Date: June 5, 2006
Author(s): Rob Williams
If you are looking to upgrade to a dual core rig, but want to do it cheaply, then Intels 8** series may be for you. We are taking a look at the 820 D, a 2.8GHz chip with 2*1MB cache. What makes it a great value is the fact that it overclocks to almost 4GHz on air. But compared to other similar chips on the market, is this one worth your cash?
It wasn’t too long ago that we were all asking the question, “Should I go dual core?” Times have proven that dual core is the future and has clear benefits, which turned that question into, “Which dual core should I get?” Even if you are buying the lowest dual core on the scale, you are essentially getting twice the computing power than from a single chip. As more and more applications take advantage of dual cores, it makes the buying decision easier. Of course, the ability to multi-task without slowing your OS down is another huge benefit.
Intel and AMD both have a great selection of dual core chips, for all wallet sizes. Intel dual cores almost always cost less than AMD’s, which makes them look more attractive. Generally speaking though, tests have proven AMD dual cores better for gaming on high-end systems. But, if you are looking for a cheap solution to get yourself a dual core rig built, then Intels offerings are worth looking at.
Today’s processor in question is the 820 D. It was the bottom of the barrel when it came to Intel DC’s, until the lower clocked 805 D came out. But before we get into specifics and feature comparisons, let’s delve a bit deeper into why dual cores are beneficial to you.
PCs that have two CPU cores are not necessarily new, although it is for the consumer market. A few years ago, you would need two separate CPU’s in the same machine. You can now have the same benefit but with only one CPU. There was a lot of speculation when dual cores came to be, but there’s no denying just how beneficial they can be. Because you essentially have two CPU cores under the same IHS, multitasking proves less groggy with larger applications. As you open new instances of applications, the CPU will decide which core to use. The goal is to allow you to do more on your PC than with a single core, without having a sluggish experience.
Even though you have a dual core, things can still slow down your PC, but that primarily only happens with multiple intensive applications that are stressing the same components at once, such as ram or your hard drive. As a great example though, you could play a game and convert a video file at the same time and feel virtually no lag or slowdown. Try doing that on a single core and it will not be a fun experience. Because of these possibilities, it’s no wonder why dual cores are growing in popularity.
If we compare our 820 D to Intels other dual cores, we immediately see it’s missing a few key features. Considering this is a budget processor, it does not include EIST, or Enhanced Intel SpeedStep Technology. Also missing is the Intel Virtualization Technology. That specific feature is not available on any of the 8** series however, rather only the 9**.
At this point in time, when compared to Intels other dual cores, the 820 D seems like an unwise choice. If you were to consider the 805 D instead of the 820 D, there are a few things to consider. Most notably, the 805 D only has a 533FSB and is clocked 140MHz slower than the 820 D. Though the clock speed won’t necessarily make a huge difference, the FSB speed will.
The 920 D though, and the 9** series in general, has some larger differences. While the 820 D has only a 2*1MB level 2 cache, the 920 D has 2*2MB. This would not likely make a huge difference in most everyday tasks, but it should be evident through synthetic benchmarks. The largest different though, is the process the cpu is based on. The 820 D is built on the ‘last-gen’ 90nm while the 920 D is based on 65nm. What this generally results in is better overclocking ability, and cooler temperatures. Besides these points though, both the 820 D and 920 D share the same clock speed and FSB.
Below is a quick sum-up of all the Intel Pentium D’s currently on the market. Celeron D and EE Dual Cores are also available, but we won’t be getting into comparisons with them today.
I grabbed the prices from a popular e-tailer to get a grasp of the pricing on the cpu’s as a whole. The 820 D currently retails for $175US. At that price point, it would be well worth it to throw in another $25 to upgrade to the 930 D which offers a 65nm process, 2*2MB cache and 200MHz on top of what the 820 D offers. Oddly enough, the 930 D is more cost effective than the 920 D. You don’t see that happen too often.
Because the 820 D carries a ‘budget’ moniker, overclocking was going to be interesting. If you buy a high-end chip, you already know to expect good overclocking ability. When you take the low road though, it’s always hit or miss. My goal originally was to achieve a 4.0GHz, as unlikely as that may seem. Up until last month, the CPU was cooled using Corsairs Nautilus 500. That setup is how I acquired all my found overclocks. However, I have since moved the N500 into another PC, so the Zalman 9500AT is the current cooling in place.
I chose to skip overclocking using the stock cooler, because temps were far too high. Of course if you do plan on doing any overclocking, you should almost always upgrade to an aftermarket cooler.
Using the water cooling, the max overclock I have achieved was 4.2GHz. Though this was rather unstable, and put me in an odd predicament. After some troubleshooting, I had found that one of the cores was not as overclocking friendly as the other. I came to this conclusion because Windows would not boot at 4.2GHz, but if I changed it to 4.2GHz when Windows was already loaded I would have no problems. While in Windows at 4.2GHz, the CPU proved 10 hours Prime95 stable.
Here’s another oddity. When I ran that 10 hour prime run, it was only on the single core. The second core was left completely alone, because I wanted to first see if 4.2GHz was stable using the CPU as a single core. Once that passed fine, I opened up another instance to run on the second core, to which I immediately had errors. Stranger still, is that if I chose either core to stress test, one at a time, I would run into no errors. It was only when I stressed them both at the same time that the second core would error. The first core has never failed in any of my overclocking tests.
After some more investigation, I came to the conclusion that the main problem had to do with the 64-bit part of the second core. At 4.0GHz, Windows x32 booted no problem, over and over. But, trying to boot into Windows x64 at the same clock would fail everytime. It would either give an abrupt BSOD or restart during the Windows boot screen.
3.92GHZ seemed to be the ultimate stable overclock, for both versions of Windows. I will consider that my top overclock, although yours may vary if you get a CPU with two friendly cores. Regardless of the 4.0GHz+ problems, I will gladly take 3.92GHz on a sub $200 dual core any day.
Here are the overclocked settings I will be benching with. All of these have proved 100% stable. Of course this could have been eight separate overclocks, but there was no need. 3.36GHz was completely stable using the stock cooler, however I didn’t go beyond that until I had the better cooling solution. Using the Zalman 9500AT, the CPU was completely stable up to 3.74GHz. 3.92GHz is hit and miss though, in a couple regards. First, at that clock speed, Rosetta had run for 44 straight hours without a single issue. That was with the CPU at 73°C for almost two days straight. However, when it came to a couple benchmarks, we will see that the increasing heat on the CPU held back from some great scores.
The primary reason that the CPU could reach 73°C to begin with is due to my rooms ambient heat. Even on a cool day, the room hovers between 80°F and 90°F. It doesn’t prove to be the ideal testing environment, but it could not be helped. That said, if you have a computer room that’s kept at comfortable temperatures, your results may prove better than mine.
Throughout all of our benchmarks, regardless of what we are reviewing, testing is done in a clean, stand-alone version of Windows XP Professional with SP2. Prior to testing, these conditions are met:
If you are interested in using the same benchmarks as us, feel free to visit the developers website:
The testing rig used for today’s benchmarking is as follows:
Since this is the first Intel chip that we’ve reviewed, we don’t have any means to compare to other similar CPU’s.
Sandra is an extensive benchmarking tool used to stress and test multiple components in your system. The Arithmetic test is a hardcore math test that Intels always prove to be great at. The Multimedia test is another one that usually sways in Intels direction due to the better SSE units, and is primarily based on how fast your CPU will render multimedia data.
There are not that many applications that will use both of your CPU cores at once, but many benchmarking tools do. Such as Sandra, these scores are almost twice as high as my single core AMD can pull off. At 3.92GHz, the air cooler is having a hard time keeping the CPU at a normal temperature, which results in lower scores.. namely in the multimedia test. The 3.74GHz scores are superb though, for not even being on water.
Everest takes similar tests as Sandras, but work quite differently. CPU Queen is a large computation using a virtual chessboard. It uses the CPU power to calculate an outcome. PhotoWorxx is just as you’d expect… a photo processor. Of course you don’t see anything that’s going on, because it’s just calculating, not outputting. Julia and Mandel both measure single precision FP’s, and is based on the Assembly coding language. What sets them apart is that Julia calculates in 32-bit while Mandel is 64-bit.
The CPU tests in Everest also heavily benefit from dual cores. The Queen and PhotoWorxx scores are as you would expect for these speeds, but the Julia and Mandel are actually far higher than I had thought they would be. Intels math processor to the rescue.
Sciencemark is a superb CPU and memory benchmarking tool. It gives you total freedom over your tests and is more in-depth than most other tools out there. The Molecular Dynamics test is very complex, but overall the lower the time the better. The Cipher test takes a popular encryption scheme and decrypts using it. Once again, the lower the time, the better.
One example here I like to look at, is the difference in the Moldyn between stock and 3.74GHz. It skims a full 27 seconds off the time, for an overclock that can be done on air cooling. The Cipher scales quite well to our overclocks also.
Cinebench is a very unique benchmark, because it’s built atop a popular 3D modeling program called Cinema 4D. There are two separate tests that you can use, but we focus on the CPU intensive one. These CPU tests are divided into 1 Core and multi Core benchmarks. It renders an image using one core, then multiple, or in our case, two. The higher the score, the better.
SuperPI is a simple concept, but is a great way to see how fast your CPU can process a ‘simple’ mathematical calculation. In this case, it calculates PI to a specified amount of digits past the decimal. Since PI is virtually never-ending, it proves a perfect test. We choose to run 1M, 4M and 8M tests.
I was hoping that the CPU could break 30s SuperPI, but it couldn’t. On water at 4.2GHz, it came very close. But due to the second weak core, I believe it was holding back some of it’s power. For a ‘budget’ dual core, these times are very pleasing though.
Both of the following programs are great for stressing your CPU. Usually if your CPU is borderline stable, PC Mark will end up crashing your computer. Needless to say, it’s one of the tools I use to assure myself that I have a steady overclock.
3D Mark 2006s CPU test is far different than PC Marks, because it’s more gaming specific. While the 3D action is held back to 2FPS, it will compute the entire scene using your CPU. Obviously, the better the CPU you have, the better the score. Like our Sandra score, the 3D Mark 06 score at the max overclock was found slower than 3.74GHz, due to the heat.
So it’s 2006, but I am using a program from 2001? 3D Mark 2001 proves to be an enthusiast favorite for benchmarking your CPU, because it’s amazingly CPU dependent. Since our video cards are far better today than they were back in 2001, the test benefits clearly from more CPU power, rather than GPU power. Yet again, the max overclock proved a lower score. With proper water cooling, this would not happen.
For our gaming benchmarks, I chose to use Half-Life 2 and Battlefield 2. While Half-Life 2 is incredibly CPU dependant, BFII is more GPU dependant. To test, I used each successful overclock, and then played the game for 5 straight minutes. The average FPS information was recorded using FRAPS. For HL2, I used the level d1_canals_08, and maxed out settings at 1280*1024. For BF2, I used the FuShe Pass level with default settings at 1024*768. While it could prove to make more sense to play through the game at the lowest resolution possible, I am more interested in seeing if there is any benefit at standard settings.
All of the runs were manually played, not time demos. Because of this, each run could prove different with varying instances of action on the screen.
This may seem like an odd bout, and it is. The 820 D is a dual core Intel and the 3200+ Venice is a single core AMD. Obviously, these results should not be used to see which is better, because obviously the dual core should win most of everything. Since the Venice is the only other chip I have on hand, it is still interesting to see how they will compare.
Since 3.92GHz was the max stable overclock I could do with the 820 D on air, I used my previous max air overclock with the Venice also, which was 2.7GHz. Here are those system specs. Both of these overclocks, though different architectures, are pretty close to one another.
I split all the results onto two separate graphs for easier reading. As mentioned earlier, you can easily see the benefits that dual core can offer. We can also see Intel specific benefits with the math processor in the Julia and Mandel tests. They are more than double what the AMD could muster.
Things start to get interesting with even more benchmarks. While the Intel beat out the AMD chip in the Moldyn test, it was vice versa for the Cipher. Since the AMD is single core, it could not do a 2x test in Cinebench, but even as a Single core, it beat out the Intel here. The AMD proved faster in each and every Super Pi test also.
Once again, these tests should not be taken -that- seriously. A more fair comparison would be comparing a single core equivalent to the 820 D, or even comparing the 820 D to an AMD X2 in the same price bracket. Still, the results proved interesting for what they are worth.
The 820 D proved to be a great CPU, especially given it’s cost. For $175US, you can have a great dual core cpu in your rig. I compared the 820 D to my Venice 3200+, which cost $270 when it was new. Just a year later, you get twice the power for less money. That being said though, it’s hard to justify a recommendation. Let’s take a look at the current price list once again.
If you were very budget conscious, then the 805 D may be a good choice for you. At $120US, it lacks some of the power that that 820 D provides, but has been known to overclock to 3.8GHz with water cooling. If you were to step up and purchase the next step up, you would have to check out some of the better options. While the 820 D currently retails for $175, the 930 D retails for $200.
For the extra $25 it would cost you, you get a faster chip with more cache, and based on a 65nm process. If you are planning to pick up the chip for overclocking and plan to use water, it would be almost foolish to not spend the extra for the substantially better CPU. One thing to also bear in mind, is that not all new Intel boards will support the 820 D. The ones that do, usually cost a little more. Generally, I have found that the less expensive motherboards support the bigger dual cores. So you will really want to shop around, because getting a board to support a 820 D will probably cost more than a 930 D and a supportive motherboard.
Times are good though. With Conroe right around the corner, we will be seeing the prices drop even further. In fact, rumor says that we should be seeing some drops in the coming days. But then the next question would be, should we just wait until then? With the Conroe performance benchmarks around the web, it’s hard to know for sure. If you are in need of a PC now, then I wouldn’t hesitate to pick up the 820 D or 930 D or anything in that general area. If you are able to wait, then do it. If Conroe lives up to the hype, it should be well worth the wait.
If you would like to discuss this review, feel free to head on into our related thread to throw your comment. You do not need to register in order to reply to our content threads.
Copyright © 2005-2020 Techgage Networks Inc. - All Rights Reserved.