the graphics processor, but meh, who cares?

We’re not so bearish on IPT. Much like those key fobs that people carry that generate random numbers, the capability now exists in a consumer chipset and processor. Up until now though, IPT has only been offered in the Q67 chipset aimed at boring corporate boxes. With Z68, IPT could hopefully mitigate some hacking and identity theft situations. Valve has already said that it will support IPT to help prevent account hijacking.

Of more interest to consumers is the ability to switch between integrated and discrete graphics. On P67, even though all Sandy Bridge chips included a fairly powerful graphics processor, end users had no access to it because the P67 chipset had no output — those were only available on the H and Q-series chipsets. With Z68, Intel now includes the Flexible Display Interface that will connect from the processor graphics to the Z68 chipset. Some boards that we’ve seen, such as the Asus P8Z68-V Pro include both DVI and HDMI ports so you can run just integrated if you want to.

But here’s where it gets interesting – board vendors are also including Lucidlogix’s Virtu. Virtu lets you virtualize either of the GPUs. There are two modes available: i-mode and d-mode.

In i-mode, you set the board to initialize the integrated graphics port first in the UEFI. Then you connect your monitor to the motherboard’s graphics port. You’ll need to install graphics drivers for both the Intel integrated part and the discrete part. For our testing, we used a GeForce GTX 580 card. Lucid actually says the best results will come from an ATI card, but we opted for the nVidia card to see how much of a monkey wrench we could throw in it. For a just out of beta product it surprisingly worked well. Would we run it in this mode? Probably not, at least at this point.

When running in i-mode, you are primarily using your integrated graphics and only kicking on the discrete card for gaming. The main sticking point here is that you’ll need Lucid to create profiles for any game that you run in the Virtu mode. Most gamers that we know can’t wait that long when a new game is released. The other issue is power savings. As one of the key points of i-mode you'll likely see only moderate to minimal savings. That’s because unlike a mobile solution’s switchable graphics, the discrete card doesn’t completely power down. Even idling, today’s beefy graphics cards still drink too much power. We’d prefer it if vendors could find a way to power down the discrete card when not in use. I-mode is also currently incompatible with dual-GPU cards and SLI too.

We used an Asus P8Z68-V Pro board for our testing

The Asus P8PZ68-V Pro will also support a similar technology from Nvidia called Synergy. At press time, it wasn’t ready for consumption but leaks on the web indicate the technology will be free (Lucid charges board vendors for Virtu) and it will leverage the profiles Nvidia has already developed for its Optimus technology on laptops.

Virtu has a second mode available that’s likely to be more handy: d-mode. In this mode, you set the UEFI to boot to initialize the PCI-E graphics adapter first and hook the monitor up to a port on the graphics card. In this mode, the discrete graphics card is in control and any game you run will run without the need for profiles to be created by Lucid. So what would ever use d-mode for? To access the Quick Sync technology in Sandy Bridge. Yes, Sandy Bridge’s graphics performance will never best a serious GPU, but believe it or not, the transistors that Intel has dedicated in Sandy Bridge for encoding and transcoding are mean mothers. How mean?

We took an Asus P8Z68-V Pro board, plugged in a Core i7-2600K, 8GB of DDR3/1333, a 1TB Western Digital Black drive and a GeForce GTX 580 card. Running in d-mode, we used CyberLink’s MediaEspresso 6.5 to transcode a single VOB file to a generic WMV file suitable for playback on an HTC smart phone. Using the GeForce GTX 580 card took 142 seconds. We then used Virtu to allow us to access the QuickSync mode on the Core i7-2600K chip which took 109 seconds. That’s about a 30 percent faster for the integrated grapics. Now think about a transcode that would take three hours. Would you rather use the GeForce GTX 580 or the Core i7-2600K’s Quick Sync?

Lucid’s Virtu control panel lets you switch on the fly between the discrete card or the onboard graphics.

The coolest new feature of the Z68 isn’t switchable graphics or anti-hacking features though, it’s going to be the Smart Response Technology that’s built into the, umm, Rapid Storage Technology drivers (Intel, who seriously comes up with these names?).

SRT allows the Z68 chipset to use an SSD to cache often used data from a hard disk. This, in theory, offers up to a 4x improvement in performance over a hard disk drive alone.

Setting up SRT

Normally, setting up new technology is pretty self explanatory. That’s not the case with the SRT. With Intel’s new 20GB Larsen Creek in one hand, and a 1TB 7,200 RPM Western Digital Caviar Black drive in the other, we weren’t sure how to set up SRT initially. The process is actually quite simple. All you have to do is attach both drives to the board – an Asus P8Z68 V Pro board. Boot the system and go into the UEFI and set the Intel controller from its default of AHCI to RAID. Now boot to your install disc and install Windows 7 to the hard drive as normal. You can’t enable the SRT until after you’ve installed all of the drivers. Once you do, you simply go to the RST driver and click on the Accelerate button.

From there, you select the drive you want to use as the cache drive, pick the disk you want to accelerate (normally C:) and then pick the mode: Either enhanced or maximized.

Enhanced is the safer of the two and ensures that all data is written to the HDD. In this mode, read performance is improved but write performance will be no better than the hard drive in use. Think of Maximized mode as write caching. Data is written to the SSD first and then synced to the hard disk as time allows. If there is a power outage, or a blue screen before all of the data can be synced – poof your data is gone. The maximized mode, however, does offer write times close to what the SSD should be able to write to.

Ideally, SRT is intended for people who can’t afford massive 240GB SSDs but want “SSD-like” performance. Using SRT, these folks can buy smaller SSDs and get that SSD-feel. In fact, Intel is hawking its new Larsen Creek SSD to these people. A 20GB SSD, the 311-series drive offers reads up to 190MB/s and writes up to 100MB/s all for $110.

For our testing, we used the aforementioned Asus P8Z68-V Pro board outfitted with a Core i7-2600K, 8GB of DDR3/1333, a GeForce GTX 580 and a 1TB Western Digital Caviar Black. We used 64-bit Windows 7 Professional with SP1 installed and latest Nvidia drivers available. We ran PC Mark Vantage 64-bit and CrystalDisk 3.01 with just the 1TB drive initially. We then turned on SRT in maximized mode using the 20GB Larsen Creek Intel SSD and reran both tests.

We didn’t want to be gated by the Intel drive so we also used OCZ’s Kick-Ass 240GB Vertex 3 in the mix. We ran the same tests using the Vertex 3 in Maximized mode and Enhanced mode. Since the Vertex 3 is SATA 6Gb/s, we also made sure it was running off of the Intel PCH’s SATA 6GB/s controller. One thing to know: the SRT mode has a maximum cache size of 64GB. The rest  of the space on the SSD can then be partitioned into a separate drive. For this test, we opted for the 64GB partition which could contribute to the performance difference. We suspect, however, that’s really the exemplary performance of the Vertex 3 that’s really to be credited.

The results speak for themselves.

The hard drive trace tests in PC Mark Vantage as well as the synthetic tests in CrystalMark show that the SRT mode is pretty phenomenal. The best results came from the very fast and very expensive OCZ Vertex 3. But even the $110 Intel SSD yields very good results. You can’t benchmark “OS feel” but we can tell you that using the 1TB 7,200 RPM drive as our primary device – even with 8GB of RAM and the Core i7-2600K – was teeth gnashing slow. We’ve just forgotten how slow hard drives are compared to SSDs. With SRT and the Intel drive in place, the drive noise was greatly reduced as we no longer had to listen to the heads clattering all over the platter and it “felt” faster.

To get a feel for how much you give up in maximized vs. enhanced modes, we ran the Vertex 3 in both modes. As expected, generally the write speeds suffer greatly in enhanced mode while read speeds are mostly intact.