AMD Athlon II 620 X4 Page: 1
Introduction
 
x4For a long time AMD held the performance crown. Anyone with a passing interest in building their own computers, or having the fastest, ran AMD. The Barton processors were almost legendary. Then Intel came along with the 775 series of chips and AMD started lagging behind somewhat.
 
The upgrade of AM2 to AM2+, and then the switch to the DDR3 based AM3 platform redressed the balance and the recent various takes on the Phenom II core in processors ranging from the X2 240 all the way up to the full fat Phenom II X4 965BE have all provided good value for money and ensured that there is a chip for any budget, without requiring the user to purchase a new motherboard and relevant components every time they want to upgrade.
 
Today on Overclock3D we're looking at very aggressively priced and featured processor in the Athlon II series, the X4 620. Second from the top of the Athlon II line, the X4 630 is clocked slightly faster, does it provide the ultimate in performance for price, or have too many compromises been made?
 
 
Technology
 
AMD have codenamed this chip the "Propus" core and it very much appears to be a cut down version of the Deneb core found in the Phenom IIs.
 
 Athlon II X4 620Phenom II X4
Cores
 QuadQuad
L1 Cache4 x 128KB4 x 128KB
L2 Cache
4 x 512KB4 x 512KB
L3 Cache-- 6 MB
Transistors300 million758 million
Process45 nm45 nm
 
At first glance it does appear that it's a Deneb core with no L3 cache and under half the number of transistors. That should improve latency and help keep the processor cool. Always something we're interested in because as we all know, heat is the main problem when attempting to overclock a processor, or in these modern times, having one that is energy efficient.
 
The loss of the L3 cache appears to be the major thing missing. No other obvious compromises have been made. The X4 620 supports x86, x86-64, MMX, 3D Now, SSE, SSE2, SSE3 and SSE4a instruction sets.
 
One of the few faults I can find with the initial specifications is that this is not one of AMDs Black Edition range with unlocked multipliers, so this defaults to a multiplier of 13x, and can only be adjusted downwards. However, AMD are clearly aiming this processor at the mainstream and cost-effective upgrade end of the market, and it's by far the cheapest Quad Core you can buy, so to complain about something like an unlocked multiplier is a little harsh.
 
X4 620
 
Considering the X4 can be found for about half the price of a Phenom II, is the lack of L3 cache going to cripple this, or will the decrease in latency mean that this is one of the best bargains to be found for those wishing to dip their toe into the quad-core waters?


AMD Athlon II 620 X4 Page: 2
Test System
 
As mentioned on the previous page this is very much a processor for those seeking either a cheap all-round system, or to install a better processor into their current set of hardware. The AMD Dragon platform is build around this very philosophy. To this end it would be defeating the object somewhat to drop it into a high-end all-singing system and so we've gone for a more budget conscious one to find out exactly how much performance can be obtained on a very tight budget.
 
Motherboard : Gigabyte MA785GT-UD3H
Processor :
AMD Athlon II 620 X4
Memory :
Corsair XMS3 1333MHz DDR3 2x1GB
Cooler :
Thermalright Ultra 120 with Yate Loon 120mm fan
Graphics Card :
ASUS 4850 512MB TOP using Catalyst 9.10
HDD : 
Samsung Spinpoint F1 1TB
Operating System : 
Windows 7 Home Premium 64
 
So far, so very average. However, with a little overclock this could be quite a useful rig, so let's see what we can get the processor to do before we move on to some benchmarking.
 
 
Overclocking
 
CPUzOverclocking the X4 620 was an exercise in both joy and frustration in equal measures. Starting from the defaults of 2000MHz NB and HTT, 200MHz bus x13 multiplier to give the default 2.6GHz we kept the voltage stock and, like all good overclockers, gradually bumped it up 5MHz at a time and tested for stability each time. The first point of call with any processor that starts in the 2GHz bracket is to hit the 3GHz barrier, and 230 bus speed saw this appear without even the slightest difficulty. 240 bus speed came and went, and 250 was hit with only a minor tweak to the voltage, up to 1.39v. Load temperatures were still around 40°c and there seemed no end to how far this could be pushed.
 
Unfortunately that seemed pretty much to be that. An increase of the bus to 255 provided a blue screen, and so the voltages were increased. +0.1v on the northbridge and HTT, and the processor was increased to 1.45v. It was still very unstable and whilst it would at least boot into Windows now, it was only a couple of minutes before the dreaded blue screen appeared again. Incredibly frustrating. No reasonable combination of voltages or tweaks would get it stable at anything other than the 3.25GHz that was got very early and pretty much at default voltages.
 
As a final attempt the core voltage was upped to a staggering 1.55v in an attempt to get it stable around the 3.3GHz mark. This was met with a annoying lack of success. There is a definite block with a bus speed of over 250 which has baffled us all, especially considering the ease at which that 250 bus was obtained.
 
Therefore the voltage was lowered back to just above stock, with the bus at 250, northbridge at 2000MHz and HTT at 2000MHz to run the overclocked tests. A 650MHz increase with absolutely zero effort on a chip with this much value is not to be sniffed at, so let's run some benchmarks and see if all this effort was worthwhile.


AMD Athlon II 620 X4 Page: 3
Synthetic Benchmarks
 
Whilst not always an indicator of actual performance, synthetic benchmarks have a place in our hearts because there is nothing like the purity of numbers to give a feel for the capability of a processor or system, and you can also get a feel for how much extra performance to expect in general usage from an overclock.
 
 
SiSoft Sandra shows that the overclock of 25% gives an improvement of around 23% in the Arithmetic tests. Not bad at all as we know that you never get a one to one improvement.
 
 
The Multi-Media test showed an even higher improvement. Maths tells us that if we saw a one to one improvement via the clock speed then our Multi-Media Int overclocked result would be 149591. We actually got 149134 showing that this is one theoretical test that gets very close to a direct increase when the processor is overclocked.
 
 
7-Zip 64 bit comes with a built in benchmark. For this test we used the 32MB dictionary and ran 6 passes. This more real-world test showed similar results to the above two, in that this scales very well with little loss in performance between theoretical increase and that seen by the end user.
 
 
It wouldn't be CPU review without the old favourite SuperPI. In this case, whilst there is a clear improvement between the stock and overclocked tests, the 1M result isn't exactly earth shattering. Not bad for a disappointing overclock on a L3 disabled processor, but I was expecting around 20 seconds.
 
Let's go and see some more synthetic results.


AMD Athlon II 620 X4 Page: 4
Synthetic Benchmarks cont...
 
 
Lavalys Everest Ultimate has a suite of great benchmarking tools and information available as we discovered here. The AES encryption test shows a linear increase between clock speed and score. PhotoWorxx displays a smaller increase, but this is also more dependant upon subsystems. CPU Queen however shows a 225 point increase over the theoretical improvement. Very impressive indeed. Although it's within tolerance nonetheless it's a good showing.
 
 
As we start to move away from purely CPU synthetic benchmarks to those that stress the whole system the difference lessens. PC Mark Vantage does seem more useful for testing everything, and it's here than the use of a standard HD and lower amount of memory really shows through. A 600 point increase for free isn't bad, but let's see if an actual program shows a similar lack of improvement.
 
 
PovRay is a script based freeware rendering program. As rendering is very processor intensive the improvement of 191 seconds (or 3 minutes 11 seconds) off of the stock speed time is very impressive. Again, like the CPU Queen test, faster than theoretically possible just based on the clock increase but within benchmark variances.
 
 
Cinema4D sees a reasonable improvement. One of the things that this graph doesn't show is that the normal clocked test showed a 3.64x processor scaling between single and multi-processor rendering, whereas the overclocked test showed 3.44x scale. Still a good result and proof that if you have any interest in graphics or rendering, this chip definitely should be on your shortlist.
 
Phew. Let's go and have a look at some game performance now we've covered every synthetic benchmark anyone could ever need.


AMD Athlon II 620 X4 Page: 5
Gaming Benchmarks
 
Having exhausted the synthetic tests, it's time to run some gaming benchmarks and see how this budget system performs.
 
 
If there is any form of entertainment that it's impossible to have enough horsepower for, it's Chess. Quad Core processors are a boon for these, and Fritz 11 is one of the finest available. Using the default engine and the in-built benchmark you can see that the difference between the stock and overclocked processors is marked. The amount of extra thinking the CPU can do in a given time ensures it will play a much sterner game.
 
 
Devil May Cry 4 is, along with Street Fighter IV, one of the games that Capcom uses to make a mockery of anyone who says that a good console port can't be done. As with all CPU based tests the resolution and details were reduced to ensure that the loading was on the CPU rather than the graphics card. Surprisingly this benchmark showed almost no difference between the two. It's still more than playable, and the graphics can be ramped to the max without issue, but this is one case where having no result speaks as much as having a good one.
 
 
Far Cry 2 is an old favourite of OC3D, so it's no surprise to see it here. It scales very well and has a very comprehensive built in benchmarking utility. Firstly a typical low detail, low resolution run. You can see that even at default speeds it provides a perfectly playable frame rate. The improvement in overclocking the processor is typical for gaming, which doesn't tend to see as much reward for high overclocks as from a better graphics card. Finally, let's crank those details right up to something most people will actually play at.
 
 
At 1680x1050, and everything set as high as possible, we're still averaging just under 50 frames per second. Considering that the pace of this game ensures it's playable at anything over 40, and that the Ranch Small benchmark is very heavy with the fire and explosion effects, this is a sparkling result indeed.
 
Ok, enough of being Mr Nice OC3D, let's really put this system on its knees with some high-res, high image quality testing. Turn that page!


AMD Athlon II 620 X4 Page: 6
Further Gaming
 
Ok. We've tried a few of the CPU based benchmarks. We've done some gentle gaming as befits this specification. Let's run some OC3D classics with all the eye candy on. We know you're all gagging to know how our three regulars do and OC3D always wants to keep its readers happy. All games are run at the 1680 x 1050 resolution, with maximum possible settings using in game menus. AA and AF levels are noted after each graph.
 
 
Crysis is a well-known system destroyer. 1680 resolution with everything set to very high, but AA and AF off, should really see how much juice the CPU has got, because we all know the 4850 definitely can't cope with this level of detail. Or at least that would be the belief before seeing the results. Yes the actual frame-rates aren't that impressive in terms of pure numbers, but make no mistake, this is buttery smooth. The particular test used here was a run into the first major fortified structure you come across in the game. Anyone who has wandered into the minefield knows where we are. Lots of bullets, lots of people, lots of water effects and dense forest. Crysis certainly isn't shy about making 25-odd FPS feel like 60. The 620 X4 remained cool even overclocked and following multiple test runs. Very impressive.
 
 
Call of Duty 4 : Modern Warfare is coming to the end of its life as a useful test of modern hardware, especially with the forthcoming release of MW2. Still, for the time being we all know it and love it. In this particular case apart from the obvious 1680, maxed settings we also added 8xAF and 4xAA. The tests were run through the Epilogue level as it's definitely a firefight with good smoke effects and some explosions. It's also easily repeatable, something that manual benchmarks delight in. As you can see from the above graph it posed absolutely no problem for either the processor or the graphics card. Both of them breezing through with amazing frame-rates.
 
 
Race Driver GRiD is, like COD4, about to be supplanted by a similar game. In this case Dirt 2. However it's still one of the more playable racing games around and works happily no matter what the hardware configuration. It's especially kind to ATI cards so the frame rates we're getting here are no shock at all. 1680, everything on High or Ultra, 4xMSAA around the very testing US circuits, and even at default clocks on a low-end graphics card the minimum FPS was 59, just one shy of the magic 60. It averaged into the 80s even from the back of the pack on the demolition derby track when all hell was being let loose. Along with Far Cry 2 a seriously impressive result.
 
Time for a summation of this.


AMD Athlon II 620 X4 Page: 7
Conclusion and Final Thoughts
 
It's difficult not to be impressed by the AMD Athlon II 620 X4. It provides a genuine 2.6GHz quad-core processor that will work perfectly well in either your current AM2+ motherboard, making it even more of a bargain, as it would provide 4 core goodness for a very small investment. As part of a new build for the budget conscious it clearly would be a good starting point and the interchangeability of AMDs platform ensures that should you require even more horsepower a Phenom II is just a short stones throw away. However, today is all about this particular chip, so how did it fair?
 
Initially the lack of L3 cache would appear to be a major hindrance and one that seemed like a strange decision on AMDs part. This certainly isn't like some of the Athlon II dual-cores in which a quick visit to Advanced Clock Calibration could provide a nice surprise. No the L3 cache just doesn't exist. There are reports that in the very very early stages of release the 620 X4 was just a Phenom II with disabled cache, but that certainly isn't the case here, and nobody should buy one looking for a cheap Phenom II.
 
However, testing clearly showed that this L3 cache actually makes very little difference to the real-world performance. Synthetic performance was exceptionally impressive, particularly once it had been overclocked. The 3 minute reduction in rendering time in PovRay stood out as one of the more impressive things that we came across during testing, and certainly for anyone looking for an all-rounder this is a genuine bargain.
 
Gaming performance, especially considering the mid to low-end graphics card used in testing, was quite surprising. Devil May Cry 4 provided playable (60fps +) frame rates from the medium resolution and details used in the review. Far Cry 2 was a revelation. The Dunia engine scales very well and, once we'd seen the results of the low resolution test with 70+ fps average, hopes were high that this processor could really make the 4850 sweat. It easily dealt with everything set to very high and 1680x1050, providing a very playable experience. Crysis performed about as expected although I was surprised that even the lowly 4850 could provide nearly 30fps average, and again it must be stated that this at no point felt jerky thanks to the 620 X4s ability to push data at a very consistent rate. Grid performed fantastically as it always does on ATI hardware, but as someone who's run it on this graphics card but with a lesser processor it has made a definite improvement all across the board.
 
A motherboard, processor and graphics card combination that can play Far Cry 2 at 50fps, gets great frame rates in Crysis, and decimates COD4 and Grid for around £220? Sounds about as good value for money as it's possible to get, and speaks very highly of AMDs Dragon platform.
 
The overclocking results are a mixed bag. It hit 3.25GHz so easily that anyone could do (put bus speed at 250, done). However the initial promise this gave that we might have the next Q6600 on our hands was quickly dispelled as 250 seemed to be the wall. No amount of tweaking and adjustments could get the system remotely stable over this despite the fact it was 30°C underneath the thermal limit. It's not clear yet whether this is a problem with the chip or with the motherboard, but you can rest assured that this wont be the last time we torture this chip to try and unlock its obvious potential.
 
So in conclusion if you have an AMD motherboard or are looking to build a system on a budget, this processor is highly recommended. It's got enough grunt to power some heavyweight applications and yet still provide good gaming performance. In fact the money you save basing your system on this great value chip could easily be spent on a 5850 or similar and we are in no doubt that the AMD Athlon II 620 X4 could provide the required grunt to make it sweat. Highly recommended and I have absolutely no qualms about giving it the OC3D Value For Money Award. A 3.25GHz Quad Core for around £80 seems to good to be true, and yet isn't. Congratulations to AMD.
 
The Good
- Price.
- Performance.
- Cool even under overclocked conditions.
- Doesn't require a whole new subsystem.
 
The Mediocre
- It overclocks well, but promises much more headroom than seems to be available.
- No L3 cache, although this didn't seem to make much difference in the tests.
 
The Bad
- Nothing springs to mind.
 
Value    
     
Many thanks to AMD for providing the Processor used in todays test. Discuss in our forums.