AMD Athlon XP3000+ 10th February 2003
Today we bring our readers a review of AMDs Athlon XP3000+ which uses the Barton core. The main improvement over the thoroughbred core is the increase of the level 2 cache from 256K to 512K. All other settings (FSB and clock speeds etc) remain the same as the XP2700+. This is quite useful for evaluating the performance of this processor as improvements will be down to the increased cache size and we should be able to see this clearly. Does the increased cache justify the extra 300 performance points? Read on to find out.
On the left is the Thoroughbred XP2700 and on the right is the new Barton XP3000. The Barton is slightly longer but to see why we need to look at the architectural differences.
With these sort of diagrams it normally takes a Sim City guru to make sense of them but in this case we can see that the extra real estate is taken up entirely by the additional cache memory.
Our ASUS board didn't allow us to change multipliers so we could only increase the FSB speed. 183MHz were reached without problem but we are sure we could have gone higher with patience and the right AGP and PCI dividers. Tests are conducted at stock speed but there does seem to be considerable room for overclocking.
On the left is the new Barton CPU and on the right is the XP2700+ based on the thoroughbred core. The two processors are identical apart from the increased level 2 cache.
Here's a summary of the system and the reference systems we compared it to:
The last two are primarily used for Desktop Video and so have weak 3D graphics so we will not be comparing them for some benchmarks. The XP3000+ and XP2700+ shared the same system so this leaves the Athlon XP2700+ & XP3000+ up against Intel's flagship P4 3.06GHz with Hyper-Threading and the P4 2.8GHz. We used our trusty Ti 4400 in all the benchmarks as well as TwinMOS DDR333 & DDR400 memory (512MB in each case). We quickly realized that DDR400 gave slower benchmarks than DDR333 because of asynchronous operation with the FSB so we performed the following tests with DDR333 memory (dual channel mode).
Let's start with the synthetic benchmarks. Readers should note these are largely dominated by factors that are not that important for most uses so they should be viewed from an academic standpoint.
As expected the NForce2 board is no match for the PC1066 memory in the P4 system. This benchmark does not stress the cache and therefore the XP3000 and XP2700 behave identically
Again the bandwidth of the P4 system shines through in this synthetic benchmark. As expected the extra cache memory of the XP3000+ provides some assistance.
Things are getting closer now. Again the extra cache does not help here.
Hyper-Threading appears to give the P4 a big boost and the Athlons struggle to keep up.
PC Mark 2002
The hard disk scores are not relevant as completely different drives were used. The CPU scores show the Athlons to be almost as good as the P4 despite the latter running about 50% faster. The extra cache of the XP3000+ gives a nice boost to memory performance.
3D Mark 2001
Before we show you the graphs we'd like you to see the score the XP3000+ gave as it first came up for us.
This score would not be out of place on a Radeon 9700 Pro but all we're using in our tests is a Ti4400 so this is truly a remarkable score. Let's see how the others did.
The XP3000+ blows away all the competition including Intel's flagship processor. In fact the extra cache seems to give a performance boost of almost 20%.
Unreal Tournament 2003
Now for some real world benchmarks starting with UT2003 Flyby.
The extra cache gives another performance boost but the main thing to note is that the Athlons comfortably beat the P4s at the higher resolutions. This seems to be due in large part to the 166MHz FSB. Let's examine the Botmatch scores.
Here we see something completely unexpected - the Athlons leave the P4s behind and shoot ahead, especially at the higher resolutions. We repeated this test several times with the same results - the KT400 anomaly seems to be a deficiency in the chipset and as we've pointed out before the NForce2 is much better at realizing the potential of CPUs than the flawed KT400. These scores should have games players rubbing their hands in anticipation..
Let's turn to an area where we know that fast CPUs will make a difference - Audio and Video encoding. This is becoming more and more popular and is very computationally intensive with long processing times (relatively speaking that is, this field is not for those that complain about how long their Outlook Express takes to load).
For consistency we will use Jet Li's The One as our test matter. It is not interlaced and contains a mixture of action types and is not too long. There will be three tests all using Divx 5.02. Audio will be encoded separately. I will try and keep my commentary to a minimum as all configuration information is shown in the images below.
and here are the results
The results show that bandwidth is king here and video encoders would be better off with a dual CPU rig than a single Athlon. We are hoping to have a review of a Dual Athlon MP2600+ system soon.
AviSynth and VirtualDub
No serious Divx encoder uses Xmpeg alone and it's just used by the media for benchmarking purposes so let's get serious. We ripped our source material to hard disk and created a DVD2AVI project file using forced film (it was 99% film). Loading this into Gordian Knot we first saved an .avs file with no changes at all (720x480) and no filters of any sort. This was loaded into VirtualDub with the following CODEC parameters :
After encoding we got these results:
Here we see the P4 pull ahead again mainly due to the P4 optimizations in VirtualDub. The XP3000 is on a par with the Dual Athlon MP1800 rig in this test where cache memory is not important.
This is all good stuff but how about a real-world test? To simulate a realistic test we added a neutral bicubic resize filter in the .avs file and used the following CODEC parameters (including two popular Pro settings) which are designed to total 700MB (when the audio is muxed in):
Which resulted in the following.
The addition of computationally heavy filters put less emphasis on the memory bandwidth and more on raw CPU speed. It also shows that the extra cache on the XP3000+ helps. This difference would be greater with more filters used in the encoding.
What about audio? We took the AC3 track from the above sample material and used HeadAC3he to convert it into Vorbis format so our final muxed file could have Ogg containment. There isn't space here to go into the advantages of Ogg Vorbis over MP3 and AVI so let's just say that Vorbis sounds about the same as MP3 for half the file size or twice as good for the same file size (that is subjective though).
Since it is more meaningful to show throughput than time taken (which depends on the length of the source) we display the results thus:
For the second time today we had to take a step back in surprise. The increased cache of the XP3000+ has resulted in a whopping 33% increase in performance in this computationally intensive benchmark. Even the 3.06GHz P4 (in a test where Hyper-Threading adds a 20% increase) is left trailing in the dust.
The Athlon continues to go from strength to strength - just when it seems the design has reached its limits AMD add an architectural improvement to give it more breathing room. The test results show consistently that where processing power counts the XP3000+ has no trouble beating the competition. Home users will be delighted at the gaming performance and lightning fast MP3 conversion. The only place the XP3000+ lags is in semi-professional media encoding which it's not really aimed at. Such users would benefit from having a Dual CPU system or one of AMDs upcoming Opteron systems.
With the Athlon64 now scheduled for a Q4 release it seems that the Barton CPUs will then be moved to the midrange and next year as the entry-level solution. The XP3000+ represents an ideal upgrade path for current Athlon users (check with your motherboard manufacturer's web site first though). In the ongoing game of leapfrog between AMD and Intel the XP3000+ appears to have hopped over the P4 3.06GHz and we can all look forward to the next development.
We would like to thank AMD UK for the loan of the test equipment.
All trademarks are the property of their respective owners.