AMD FX-8150 review

AMD’s FX-8150 is the first processor we’ve seen that uses the new Bulldozer core – the first all-new AMD core since 2007’s K10. This top-of-the-range model has a whopping eight cores running at 3.6GHz, and can boost up to 4.2GHz in Turbo mode when lightly-threaded applications aren’t using all the cores.

Like AMD’s Llano processors, you’ll need a new motherboard to use a Bulldozer chip – it requires a Socket AM3+ motherboard with an AMD 990FX chipset. The cheapest board currently available is Gigabyte’s £115 GA-990FXA-UD3, but prices should come down as more boards are launched.

The FX-8150’s architecture is significantly different from that of its main rival: Intel’s Sandy Bridge. Sandy Bridge chips consist of up to four cores, each with their own Level 1 and 2 cache, sharing a pool of Level 3 cache. The FX-8150 has four modules, each of which contains two Bulldozer cores. Each core has its own level 1 cache, and the two cores within each module share 2MB of level 2 cache. In turn, the four modules share an 8MB block of level 3 cache. AMD claims this means that single-threaded applications have access to more resources, as they can access 2MB of level 2 cache and 8MB of level 3 cache even when using just one core; with Sandy Bridge, single threads only have access to 256KB of Level 2 cache and 8MB of level 3.

The top-of-the-range FX-8150 has eight cores arranged in four sets of two

AMD’s FX range also doesn’t have built-in graphics, but every processor in the range has an unlocked multiplier for easier overclocking. At launch, there will be three FX models: two eight-core chips running at 3.1GHz and 3.6GHz, and a six-core model running at 3.3GHz (see table below). Later on AMD will launch a four-core version, starting at £100.

All the above prices are from www.aria.co.uk. Like the cost of the motherboards, this price may well drop through competition when more stock is in the channels, but at the moment it’s around £30 more than the Intel Core i5-2500K, which is also unlocked and which we consider the FX-8150’s chief rival.

Next page – performance testing

PERFORMANCE TESTING

We paired the processor with 4GB RAM and ran our benchmarks, which consist of intensive image conversion, video-encoding and multitasking tests. In these tests the Core i5-2500K manages 100 in each benchmark. The FX-8150, however, could only manage an overall score of 85 – see below for the full results.

The problem seemed to be that our benchmark software was struggling to take advantage of all the cores. Even when video-encoding, where multi-core processors are normally strong, Windows’ Task Manager only showed the FX-8150 at 45-50% processor load. In contrast, during the same test the Core i5-2500K stayed at between 90% and 100% usage. Windows’ processor usage graph (below) shows that one FX-8150 core was under heavier use than all the others, showing that the program’s threads aren’t evenly balanced across all the cores, which caused a bottleneck.

When video-encoding, the first core is under heavy load, but the rest hover below the 50% mark

To double-check we installed Sony Vegas Movie Studio HD Platinum 11, and encoded a 1080p AVCHD video to 1080p Blu-ray format at 24fps. On the Intel Core i5-2500K we saw processor usage of 80-85% and the video encoded in 12m 43s, but on the FX-8150 the processor sat at between 65 and 75% usage and it took 14m 11s to encode the video.

This shows that even multithreaded applications such as video encoding can’t take advantage of the FX-8150’s eight cores efficiently, and as an individual Bulldozer core seems slower than the equivalent Intel Core i5-2500K core, despite the Bulldozer’s 300MHz faster clock speed, this holds it back in application performance. To compare core-to-core performance, we disabled all but one of the cores on both our Intel Core i5-2500K and AMD FX-8150 test systems and ran our benchmarks again. The Core i5-2500K managed 29 overall and the Bulldozer managed 25, showing each of AMD’s cores to be 14% slower than its rival from Intel. We also underclocked the FX-8150 to 3.3GHz (the same clock speed as the Core i5-2500K) and re-ran the test. With one core and the same clock speed, the FX-8150 scored 24, compared to the Core i5-2500K’s 29.

AMD single-core

Intel single-core

AMD’s latest can’t match Intel’s rival in a core-for-core comparison

To try to run all the FX-8150’s cores at maximum usage, we ran two sets of our benchmarks at once. Here, both the FX-8150 and the Core i5-2500K ran at 100% processor usage at all times. Once we’d calculated the average of the two benchmark scores, we saw that AMD’s chip had finally managed to pull ahead, with an overall score of 82 compared to the Intel processor’s 63. This shows that the FX-8150 can have an advantage, but only if you run incredibly intensive tasks such as encoding two HD videos at once to another format or encoding two HD videos while playing back a third.

AMD multi-benchmark

Intel multi-benchmark

Under incredibly intensive multitasking, the FX-8150 has the edge over Intel’s Core i5-2500K

Next page – overclocking and conclusion

OVERCLOCKING

As the FX-8150 has an unlocked multiplier, you can overclock the processor without overclocking your RAM and the motherboard’s chipset. With the stock cooler, we only managed to get it up to 4.2GHz, but with a Cooler Master V8 cooler (£45 from www.expansys.com) we could set the processor multiplier to 45x to see a stable 4.5GHz – an increase of 900MHz. At this speed, our overall benchmarks score increased to 96 – a 10 per cent improvement. It’s a useful speed boost, but it’s overshadowed by the overclockability of Intel’s Core i5-2500K. Even with the stock cooler we could get this up to 4GHz, an increase of 700MHz, and this led to an overall score of 118 in our benchmarks – far ahead of AMD’s chip.

CONCLUSION

It’s great news to see a new core architecture from AMD after four years, but the Bulldozer-based FX-8150 isn’t as quick as we’d hoped. The main problem is that each individual core isn’t as fast as each individual core in Intel’s Sandy Bridge processors, leading to slower overall performance in our benchmarks. Part of the issue is that most software isn’t written to take advantage of so many cores, so it’s hard to push this processor to its limits. Running two sets of our benchmarks at once to push the processor showed it could be faster than its Intel rival when all cores have to be used, but this is unlikely to happen in everyday use. Given that you need to buy a new motherboard to use the Bulldozer chips, you may as well buy an Intel Sandy Bridge motherboard and processor.