Print this page
Published in Reviews

Intel QX9650 reviewed [updated]

by on29 October 2007

Image Image

Review: Penryn goes up to 4.30GHz with air cooling


German version available here.

Two weeks and some days ago we informed you that we received the new QX9650 Penryn-based Yorkfield CPU from Intel. Today, we compare the new quad-core with the old QX6850 Kentsfield and check out what the new CPU has to offer.

Image

We will not go into the technical details too extensively. The new CPU is produced using the new 45nm process with a die-size of 214mm² and 820 million transistors. The old Kentsfield is 282mm² with 582 million transistors. Obviously, the increased transistor count comes from its massive 12MB cache, as each dual-core CPU die has its own shared 6MB. The new cache is now 24-way associative, compared to the 16-way associative cache of the older models.

Intel also added an extension to the SSE instruction set called "SSE4" which will help video algorithms work much faster; that is, of course, only if supported by the software. The most important changes are the power-saving functions inside the CPU. While most mainboard BIOSes will not support VCore reduction of the CPU when overclocked, the new family can power down on its own. If the CPU is idle, it will go into idle mode regardless of the BIOS. 





Testbed:

Motherboard:
ASUS Blitz Exterme (provided by ASUS)
Intel P35/iCH9R

CPU:
Intel Core 2 Duo E6700 (provided by Intel)
Intel Core 2 Quad QX6850 (provided by Intel)
Intel Core 2 Quad QX9650 (provided by Intel)

CPU-Cooler:
Scythe Andy Samurai Master (provided by Scythe-Europe)

Memory:
Patriot 2GB Kit PC3-10666U PDC32G1333LLK (provided by Patriot)
CL7-7-7-20-CR2T up to 1.90V

Graphics Card:
AMD ATI Radeon X1950XTX (provided by AMD)

Power supply:
Silverstone Element SF50EF-Plus (provided by Silverstone)

Hard disk:
Western Digital WD4000KD (provided by Ditech)

Case fans:
SilenX iXtrema Pro 14dB(A) (provided by PC-Cooling.at)
Scythe DFS122512LS





In our QX6850 review we had no time to do any overclocking. This time, we include the Kentsfield CPU in our tests to show you the differences.

First, we have to caution anyone who wants to overclock. If you own a Kentsfield the maximal recommended VCore is between 1.4000V and 1.4250V. Also, you need a very good temperature regulating air-cooler. We don't have a test-rig with a water-cooler, but if you do you can expect to go even higher than we did. Not any CPU will go that high, but it should give you an idea how far you can go. Always watch the temperature of the cores, as if they go over 70°C you are in big trouble.

Overclocking FSB:

QX6850:
We have heard and read statements that the QX6850 is not such a good performer on the FSB side. Actually, we were surprised because we managed to go beyond 400MHz FSB (1600MHz) with ease.

Image Image


QX9650:
After the nice results achieved with the QX6850, we thought that the QX9650 would go much higher, but we only got a mere 10MHz more.

Image Image





Increasing the CPU clock:

QX6850:

The result is quite good, but Kentsfield is like a little reactor:  it gets hot very easily even when it's idling, so hardcore overclockers need to use water-cooling or other similar methods to keep the CPU cool. It would be possible to squeeze some more MHz out of our CPU but we stayed on the conservative side, so our result was "only" 3750MHz, which is still quite nice.


Image Image

QX9650:

The Penryn, on the other hand, is a very cool customer. You can go up to 1.6000V VCore and it will not hurt it, which is quite amazing. We could go up to 4300MHz and did some tests, but we did not check if this configuration was prime stable. For our benching we decided to use a common frequency of 4GHz.


Image Image




Please note that different bios revisions may give different results. All benches are done with AUTO settings, without altering any BIOS option besides CPU VCore and major CL settings. DDR3 speed is always 1066MHz or as near to 1066MHz as possible.

Cinebench R10:

QX6850@3000MHz:
Image

QX9650@3000MHz:
Image

QX9650@4000MHz:
Image

As you can clearly see the QX9650 is 7.5 percent faster. The bigger cache does not count for this increase, we believe that is caused by support for SSE4.





Gordian Knot/XviD 1.1.3:

For our Gordian Knot testing we took a PAL episode of "Babylon 5" with a length of 41 minutes, 57 seconds and 8 frames.

We tried to "emulate" the most common usage of Gordian Knot:

1st: We have a perfect master, so we only de-interlace the content and resize it, without any other manipulations. We marked this as "fast."

2nd: You get bad mastering on many DVDs, especially "old" stuff or when the studios are in a hurry for the release. In this case you may want to improve the picture quality, which is done by filtering the content. You can choose from many filters for any purposes you can think of. We only used the most common "undot," "FluxSmooth" and "MSharpen." Of course, we also de-interlaced, filters were done before any resizing took place (which is slower). We marked this as "slow."

If you need more information about filters, we recommend reading the doom9.org forum

Image


x264:

With this review we used x264 encode for an additional video encoding bench. x264 is a H.264/AVC codec which supports four threads and is available for free. We used the same "slow" settings like XviD. H.264 gives you a huge advantage in compressing size, and you need much less bitrate to achieve the same or better quality compared to XviD. While the H.264/AVC is much more advanced compared to MPEG2 encoders, now the CPU-power is available to do the encoding in an acceptable time.

You can clearly see that a quad-core brings an massive advantage in encoding speed; first passes seems to run on only two cores, but the second pass takes advantage of all four cores. With a quad-core you can nearly reach the speed of XivD, which can only use two cores for the time being.

The codec is open source, still in heavy development and you can grab it here.

Image

Update: While testing a Q6600 we noticed discrepancies with the results of the QX6850. We rerun the tests and updated the chart. We apologize for the inconvenience.

LameMT:

We used the same episode for our MP3-testing. We don't recommend using MP3 for encoding, because AC3 can do the job better. The audio is almost 42 minutes long and it gives us approximately the length of most of album. A measurement in seconds, as many sites use, is useless, because the differences are too small.

We used the built-in play/CPU ratio, which means the CPU is encoding x-times faster then the track-length. Fast memory does not play an important role here. They will be produced with any other L.A.M.E. version, because only LameMT can do more than one thread and take advantage of a second dice. We used this setting: lamemt --vbr-new -q 2 -V 2 -m j --strictly-enforce-ISO --resample 48.

A quad-core does not improve the speed, because it can only utilize two cores.

Image





 
3DMark2003:

QX6850 @ 3000MHz:
Image

QX9650 @ 3000MHz:
Image

QX9650 @ 4000MHz:
Image

About 300 Points more is not very impressive. The CPU optimization and the bigger cache does not help on this GPU bounded test.




3DMark2006:

QX6850 @ 3000MHz:
Image

QX9650 @ 3000MHz:
Image

QX9650 @ 4000MHz:
Image

A 90 point overall score difference between QX6850 and QX9650 is not very impressive, but you see the CPU-score increased quite nicely. When the QX9650 is overclocked you get an impressive boost of the CPU-score, but only a mere 375 more points; this is expected, because the bench test focused mostly on the graphics card. 




Power Consumption:

And now we come to the most interesting part. Intel claimed it had improved the power-saving on the CPU and yes, they did, with impressive results. While our board does not decrease the voltage of the CPU when overclocked in idle-mode, the Penryn does not care. It shuts as much down as possible, a quad-core is using less (!) power than compared to an overclocked dual-core E6700.

Image


Under load the picture doesn't change one bit. The quad-core uses only as much energy as the the dual-core; this is really impressive. The full load graph is only an indication, because it's hard to get all four cores at full load. We used two cores for x264 encoding, while Super-Pi and 3DMark06 were allocated to the other two cores. It's unclear why Intel stated a 130W TDP, as the CPU not overclocked is capable of much less.

Image




Conclusion:

While the bigger cache and SSE4 do not help any application, it is a huge step forward for Intel. Now you can use a quad-core and not have to worry about your electricity bill. If the cores are idle, they stay idle, and consume much less power than any Core2 so far. Even under load it does not use more energy than  our E6700. In short, it's fast, it's cool and it's energy efficient.

For gamers and "normal" users a quad core simply won't provide any advantage compared to any dual-core CPU, but in the near future some games will make use of them; for example, Microsoft Flight Simulator X already does.

Data center operations and rendering farms will take advantage of the new family and we believe sooner, rather than later. Rendering farms will have their software updated for SSE4, which will increase in speed and save "Hollywood" a lot of cash. The same is true for the energy efficiency for data centers, as 45W x 1000 machines saves a lot of energy costs, too.

Now it is AMD's turn. We will soon see if the new K10 can compete against Intel, especially in speed and energy savings.

We recommend the Penryn series of processors over all other currently available CPUs, primarily due to its high performance and low power consumption, which is a direct result of the new 45nm process and the built-in energy management. It will be rather difficult to get one right now, as the first batches go into Xeon production and the few desktop CPUs will be used by Dell, HP and some other high-end manufacturers. We expect mass availability, especially of the smaller models, in February 2008 but maybe Intel will surprise us all with faster and wider availibility of the Penryn.

The price tag ? It is the usual US$999, but we think it will not deter professionals who will appreciate all the benefits of the new architecture and utilize its full potential. This is simply the best CPU you can buy today and it will even overclock to 4.3GHz with an air cooler. What more can you ask for?

Image

Last modified on 01 November 2007
Rate this item
(0 votes)