Roughly this time last year we got our first look at the new AMD FX series which was based on the Bulldozer architecture. Unfortunately Bulldozer wasn’t all it was cracked up to be, falling short of competing with the Intel Sandy Bridge processors which at the time were already 9-months old. Worse still was the fact that in certain instances the FX-series failed to surpass the Phenom II range.

Concluding that article we noted that there was still more to come from Bulldozer, and AMD claimed that there was also much more to be seen from the FX range. Little has changed in the last year, particularly on the desktop front. Although we did recently see the release of the Trinity based APUs, little has changed when it comes to the FX series.

Today AMD is giving the FX series a refresh with the release of the second generation Vishera FX-series based on the Piledriver architecture. Piledriver is essentially an enhanced version of Bulldozer with the focus of improving instructions per clock and frequency.

Those hoping AMD would abandon Bulldozer might be disappointed as Piledriver picks up where Bulldozer left off. Fundamentally the architecture remains the same and rather than redesign from the ground up AMD has just made a number of small improvements.

Perhaps the most controversial issue with the FX-series was the new integer core/module design. With Bulldozer the floating-point unit featured a complete redesign. It had been improved to support many new instructions while also allowing resource sharing between cores. There are now two 128-bit FMACs shared per module, allowing for two 128-bit instructions per core or one 256-bit instruction per dual-core module.

Although each module appears in the operating system as two cores they are not true cores in the sense that they do not have as many resources as traditional AMD cores. So while AMD claimed that the FX-8150 was the first true eight-core desktop processor, in reality it wasn’t and the performance reflected that. Again Piledriver doesn’t change this architecture in any significant way, meaning that the module design still exists.

As we saw recently when testing the A10-5800K Trinity based APU there was a dramatic improvement in power consumption which allowed AMD to crank up the frequency. AMD has said that the 2nd generation FX-series processors will offer up to 20% to 30% better performance increase under digital media workloads, which is obviously a significant leap forward.

The new Vishera FX-series features 4, 6 and 8 Piledriver core models, all of which support the existing Socket AM3+ and 9xx series chipsets of the 1st generation FX-series Zambezi processors. These new processors also use Turbo Core 3.0 technology and employ the same dual-channel DDR3 memory interface.

The processors being released today include the FX-8350, FX-8320, FX-6300 and FX-4320, the specifications are listed in the table above. Although AMD is releasing four new FX-series processors we are only testing the FX-8350 and FX-6300. When compared to their predecessors the FX-8350 comes clocked a little over 10% higher than the FX-8150, though the maximum turbo core frequency is the same at 4.2GHz.

The six-core FX-6300 is clocked 6% higher than the FX-6100 though it is clocked lower than the FX-6120, FX-6130 and FX6200 processors. The FX-4300 is clocked at the same 3.8GHz frequency as the older FX-4150 though whereas the older model featured the full 8MB L3 cache the FX-4300’s cache has effectively been cut in half, reducing it to just 4MB.

Here we see that nothing has been changed to improve the memory subsystem of the Piledriver-FX processors. Although the AMD FX-8350 did have 4% more read bandwidth at its disposal, keep in mind it is clocked 11% higher than the FX-8150. Meanwhile when compared to the Core i5-3470 the FX-8350 was 24% slower when measuring read bandwidth and 44% slower when comparing write bandwidth.

Something we found odd when testing these new Piledriver FX processors was the L2 cache performance that was being reported by AIDA64. Using the latest version of the software the L2 cache performance was extremely low. While the write performance is similar to the Bulldozer processors, the read performance has effectively been halved.

Either something has gone drastically wrong with the Piledriver processors or the more likely explanation is that the AIDA64 software needs to be updated. In either case we will be looking into this further over the next few days.

First up we have the SolidWorks test from the SPECviewperf v11 benchmark. Here we see that the AMD FX-8350 is 6% faster than the new FX-6300. However when compared to its predecessor the FX-8150 the FX-8350 delivered virtually the same performance, despite featuring an 11% clock speed advantage.

The FX-6300 was also just 4% faster than its predecessor the FX-6100 though again it is clocked higher, 6% higher to be exact. Despite the only very slight performance advantage delivered by the new Piledriver FX processors, we found that the FX-8350 was able to match the performance of the Core i7-3770K while beating the Core i5-3470 by an impressive 18% margin.

The SPECviewperf v11 Maya test results showed very different performance trends when compared to SolidWorks. Although the AMD FX-6300 was again only slightly faster than the FX-6100, 3% faster to be exact, and the FX-8350 was just 4% faster than the FX-8150, both were slower than the Core i5-3470. The FX-8350 was 4% slower than the Core i5-3470 and 29% slower than the Core i7-3770K.

The CINEBENCH R11.5 CPU performance benchmark again shows only a slight performance advantage when comparing the new Piledriver FX processors to the older Bulldozer models. Here the AMD FX-8350 was 21% faster than the Core i5-3470, yet 9% slower than the Core i7-3770K.

When testing with the inbuilt WinRAR benchmark we found that the AMD FX-8350 was 10% faster than the FX-8150, while the FX-6300 was just 2% faster than the FX-6100. This meant that the FX-8350 was 8% slower than the Core i5-3470 and 15% slower than the Core i7-3770K.

Again we see virtually no performance advantage is delivered by the new Piledriver FX processors when compared to the older Bulldozer parts. The AMD FX-8350 was just 2.5% faster than the FX-8150, while the FX-6300 was just 2.5% faster than the FX-6100. This meant that the FX-8350 was 13% faster than the Core i5-3470 while it was 23% slower than the Core i7-3770K.

The Photoshop CS5 test saw the AMD FX-8350 beat the FX-8150 by a slim 3% margin, while the FX-6300 was just 4% faster than the FX-6100. Meanwhile the FX-8350 beat the Core i5-3470 by a 7% margin, though it was also 26% slower than the Core i7-3770K.

For only the second time thus far we saw the AMD FX-8350 outperform its predecessor by a 10% margin or greater. When testing with Fritz Chess 13 the FX-8350 was 12% faster than the FX-8150, while the FX-6300 was 11% faster than the FX-6100. This meant that while the FX-6300 matched the performance of the Core i5-3470, the FX-8350 was 19% faster.

The new Piledriver FX processors failed to deliver much in the way of extra performance in our HandBrake test, as the FX-8350 and FX-6300 were just 1 – 2% faster than their predecessors. That said the good news was that the FX-8350 was 17% faster than the Core i5-3470 and just 6% slower than the Core i7-3770K.

The good news for AMD when testing with the x264 HD Benchmark was that the FX-8350 was 6% faster than the FX-8150, while the FX-6300 was 4% faster than the FX-6100. The bad news being that the FX-8350 was 16% slower than the Core i5-3470.

Unfortunately AMD finished off our encoding testing phase on a rather poor note, as both Piledriver FX processors provided no real performance gain when compared to their Bulldozer parts. Worse still this meant that the FX-8350 was 23% slower than the Core i5-3470 and 31% slower than the Core i7-3770K.

When testing with Just Cause 2 at 1920x1200 using the GeForce GTX 580 graphics card we see that there is just a 7% performance difference between the slowest and fastest processors. This leaves very little room for improvement and therefore we were not surprised to find no real difference between the AMD FX-6300 and FX-6100. That said the FX-8350 was found to be 3% faster than the FX-8150, likely due to the 11% bump in clock speed.

As a result the FX-8350 was able to deliver 1fps more than the Core i5-3470, though we should point out that it was only just able to match the performance of the Phenom II X4 980.

This time we find that the fastest processor is 11% faster than the slowest when testing with The Witcher 2. That said the AMD Piledriver FX processors were unable to improve on the performance set by the Bulldozer processors and as a result the FX-8350 was 6% slower than the Core i5-3470.

This time when testing with Crysis 2 we again saw almost no difference in performance when comparing the new Piledriver FX processors to their Bulldozer counterparts. This time there was a 13% performance difference between the fastest and slowest processor, while the FX-8350 was 5% slower than the Core i5-3470.

When we tested Trinity last month we found that the 3.8GHz A10-5800K consumed 4% less power than the A8-3850 while being clocked 31% higher. The new Piledriver-FX processors are not quite that impressive as we found the AMD FX-8350 to consume roughly the same amount of power as the FX-8150, though it is clocked 11% higher so some improvement has been made.

The FX-6300 system also consumed roughly the same amount of power as the FX-6100 system, which isn’t bad given the FX-6300 is clocked 6% higher. Still when you consider that a system featuring the FX-8350 consumes roughly the same amount of power as a system using the Core i7-3960X, that’s not good news from AMD. Worse still the Core i5-3470 system consumed just 122 watts which is 52% less than the FX-8350 system.

When the original Bulldozer based FX processors landed roughly a year ago they for the most part offered no real performance advantage over the existing Phenom II processors, particularly in our gaming and encoding benchmarks. Furthermore, when it came to efficiency, aka power consumption, almost no improvement was made here either.

Then jump to the Trinity release last month and here we found that the Piledriver based A10-5800K provided roughly the same performance as the A8-3850 (clock-for-clock). However the A8-3850 doesn’t use Bulldozer cores and instead is based on the AMD 10h, or as it is commonly referred to K10 architecture used by the Phenom II processors.

So then if we are going to take what we already know about the Trinity (Piledriver) and Llano (K10) comparison, then it would be safe to assume that the FX based Piledriver processors wouldn’t really be any faster than the Bulldozer FX processors and sadly this appears to be the case.

With the exception of WinRAR and Fritz Chess, the AMD FX-8350 was no more than 6% faster than the FX-8150 in any one of our tests, despite being clocked 11% faster (though we should keep in mind both processors feature a max turbo frequency of 4.2GHz). It was a similar situation when comparing the FX-6300 to the FX-6100, though the margins here were even slimmer as the Piledriver part is clocked just 6% higher.

Honestly, despite what we saw from Trinity we were still expecting more from the Piledriver-FX processors. Unfortunately we have had just 2 days to test these new processors, which obviously isn’t a great deal of time, and much of it was spent trying to work out if there was something off with our test bed.

In the end the results we were receiving were consistent and we had no reason to doubt them despite the fact that AMD was claiming up to a 23% performance advantage in their reviewers guide document. That said, the words "up to" are quite different from average performance and this is what we found.

Moreover after going through all of AMD’s data which compared the FX-8350 to the FX-8150, we found just one test that showed a significant difference and that was the 3Dmark Vantage CPU1 - AI Test which found the FX-8350 to be 21% faster. However when we compiled the data from the 130 game related benchmarks that they ran across 20 different games we found that their own results showed that the FX-8350 was just 5.5% faster.

So then with no real performance advantage based on our findings as well as AMD's, where does this leave the new AMD Piledriver based FX processors? Well performance aside the most important factor really is the price and this is how AMD plans on being competitive.

The new flagship part, the FX-8350 is set to begin selling at just $200, which is the same price as the Core i5-3470. The FX-6300 on the other hand will cost just $135, which is considerably cheaper than the most affordable Ivy Bridge Core i5 processor. In fact it is also much cheaper than the most affordable Sandy Bridge Core i5 processor the i5-2300 which costs $187. In fact the FX-6300 is set to cost just a fraction more than the Core i3-3220 ($130) which we also featured in our review.

Therefore if we break down our testing we found that the FX-8350 was 7% faster than the Core i5-3470 in our synthetic benchmarks, 13% faster in the application tests, 7% slower in the encoding benchmarks and 3% slower in the gaming tests. This meant that overall the FX-8350 was just 2.5% faster than the Core i5-3470.

However the FX-6300 was 17% faster than the Core i3-3220 when comparing synthetic performance, 44% faster in the application tests, 30% faster in the encoding benchmarks and just 2% slower when comparing gaming performance. This means that overall the FX-6300 was 22% faster than the Core i3-3220, and given the similarity in pricing this is a good result for AMD.

Of course there is that little issue of efficiency that AMD’s aggressive pricing strategy can’t really erase. When compared to the Core i3-3220 the FX-6300 system consumed 86% more power to deliver just 22% more performance on average.

Bottom-line is the new Piledriver-FX processors provide a quick and cost effective upgrade for those still using lower-end K10 hardware. That said, those using high-end Phenom II X4 and in particular X6 processors, there still isn’t a lot to be gained for the upgrade, despite how cheap these new processors might appear. The real issue for AMD has to be that for those building a new system from the ground up, the Ivy Bridge range is still a more attractive option thanks to its much better single thread performance and vastly superior efficiency.