AMD’s three new FX processors are here: the FX-8370, FX-8370E, and FX-8320E. The company also made several older models more affordable. As we'll see though, street pricing still lands close to the respective MSRPs. And the performance benchmarks are going to suggest that AMD needs a couple rounds of cuts before mainstream enthusiasts can get excited about its latest efforts.
AMD sent over an FX-8370E operating at a 3.3 GHz base clock rate. It accompanied the chip with a motherboard, since the new processor wouldn't run on our older Socket AM3+-based platforms; the requisite BIOS updates weren't available yet.
Furthermore, AMD told us that its latest and greatest was rated for a TDP of just 95 W, which would certainly be an improvement over previous FX models. Hold your horses, though. How did the company's engineers achieve this using a familiar architecture and the same manufacturing technology? Is it binning? A less aggressive Turbo Core implementation? Some revision to the die?
|
|
|
|
|
|
|
A New Focus on Efficiency? Fine by Us!
We’ll take AMD’s word (and TDP rating) as accurate and switch things up a bit to reflect on the new CPU's focus. Of course we'll give you performance benchmarks as well, but only at the end and after a special preamble.
Our purpose is to figure out how much power this processor draws at its stock and Turbo Core clock rates. We want to know where its sweet spot is, and to explore when overclocking just doesn't make sense anymore. From there, we'll benchmark at three different clock rates for a more in-depth look at efficiency.
Unfortunately, AMD shot itself in the foot by using ASRock's Fatal1ty 990FX Killer. Its idle power consumption is a lot higher than what we’re used to from other boards with the same socket. Obviously, this results in the CPU's total power consumption measurements, measured from the motherboard, turning out a lot higher than they need to be.
In the end, this is just another FX processor, so we’re not going to cover the architecture in-depth. If you're new to the FX and the technology inside of it, check out AMD FX-8350 Review: Does Piledriver Fix Bulldozer's Flaws?
Where the new CPU lands in AMD's FX family, sporting four modules able to schedule eight threads concurrently, can be gleaned from the following table:
| Products |
AMD FX-9590
|
AMD FX-9370
|
AMD FX-8370E
|
AMD FX-8370
|
AMD FX-8350
|
AMD FX-8320
|
AMD FX-8320E
|
AMD FX-8330
|
| Pricing |
|
|
|
|
|
|
|
|
| Base Clock | 4.7GHz | 4.4GHz | 3.3GHz | 4.0GHz | 4.0GHz | 3.5GHz | 3.2GHz | 3.3GHz |
| Max. Turbo | 5.0GHz | 4.7GHz | 4.3GHz | 4.3GHz | 4.2GHz | 4.0GHz | 4.0GHz | 3.9GHz |
| L2/L3 Cache | 8MB/8MB | 8MB/8MB | 8MB/8MB | 8MB/8MB | 8MB/8MB | 8MB/8MB | 8MB/8MB | 8MB/8MB |
| NB Clock | 2.4GHz | 2.4GHz | 2.2GHz | 2.2GHz | 2.2GHz | 2.2GHz | 2.2GHz | 2.0GHz |
| TDP | 220W | 220W | 95W | 125W | 125W | 125W | 95W | 95W |
Test System And Measurement Methodology
Our German lab went the extra mile for drilling down into power consumption, cutting the braiding from our power supply's cables to give us the same measurement capabilities you've seen in our graphics card launch coverage. The readings are based on the four-channel HAMEG HMO 3054 oscilloscope.
We’ll first take a look at the power consumption, which we measured via the motherboard’s power connector and includes any losses due to the voltage regulators. Depending on load, these losses can reach eight percent. Because the ASRock motherboard AMD sent over doesn’t let us analyze the VR data, we weren’t able to factor out those losses, though. The FX-8370E’s actual power consumption is a bit lower than the values reported here.
| Test System | |
|---|---|
| System | AMD FX-8370E be quiet! Dark Rock Pro Air Cooler be quiet! Shadow Rock Slim Air Cooler ASRock Fatal1ty 990FX Killer 16 GB Radeon DDR3-1866 Samsung 850 EVO 512 GB be quiet! Dark Power Pro 1200 W Microcool Banchetto 101 |
| Methodology | No-contact current measurement at all rails Voltage measurement IR real-time monitoring |
| Equipment | 1 x HAMEG HMO3054, 500 MHz Four-Channel Oscilloscope 4 x HAMEG HZO50 Current Probe 4 x HAMEG HZ355 (10:1 Probe, 500 MHz) 1 x HAMEG HMC8012 DSO 1 x Optris PI450 80 Hz Infrared Camera + PI Connect |
Optris' PI450 is an infrared camera that was developed specifically for process monitoring. It supplies real-time thermal images at a rate of 80 Hz. The pictures are sent via USB to a separate system, where they can be recorded as video. The PI450’s thermal sensitivity is 40 mK, making it ideal for assessing small gradients.
We're also using be quiet! Dark Rock Pro. It’s a large dual-tower cooler with two fans that spin based on the CPU’s temperature. Even our highest overclock doesn’t pose a problem for it. As a result, the fans max out at 800 RPM, and our acoustic measurement equipment can’t pick it up. So, we're forgoing our usual noise level measurements.
Even after a lengthy test run, idle temperatures are extremely low. We start a new benchmark once the heat pipe cools down to 30 degrees Celsius.

Because we want to uncover the sweet spots for overclocking, voltage, and cooling, we’ll switch out the Dark Rock Pro CPU cooler for a less expensive model later, and then make our recommendation.
Here's another of the motherboard's shortcomings: a measurement of 75 degrees is way too high for a chipset at idle. We even measured well over 80 degrees Celsius on the chipset’s surface after running the motherboard inside of a case. That's enough to hurt your fingers, as I now know from experience.
AMD FX-8370E at 3.3 GHz
Core Voltage
The core voltage, provided by the VRM, plays a prominent role in determining power consumption and how much waste heat is produced. A real 1.17 V reading is a bit lower than the BIOS setting of 1.1850 V. Interestingly, the value fluctuates a lot when the BIOS is set to regulate voltage automatically, whereas it doesn't if you switch the firmware to manual control.
The Turbo Core clock rate falls all the way to the base frequency during our stress test.
Power Consumption
At idle, we’re looking at 17 W. Under a taxing load, that number jumps to 75 W measured at the rail supplying the CPU. This is both unexpected and pleasant. It's the sort of power figure we've always wanted to see from a top-end AMD processor. After all, once the voltage regulator losses are calculated, we should be looking at right around 65 to 68 W. Sure, Intel's CPUs are lower still (and a great many faster, too), but not as far as you’d think.
Temperatures
A lower-power processor is bound to demonstrate better thermal performance. And indeed, the temperatures we're reporting are good at idle and under load. We could have probably used AMD's stock cooler without creating a bunch of annoying noise.
Heat pipes on the cooler we did use only warm up to 34 degrees Celsius. A 40-degree core temperature is also impressive.
Looking at the power consumption-to-performance ratio, we have a really interesting eight-core CPU that’s definitively better than what AMD offered previously in the FX family. Even though a lower clock rate results in a performance hit, the FX-8370E is exactly what we were hoping to see from AMD to begin with.
AMD FX-8370E At 3.5 GHz
The first step in our exploration is increasing the FX-8370E’s frequency by just 200 MHz. All of the basic BIOS settings are left alone otherwise.
Core Voltage
The core voltage curve is now smooth. Also, it drops to 1.14 V with a Vcore setting of 1.1850 V manually set in the firmware.
Power Consumption
Power draw measured directly from the PSU is 78 W, representing a meager 3 W increase. This proves two things. First, the motherboard’s automatic regulation isn't well-optimized, since all other settings are the same as they were during our stock 3.3 GHz testing. Second, this CPU is a lot more efficient than its predecessors.
Temperatures
The temperatures don’t rise in spite of the higher clock frequency, which is probably due to the lower core voltage.
The AMD FX-8370E remains easy to cool at 3.5 GHz. A sub-$20 cooler should suffice.
AMD FX-8370E At 4.0 GHz
Let’s take a big step up and push the FX-8370E all the way to an even 4.0 GHz.
Core Voltage (Vcore)
An average of 1.17 V is good enough for stable operation, and we had to change the BIOS setting to 1.2125 V to achieve it. The curve continues to look very smooth.
Power Consumption
We measure 90 W at the appropriate rail, which should work out to somewhere between 80 and 82 W for the CPU alone, once the losses due to the voltage regulator are subtracted. Nice!
Temperatures
The thermal readings remain in a range that could be handled by AMD's reference cooler without forcing its fan to spin uncomfortably fast. Our third-party cooler's heat pipe only climbs to 35 degrees Celsius.
The FX-8370E can be overclocked stably with a minor voltage increase. It stays well below its official TDP of 95 W, even at 4 GHz. If relatively mainstream performance is acceptable, you can get it from this economical CPU that's easy to keep cool. Still, we can't ignore the fact that lower power consumption is enabled by a reduced clock rate compared to AMD's higher-end models.
AMD FX-8370E at 4.5 GHz
For some reason, everyone seems to want to hit a nice, round 4.5 GHz when they overclock. We're not going to be the exception, even though it might not be an efficient approach to tuning AMD's Vishera core. Really, this CPU can go higher. We even managed 5 GHz long enough to boot into Windows. That required an unsafe 1.5 V though, which is unreasonable for running benchmarks.
Core Voltage
The core voltage at 4.5 GHz was set to 1.315 V in the BIOS, resulting in an actual reading of 1.26 V. This is already quite high, even though it’s still a lot more manageable than 1.5 V.
Power Consumption
At 116 W for the CPU and VRM combined, AMD's FX-8370E is still doing well, especially since the CPU’s contribution is probably under 100 W.
It's annoying that the motherboard and its voltage regulation circuit hit almost 80 degrees Celsius, though. If massive overclocking is your goal, then a more efficient platform would help.
Temperatures
In light of the previous findings, especially the relatively moderate power consumption, a CPU package temperature of 47 degrees Celsius isn't surprising. The processor's heat spreader registers 53 degrees Celsius according to the sensor under it, which isn’t much higher.
The infrared snapshot shows us that a big aftermarket cooler is capable of handling the FX processor's thermal load. Its two fans remain barely audible.
Efficiency
Does it make sense to run at a higher clock rate, trading better performance for increased power consumption? The answer isn’t clear-cut. We do see that the CPU speeds up while essentially consuming the same amount of power when it’s overclocked to 3.5 GHz. This is achieved with mostly untouched BIOS settings and a manual Vcore configured to match AMD's stock reading. Then there's the sweet spot (the point beyond which efficiency starts to decrease). We calculate this to land right around 3.8 GHz. And once you crest 4.2 GHz or so, overclocking AMD's FX-8370E probably doesn't make much sense. It mostly wastes electricity.
Again, if you're gunning for the sweet spot, it's right around 3.8 GHz. That's where you'll get sub-90 W power consumption and respectable performance. If you care more about benchmark results than your power bill, consider calling it a day around 4.2 GHz. Yes 4.5 GHz and higher is possible, but at a certain point you're going to spend too much on a beefy motherboard and high-end cooler, negating the value of overclocking outright.
Once I realized that a loss in efficiency (and not cooling headroom) was limiting the usefulness of my overclocking efforts, I swapped over to a more sensible cooler. The be quiet! Dark Rock Pro is a $90 heat sink. Slap it on top of a $200 processor and you're looking at an expensive, unbalanced combination. So, I gave the same company's $50 Shadow Rock Slim a shot with its fan running at under 600 RPM. Noise still wasn't an issue; I measured 31.4 dB(A) from a distance of 50 cm.
Core Voltage
At 4.2 GHz, we measured an average of 1.18 V. This doesn’t change, even if we drop the BIOS' default 1.215 V setting. However, the system becomes unstable at settings of 1.2 V or less. Avoid undervolting; it's not going to happen.
Power Consumption
We measure a maximum power consumption of 103 W between the host processor and VRM. The FX-8370E’s actual power use is probably just over 90 W. The rest shows up on our infrared pictures as a pretty chain of lights made of voltage regulators and waste heat. It looks a lot like what we already showed you at 4.5 GHz.
Temperatures
The CPU package temperature is 47 degree Celsius, and the heat spreader stays at 51 degrees, according to the FX-8370E’s sensor. That should alleviate the worries of enthusiasts who are afraid of cooling AMD's less heat-tolerant processors. Any $15 budget-oriented solution specified for 90 W or more should be sufficient for AMD’s latest FX model.
Most gamers will happily settle for 4.2 GHz, which yields a "free" 30-percent performance boost. If you want to me more environmentally friendly, that 3.8 GHz is a better bet. re better served by the optimal 3.8 GHz.
We're adding five other CPUs to the mix for our performance comparisons. The FX-8370E represented at three different clock rates takes the total number of bars to eight. Intel’s Core i7-4790K hails from another price range of course, but we still wanted to show what a Haswell-based CPU also capable of scheduling eight threads can do.
Per-clock performance is AMD's most glaring problem area. What happens when an application runs on just one core? The outcome is demonstrated nicely by Cinebench's single-threaded benchmark.
It's made clear in this test that AMD’s new CPU runs at a lower clock rate. Taking scaling into account, we’re basically left with the same performance that AMD’s other FX processors offered. At 4.2 GHz, the FX-8370E nearly catches the non-E version. Haswell blows them both out of the water, though. AMD has nowhere to go but up, and it's a shame that we don't have a Steamroller-based eight-core chip to fold into our chart. The A10-7800 is as much representation as we get from AMD's most modern architecture.
The situation improves when all eight of the FX's integer cores are utilized. Comparing AMD to Intel's Core i7-4790K still yields an ugly outcome, but the Core i5 isn't able to keep up as well in this case.
Well-parallelized tasks tasks allow the CPUs able to handle more than four threads at a time to show off. This is another well-deserved high point for AMD’s new processor.
Unfortunately, real-world tasks typically aren't as dramatically optimized, so they knock AMD's FX-8370E back down to earth. At least the 4.2 GHz configuration almost keeps up with the non-E version.

We could have added more synthetic and application benchmarks, but they wouldn't change the bottom line. Over the past few years, we've covered the Piledriver architecture in much depth, including this Vishera-based implementation. At any given frequency, we know how these CPUs fare, and that won't change due to an optimization for power consumption.
The AMD FX-8370E is an interesting option for processing-intensive applications able to exploit more than four cores. Then again, the competition sells even faster processors available at lower power ceilings, if you have more money to spend.
This brings us to the fun part of the review: the FX’s gaming performance. Typically, when we review a CPU, we use a high-end GPU to avoid bottlenecking host processor performance. That makes sense from a theoretical approach, but it's naturally not going to be practical. At what point does graphics hold back a CPU, anyway?
For this reason, when there are large differences between CPUs, we test each game twice: once with a high-end graphics card (in this case, AMD’s Radeon R9 295X2) and once with a more mainstream graphics card better balanced to match the FX. AMD's Radeon R9 270X or 285 are good matches.
When Does The Graphics Card Become The Limiting Factor?
DiRT 3's ability to scale makes it a good example. If you pair a fast CPU with a high-end graphics card, this title really flies. But if you handicap your platform with a lower-end GPU, the bottleneck becomes obvious.

Now, this isn't to say that matching a Radeon R9 270X to a Core i7-4790K is a good idea. However, if you allow graphics to become your limiting factor, the impact of CPU performance becomes less obvious. If anything, let our exploration serve to better inform you how component choice can dramatically alter the outcome of benchmark results.

We perform the same exercise in Battlefield 4, even though we measured a performance difference of only 30 percent with the high-end graphics card. First, let’s take a look at the original test:

In single-player mode, a Radeon R9 285 is enough to essentially level the playing field. It's our limiting component at Ultra detail settings and a 1920x1080 resolution. Of course, most folks still looking at Battlefield 4 are involved in the more CPU-taxing multi-player component. Unfortunately, that's difficult to benchmark reliably.

BioShock Infinite doesn’t have the reputation of being hard on hardware, but that doesn’t mean that the combination of a high-end graphics card and AMD's FX-8370E makes sense.

Once again, capping performance with a more mainstream GPU masks the potential of our various host processors. In the real-world, pairing a Radeon R9 285 and FX CPU makes sense. Substituting in a Core i7 won't yield dramatically better frame rates until you also step up to a much faster graphics configuration.

Gaming at 3840x2160 Resolution
It’s common knowledge that massive resolutions almost always lead to GPU bottlenecks. Consider it an exaggeration of what we just saw. Even at maxed-out settings, there’s almost no difference between the CPUs in spite of the relatively high frame rates. It basically doesn’t matter what CPU you pick because the graphics card creates the performance ceiling.

To put it nicely, the FX-8370E is a true middle-of-the-road CPU. Using it only makes sense as long as the graphics card you choose comes from a similar performance segment.
Depending on the game in question, AMD’s new processor has the potential to keep you happy around the AMD Radeon R9 270X/285 or Nvidia GeForce GTX 760 or 660 Ti level.
A higher- or even high-end graphics card doesn’t make sense, as pairing it with AMD's FX-8370E simply limits the card's potential.
Does the new FX-8370E represent a step forward for AMD’s FX line of processors? Giving a clear answer is somewhat of a challenge. Our test sample performs about the same as its predecessors at similar clock rates, but draws less power doing so. However, we only tested one unit, and it’s always possible that AMD sent us a choice sample. Plus, there is almost certainly variance from one -8370E to the next. And so it's hard to know if the FX-8370E is actually better. Based on our measurements, all we can say is that the CPU in our lab is better than those that came before.
Directly comparing the FX processors in general to the FX-8370E specifically, and then adding Intel's competing models, does give us some clearer answers, though. Nothing changes, which is to say that Intel is still in the lead. It’s plain to see that AMD hasn’t invested in real innovation, at least not in this model. Instead, its recomposes existing ingredients into a somewhat more savory dish.
Bottom Line
The AMD FX-8370E isn’t a bad CPU, and it actually provides decent and acceptable performance in certain areas. A comparison to its direct competition does make it abundantly clear that we're dealing with an old architecture demonstrating noticeable weaknesses, though. The only way to sell it at this point is via lower prices.
This is exactly where we encounter our main problem. For the AMD FX-8370E’s performance, an MSRP of $199 is way too high. In Europe, the first vendors have the new processor listed for 190 Euros, and it just doesn’t offer enough performance to warrant that. Locally, you'll find the processor right around $200.
We’d like to revisit our introductory comments about the importance of street pricing. In the end, what we wrote then is the bottom line now. Whether AMD’s FX-8370E makes sense depends on what you pay for it. At $200, we think AMD needs to do some trimming.
Clarification: This article was updated at 6:25 pm ET to clear up confusion about sampling concerns and our editorial stance on AMD’s greater CPU business.

























