by Ilya Gavrichenkov
01/07/2008 | 12:07 AM
Not so long ago we thought that in early 2008 we will be focusing on comparing the new AMD Phenom processors against the refreshed Intel Penryn manufactured with 45nm technological process. However, these expectations didn’t come true, and both – AMD and Intel – should be blamed for that.
It is true, at this time AMD cannot deliver mass quad-core processors working at competitive frequencies. The currently available Phenom models lose even to previous generation quad-core Intel processors, not to mention the more advanced CPUs. It is quite logical that Intel doesn’t have any significant stimulus to refresh their quad-core processor line-up, because there are simply no worthy competitors to the pretty successful Core 2 Quad on old 65nm cores these days. That is why the launch of new Core 2 Quad processors known as Yorkfield has been postponed for an indefinite period of time, at least until February or March 2008. And although Intel has found an excuse – an alleged problem in the upcoming processors caused by EMI in 1333MHz front side bus when these CPUs are used in hypothetical mainboards with 4-layer PCB design – it doesn’t sound convincing at all.
As for us, we have to state to our disappointment that it doesn’t make sense to compare Phenom against Penryn, because the former is uncompetitive, and the latter is still illusive and remains only an upcoming solution for the time being.
Nevertheless, there are more than enough interesting topics for discussion in the today’s processor market. Although Intel decided to postpone the launch of their quad-core processors on 45nm cores, the Core 2 Duo processor lineup will be refreshed with a few new models. They are going to announce three new processor models with Wolfdale codename within the next few days. They will be Core 2 Duo E8500, E8400 and E8200. These CPUs are based on the revised core manufactured with 45nm process and belong to the same Penryn family as the postponed Yorkfield CPUs. We certainly can’t disregard the arrival of mass Wolfdale processors, which promise to raise the performance bar for Intel’s dual-core solutions to a totally new qualitative level. They feature higher clock speeds, larger L2 cache and a number of other improvements. And the most pleasing thing about them is their cost, set at the same level as that of older Core 2 Duo solutions.
So, in the second half of January 2008 Intel is going to massively update their dual-core processor lineup in $160-$260 price range. This particular event became the main topic of our today’s article that will dwell on the new promising Intel processors and the changes they will bring to the mainstream desktop market.
So, Wolfdale is the codename for dual-core Penryn processors. Just like the postponed quad-core Yorkfield CPUs, Wolfdale processors are manufactured with 45nm technological process. Moreover, Yorkfield and Wolfdale are based on absolutely identical semiconductor dies: Yorkfield is traditionally made of two dual-core Wolfdale dies put together into a single processor package. So, Wolfdale can be regarded as a basic building block for the entire Penryn family, and that is why it is very interesting to us.
Wolfdale processor core is 107sq.mm big and consists of 410 million transistors. These numbers give us to understand that Wolfdale has been changed significantly since the times of its predecessor - 65nm Conroe with only 291 million transistors. You can even see it from the photographs of Wolfdale and Conroe cores: the positioning of their functional units has slightly changed.
Wolfdale on the left and Conroe on the right (the images aren't scaled).
So, Wolfdale core is not just a smaller Conroe core resulting from the transition to finer production technology. Intel engineers made a number of innovations in the new processors (you can read more about the Penryn processor family in our article called Second Iteration of Core Micro-Architecture: Intel Core 2 Extreme QX9650 CPU Review).
Most of these innovations should increase the processor performance. Wolfdale’s biggest advantage is the larger 6MB shared L2 cache memory. Moreover, Wolfdale processors support SSE4.1 instructions set including 47 new commands that can speed up 3D graphics and video processing alongside with scientific calculations in case of appropriate applications optimization.
A few changes have been made to the execution units, too. Wolfdale processors have now acquired Fast Radix-16 Divider unit that speeds up division and square root calculations. The new processor also features Super Shuffle Engine mechanism accelerating those SSE instructions that require bit-by-bit shuffling.
The above listed improvements and a few additional ones introduced in Wolfdale processors ensure that these new processors will work faster than the older Conroe CPUs running at the same clock speeds. However, we are not talking about any significant advantage here. Wolfdale offers only minor cosmetic refresh of the Core micro-architecture, that will change dramatically only in the upcoming Intel Nehalem CPUs scheduled to arrive in the end of 2008.
The most important thing about Wolfdale processors is certainly the new 45nm manufacturing process that allowed Intel not only to significantly increase the number of transistors per die without making the die any bigger in size. New production technology uses a new hafnium based material with a property called high-k, for the transistor gate dielectric, and a new combination of metal materials for the transistor gate electrode. This allows increasing processor clock frequencies further without raising their heat dissipation and power consumption. That is why these new processors will be of specific interest to overclockers, who will definitely be able to use them to set new overclocking records.
Summing up everything we have just said, let’s compare Wolfdale and Conroe technical specifications side by side:
Now that we have taken a quick look at the main Wolfdale features, let’s meet the specific members of this processor family that will become available in the nearest future.
The Wolfdale processor family announced these days that is based on 45nm cores will initially include three Core 2 Duo processor models: E8500, E8400 and E8200 with 3.16GHz, 3.0GHz and 2.66GHz respective clock speeds. Moreover, there will also be a CPU with E8190 model name, similar to Core 2 Duo E8200, but without the virtualization technology support. Later on the fifth CPU will join this lineup: Core 2 Duo E8300 with 2.83GHz clock speed. However, it should happen in Q2 2008 at the earliest. The table below offers a better look at the mass Core 2 Duo processors with 45nm cores:
The pricing info on the new Intel processors is a very important addition to the technical briefs given in this table:
It is very nice to see that Intel continues to pursue the price policy many of the computer users approve of: when the new processors are being offered at the same price point as the older ones, so that the old ones get ousted from the market in an evolutionary way. This time, Core 2 Duo E8500 came to replace Core 2 Duo E6850, Core 2 Duo E8400 steps in for the Core 2 Duo E6750, and Core 2 Duo E6550 gives way to Core 2 Duo E8200. In other words, those of you who will be shopping for new dual-core processors within the next few days will already be able to buy more advanced and higher-speed solutions at the same old price.
Let’s take a closer look at the CPUs codenamed Wolfdale:
Wolfdale (left), Conroe (right)
As you can see from the picture, the CPus on 45nm core look almost the same as their 65nm predecessors.
Wolfdale (left), Conroe (right)
Nevertheless, we can see different external electronic components at the bottom of the CPUs from two different generations.
CPU-Z diagnostic utility is already very well familiar with the new processors. It has no problems recognizing Core 2 Duo E8500, E8400 and E8200 CPUs.
Note that our test samples are based not on the first core revision – C0, and it will be the one to actually be used in mass production units.
We would only like to add one thing to the info you can get from the screenshot above. Wolfdale processors can support fractional multipliers, which allowed Intel to make the list of supported clock frequencies even richer. This is exactly what we see in case of Core 2 Duo E8500: it supports 9.5x clock frequency multiplier. Note that the mainboard BIOS also needs to support fractional multipliers for a CPU like that to function properly. However, all leading mainboard maker should release corresponding BIOS updates very soon, so there is no need to worry.
To check out the performance of the new Core 2 Duo E8500, E8400 and E8200 processors and compare their speed against that of their predecessors, we put together a few systems including the following hardware components:
I would like to specifically stress that the Asus P5E mainboard with BIOS version 0502 we used to test Wolfdale processors does support them properly and allows adjusting their clock frequency multiplier with 0.5x increment.
First of all we were eager to check out the practical value of all innovations introduced in the new Wolfdale CPUs. Therefore we compared the performance of processors with Core micro-architecture based on the old and the new core working at the same frequencies. For this test we used Core 2 Duo E6850 and Core 2 Duo E8400 processors: they both work at 3.0GHz speed and use the same 1333MHz bus.
The obtained results can hardly be disappointing. New Wolfdale processors turn out much faster than their predecessors even when working at identical frequencies. The average performance gain in this case is 6%, but in some applications it may be much higher. I would also like to add that Wolfdale processors can work at higher frequencies, which indicates clearly that Intel prepared a potential hit.
More detailed results analysis shows that larger L2 cache turns out the determinative factor for higher performance of the new processors. As we may see, the performance gain is especially high in those applications that are sensitive to L2 cache size. For example, Wolfdale’s advantage over Conroe in games reaches 11%, and on average hits 7.2%.
Introduction of Fast Radix-16 Divider also played an important part here: the performance gain in computational tasks, such as final rendering, for instance, also turns out above average. The same significant performance improvement can be observed when we work with H.264 codec from Mainconcept and during video processing in After Effects CS3: it must be thanks to the Super Shuffle Engine unit speeding up some of the SSE instructions processing.
As for the SSE4.1 instructions support, from this prospective software developers aren’t ready for the new processors yet, although these instructions are potentially very demanded. Therefore, we can’t make any specific conclusions here. In fact, only TMPGenc codec may currently boast new instructions support, as well as DivX in experimental mode. Moreover, if you enable Experimental SSE4 full search in DivX, the performance will drop, which doesn’t allow us to state that this codec is nicely optimized for SSE4.1. Nevertheless, we expect to see the applications that would be able to efficiently employ the new SSE4.1 instructions, and then the value of the new SIMD set will become evident. For example, as far as we know, the corresponding changes should be soon made in the upcoming MainConcept H.264 Encoder, Pinnade Studio Plus and Sony Vegas.
The SYSmark 2007 benchmark we picked uses typical work scenarios to test the systems performance in the most popular real applications.
SYSmark 2007 reveals about 4% average advantages of the new Wolfdale processors over Conroe working at the same clock frequencies. However, since Intel increased the working frequency of its new processors, the top Wolfdale model turns out 7% faster than the top Conroe. As for the price, the CPUs from different generations sell at the same price point, according to Intel’s official price-list.
If we take a look at the detailed SYSmark 2007 results, we will see that the new processor gain most performance in a scenario preparing a learning web-site with various media content. It involves the following applications: Adobe Illustrator CS2, Adobe Photoshop CS2, Macromedia Flash 8 and Microsoft PowerPoint 2003. The smallest performance difference between Core 2 Duo on 45nm core and 65nm core takes place during video processing that uses Adobe After Effects 7, Adobe Illustrator CS2, Adobe Photoshop CS2, Microsoft Windows Media Encoder 9 and Sony Vegas 7.
Gamers should be very excited about the new Core 2 Duo E8000 processors. As we know, the gaming applications performance is very sensitive to changes in the cache memory size, which we witness here, too. In some games the youngest Wolfdale processor, Core 2 Duo E8200 manages even to outpace the former top dual-core Core 2 Duo CPU on 65nm core. The top dual-core AMD processor, Athlon 64 X2 6400+ that has never looked too good in games, now gets knocked out completely. It falls far behind even the slowest of the Wolfdales.
These turned out pretty expected results: the advantage of Core 2 Duo E8000 processor family over their predecessors represented by Core 2 Duo E6000 family is indisputable. Although the situation should change dramatically in the near future: codecs are among those applications that should gain a lot from the upcoming optimization for SSE4 instructions. So, we would like to refrain from drawing any final conclusions concerning Wolfdale’s performance in this type of tasks.
All in all, everything looks very similar to what we have already seen in the previous tests. Well-paralleled rendering algorithms benefit from the new core. Here I would also like to draw your attention to another curious fact that cannot be seen from the diagrams. The thing is the dual-core Core 2 Duo E8500 processor during final rendering has almost caught up with the youngest quad-core AMD CPU, Phenom 9500, although it may seem unbelievable. According to our tests, this AMD processor scores 5.61 in 3ds max 9, and 7114 in Cinebench R10.
For this part of our test session we selected another four widely spread applications, which do not fit into any of the previous parts. However, there is nothing principally new here either: Core 2 Duo E8500, E8400 and E8200 are indisputably faster than the solutions on 65nm cores working at the same frequency, and of course than those selling at the same price point.
Since the new 45nm manufacturing technology should definitely affect the electrical and thermal characteristics of the new processors, we decided to pay special attention to practical investigation of these parameters in the new CPUs.
First of all, we measured the operational temperature of our testing participants in idle mode and under high workload. During the tests we used the same Zalman CNPS9700 LED cooler for all CPUs. Enhanced Intel SpeedStep and Cool’n’Quiet 2.0 power-saving technologies were activated. By the way, Wolfdale processors, just like their predecessors, drop the frequency multiplier to 6x in case the workload is low.
CPUs were loaded using Prime95 25.5 utility. The temperature readings were taken with CoreTemp 0.96 utility. The obtained results are given in the table below:
As we have expected, CPUs on 45nm core are overall cooler than their predecessors on Core micro-architecture. However, the temperature different under 100% workload makes only 4-5 degrees. The thing is that Wolfdale processor core is smaller and features higher transistor density within a semiconductor die, which makes it a little harder for the heat flow to dissipate properly. That is why Wolfdale and Conroe heat up almost identically in idle mode. As for the relatively low temperature of AMD Athlon 64 X2 6400+, with twice as high TDP as that of Core 2 Duo actually, it can be explained by the not very best location of the on-die thermal sensor that is situated quite far from the hottest spots of the processor’s semiconductor die.
Everything we have just said indicates that CPU temperature measurements still offer pretty subjective judgment. Therefore, we decided to pay due attention to power consumption tests, too, that should fully reveal the advantages of the new 45nm core. During our experiments we measured the current going though the processor voltage regulator circuitry. This way we could estimate the actual CPU power consumption disregarding the losses in the voltage regulator circuitry.
The new processors manufactured with 45nm process performed more than impressively. In fact, we didn’t expect the results to be any different, as the new production technology allowed reducing the leakage current thanks to the innovative transistors with metal gate and high-k dielectric. As a result, the power consumption of the new Wolfdale processor under workload is equal to that of two 2-3 year-old processors in idle mode. Actually, Athlon 64 X2 results stress this dramatic difference between the processor generations even more, as its micro-architecture hasn’t been optimized for high performance-per-watt standards.
Now let’s move on to the most interesting part of our test session: Wolfdale overclocking experiments. We pin a lot of hopes on the new CPUs from this stand point, because 45nm technology and some other features, such as lower heat dissipation, allow us to expect the new processors to hit frequencies unattainable for the previous Core 2 Duo models.
We used the same test platform for our overclocking experiments as we did for our performance tests. The CPUs were cooled down with Zalman CNPS9700 LED cooler. The CPU stability during overclocking was tested with a 30-minute run of Prime95 25.5 utility.
First of all we tried overclocking our Core 2 Duo E8500 sample without raising its Vcore. In this case the CPU ran stably at 3.66GHz frequency.
Of course, it is a very good result. Core 2 Duo on 65nm cores could reach these frequencies only at increased Vcore. However, we wouldn’t stop here and continued our tests with increased core voltage setting.
Actually, our tests showed that Wolfdale responds very good to Vcore increase. But we didn’t aim for record-breaking speeds, so we only increased the processor Vcore in the BIOS Setup to 1.5V that resulted into the actual 1.42-1.46V taking into account Vdroop. This voltage increase is relatively harmless for a processor using an efficient air cooler and is acceptable for systems running 24/7. However, even in this case our Core 2 Duo E8500 didn’t disappoint us at all.
We managed to get this processor to run stably at 4.37GHz. No doubt: Wolfdale CPUs will become another overclockers’ favorite, because Conroe processors could hardly hit these speeds at extremely high voltages and with special cooing systems involved. By the way, our processor running at 4.37GHz retained quite acceptable thermal mode of only 70°C under workload.
All the above described experiments were performed without changing the clock frequency multiplier: it remained at 9.5x in all cases. That is why the maximum FSB frequency in previous experiments remained 460MHz. However, it would also be interesting to see how well Wolfdale processors can cope with high bus speeds. Therefore, we performed another experiment aimed at finding the maximum FSB frequency for our CPU, i.e. its FSB Wall.
As we found out, the maximum FSB speed for our processor was 540MHz. After that the system would lose stability. So, we can state the new processors have pretty high FSB Wall.
However, great Core 2 Duo E8500 overclocking results were slightly shadowed by the less impressive performance of the youngest model – Core 2 Duo E8200. Our Core 2 Duo E8200 test sample could only work at 3.2GHs at its nominal Vcore setting.
By raising Vcore we could hit higher frequency, but stalled at only 3.88GHz.
Of course, this is not a bad result at all, but nevertheless, Core 2 Duo E8200 yielded significantly to its elder brother, Core 2 Duo E8500. So, it turns out that Wolfdale overclocking success depends a lot on the CPU sample.
Well, everything seems pretty clear. Summing up everything we have just said, we can state that new dual-core Core 2 Duo E8500, E8400 and E8200 processors on 45nm cores are great from all stand points. They are faster than their predecessors working at the same clock speeds. Besides, their working frequencies are initially higher than those of previous Core 2 Duo CPUs. And taking into account that Intel is going to sell the new solutions for the price of Core 2 Duo E6850, E6750 and E6550, we can all get “free” performance improvement of 10-15% on new Intel dual-core CPUs.
Moreover, Core 2 Duo’s transition to new manufacturing technology provides additional bonuses to the users. First, they will support of the promising SSE4.1 instructions that will show their best in the future. Second, Wolfdale CPUs are extremely economical. Third, they overclock brilliantly and hence will become overclockers’ choice.
In other words, the second reincarnation of dual-core processors on Core micro-architecture turned out a definite success. The only upsetting thing about it is that the sales of these processors will definitely shake AMD’s positions in the market even more, because they cannot offer anything of the same performance level at the time. All dual-core processors from AMD are slower than the new Core 2 Duo E8000 series, which automatically ousts them from the $150+ price segment, where Intel’s dual-core solutions will now become the only player.