Today we are taking a look at a custom iBuyPower AMD FX Ultimate PC designed to showcase the capabilities of the AMD FX 9590 CPU. The specifications are below:
- Processor: AMD FX 9590 8-Core Black Edition at 4.7 GHz
- Video Card: 2x AMD HD 7990 in CrossFire
- Motherboard: ASRock 990FX Extreme9
- System Memory: 16GB DDR3 2133 MHz AMD Radeon RG2133 Gamer Series
- Power Supply: 1200W Corsair AX1200i 80 PLUS PLATINUM
- SSD: 2x240GB Solid State in RAID 0 (By: Corsair) (Model: Force GS)
- Storage: 2TB Seagate (7200 RPM)
- Optical Drive: LG 24x Dual Format/Double Layer DVD±R/±RW + CD-R/RW
- Cooling: NZXT Kraken X40
- Chassis: NZXT Phantom 630 Full Tower
- OS: Microsoft Windows 8 (64-Bit Edition)
Keep in mind, these specifications cannot be chosen on iBuyPower?s online configurator, but at the time the PC was sent to us it was priced at a theoretical $3,999. Configuring a similar build now would be cheaper due to a variety of discounts iBuyPower currently offers for many of the components.
The AMD FX 9590 retails by itself for around $350 at the time of this writing, or $399 with a bundled liquid cooling system. However, AMD?s suggested pricing for the FX 9590 is $306. It should be noted that the FX 9590 is rated to have a TDP of 220W, and tends to run very hot. This is the reason AMD?s preferred method of offering the CPU is either through a system builder such as iBuyPower, or bundled with a liquid cooling system since many air cooling systems would not sufficiently cool the CPU. Also, the FX 9590 cannot be run on all 990FX chipset motherboards, only certain motherboards which have been rated for it due to the sheer amount of power it requires.
Now, having read that, most enthusiasts would slowly back away from the FX 9590 on the basis that any CPU with that many warnings cannot possibly be worth considering. Many enthusiasts have likely also read various FX 9590 reviews and have concluded that it either doesn?t perform as well as they like, uses too much power, or runs too hot. These are all valid concerns that we shared, and we spoke to AMD at length about this PC and the FX 9590 in order to fully understand why the FX 9590 exists and its purpose in the market.
The FX 9590, plainly stated, is a binned and overclocked FX 8530 CPU. AMD?s idea with the FX 9590 was not to compete with Intel, but rather to create the most powerful 32nm Vishera-based CPU they possibly could. We were told that the FX 9590 is not expected to sell at high volume and was created mostly for the reason of ?we did it because we could?. It was designed to appeal to AMD enthusiasts who want to get the best, most powerful CPU the company can create (in this gen), as well as gamers and other enthusiasts who want to maximize performance but are hesitant to overclock an FX 8350 to the same frequency as the FX 9590.
The FX 9590 is a Black Edition and comes with an unlocked multiplier, though it?s not expected that most FX 9590 owners will take advantage of that since this CPU comes out the gate running hot and power hungry as it is.
To really sum it up, AMD released an 8-core CPU running at 4.7 GHz just to see how hard they could push the architecture with no regard to power consumption or heat. While this may not be an optimal approach for someone who likes to maintain a respectable power to performance ratio, for someone who only cares about maximizing performance, this is a reasonable strategy. Depending on how one utilizes their processor, the FX 9590 may or may not be better than a similarly priced Intel CPU such as the i7-4820K or i7-4770K in terms of performance.
Although the focus of this review is the FX 9590, we would like to cover the fit and finish of the PC. iBuyPower did a spectacular job assembling the PC.
The cable management is well done and the interior is very clean. There are no cheap components in this PC and it is very apparent.
The NZXT Phantom 630 is a great looking chassis with a good set of features (though we won?t expand into that since this isn?t a chassis review). The power supply is Corsair?s Platinum rated 1200W PSU, which should be more than enough to handle the power hungry FX 9590 and dual HD 7990 GPUs, despite having a combined TDP of 970W.
The dual SSDs in RAID 0 are blazing fast and provide a respectable almost half GB of storage, and are complemented by the 2 TB Seagate HDD. The AMD RAM looks good, performs well, and fits in nicely with this all-AMD system.
The system came totally bloat-ware free, the number of preinstalled applications was limited to just drivers and some Microsoft C++ libraries.
The desktop also came nice and clean, devoid of anything other than the Recycle Bin, a Readme for the system, and a motherboard chipset guide.
One thing we did notice which was strange, though this is not specific to this particular computer, is that the HD 7990 GPUs support PCIe 3.0, but the 990FX chipset, which is AMD?s highest end motherboard chipset, does not support PCIe 3.0; it only supports PCIe 2.0. This discrepancy is a little awkward, but we did not see any significant performance hit as a result of the PCIe 2.0 slots.
Anyways, we begin our benchmarks with some synthetics.
First up is 3DMark11:
The Entry preset yields a result of E15137.
The Performance preset yields a result of P15512.
The Extreme preset yields a result of X9435. This number seems to be in the approximate range of an i7 4770K with 2x NVIDIA GTX 780 GPUs in SLI according to 3DMark?s website. Comparing retail prices of that CPU and those GPUs with our FX 9590 and dual HD 7990 GPUs, the i7 + 2x GTX 780 combo comes out a fair bit cheaper. This may be significant when doing a cost analysis, but remember that the purpose of this PC was to showcase the FX 9590, so the dual HD 7990 GPUs were definitely and purposely overkill, something that will become much clearer as our review progresses.
Next, we ran PCMark 8:
PCMark 8 yielded a Home benchmark score of 4363.
We then ran Unigine Heaven 4.0 on the Extreme preset:
Our system managed 41 FPS in Unigine and a score of 1035.
We then tested our SSD RAID 0 array and 2 TB HDD in AIDA64:
The top half of the results are from the SSD array, and the bottom half is for the 2 TB HDD.
The SSDs had linear and random reads ranging from 911 MB/s to almost 1056 MB/s, with average read access taking a miniscule 0.09 ms.
The HDD had linear and random reads ranging from 94 MB/s to 185.5 MB/s, with average read access taking 15.60 ms.
We also tested out the RAM in AIDA64:
Looking at some game benchmarks, we noticed the system seemed CPU limited in many cases in 1080p, though this stopped being the case once we stepped it up to 4K (3820×2160). Also, for a point of reference, we ran our benchmarks using the Catalyst 13.10 Beta drivers for the iBuyPower, and the GeForce 327.23 drivers for the Intel + NVIDIA system we compared with.
We begin with DIRT Showdown:
It is very apparent that in 1080p, the 2nd HD 7990 isn?t helping. In fact, due to the overhead of CrossFire, performance actually takes a slight hit. The CPU limitation becomes obvious once the FX 9590 is downclocked from 4.7 GHz to 3.4 GHz, as the framerate drops significantly. Running at 3.4 GHz, the framerate is only slightly higher than that of an i7 3820 paired with a GTX 670, despite the single and dual HD 7990 configurations. However, anyone who purchases a system with dual HD 7990 GPUs would not likely be playing on a single 1080p monitor, and the iBuyPower puts up respectable numbers at the 4K resolution. Again, remember what we stated earlier, which is that this system was designed to force the FX 9590 as the bottleneck to be able to fully demonstrate its capabilities.
We also ran DIRT: Showdown with 8f16x EQAA multisampling turned up on the 4K resolution, at which point the CPU is definitely not the limiting factor, and GPU scaling becomes very apparent.
Next is Hitman: Absolution, another AMD title:
Unfortunately, the CPU limitation is even more apparent in Hitman, with the iBuyPower getting no benefit from the 2nd GPU in either 1080p or at 4K. The i7 3820 and GTX 670 perform almost identically to the iBuyPower at 1080p. Again, the CPU limitation is proven by downclocking the CPU and seeing the framerate take a large hit. Thankfully these are all playable framerates, though we did notice some tearing and frame pacing issues.
Sleeping Dogs is next up on our list:
Sleeping Dogs was not as CPU bound as the previous two titles, though it certainly plays a vital role. Scaling from 2 GPUs down to 1 GPU causes a noticeable drop in framerate in 1080p, though it becomes much more apparent at 4K when it is cut in half, down to an unplayable 22 FPS. The CPU limitation is still highly apparent when taking into account that downclocking the CPU results in dropping to a lower framerate than you get when removing an HD 7990. The GPU horsepower is clearly a requirement though, as the iBuyPower clearly beats the i7 + GTX 670 combo.
Bioshock Infinite is our next title:
The CPU limitation is clear here, as going from 2 to 1 GPUs makes no difference in 1080p, and while it does make a difference at 4K, the change is not nearly as drastic as it was on some of the games we?ve covered so far. The framerates are all very playable though, so there?s no noticeable change in experience. As usual, downclocking the CPU results in lower framerates, but one might wonder why the framerate is lower with dual HD 7990 GPUs than with a single one. We suspect it has to do with CrossFire overhead becoming much more significant at the lower frequency.
Finally, we tested out Tomb Raider, definitely an AMD title:
This is where the system really shines. Tomb Raider is very GPU bound, and the difference between 1 and 2 GPUs is massive. Cranking the settings up to 4K also heavily taxes the GPU, 4x the resolution of 1080p results in approximately 1/3 of the framerate. Downclocking the CPU makes very little difference, and the gap between the iBuyPower and the i7 + GTX 670 is huge. Clearly the CPU is not the limiting factor in Tomb Raider. It should be noted that while 2 HD 7990 GPUs provide a very playable experience at 4K, a single HD 7990 results in framerates that are barely playable.
As we mentioned a few times already, the fact that this system tends to be CPU limited should not necessarily be taken as a fault of the FX 9590. Anyone willing to spend $1400 on their graphics cards will likely (hopefully) pair them with a CPU at a higher price point as well. Considering the FX 9590 has a suggested retail price of $306, it would be prudent to consider something more powerful such as an i7 4930K to complement the GPUs. The reason this system came with this level of graphics horsepower was to make sure the system would not be bottlenecked by any component other than the FX 9590.
We also took some power consumption readings. As stated earlier, the FX 9590 was not designed to compete in terms of power efficiency, and paired with two power hungry HD 7990 GPUs, we weren?t expected a cool and low power system. Our predictions were validated during testing:
- At idle: 170W
- CPU at full load (using Prime95?s In-place Large FFTs test): 420W
- GPU at full load (using Furmark at 1920×1080, 8x MSAA): 840W
- CPU + GPU at full load (using Prime95 + Furmark): 950W
Thankfully Corsair?s 1200W Platinum PSU was able to handle the system at full load with no issues.
We also ran BSN?s very own AMDMsrTweaker tool to tell us about the FX 9590?s P-states. The P-states are what determine how the CPU scales up and down when at load or at idle and the various in-betweens.
While the FX 9590 has 7 P-states, readers will notice that the top 3 all run at the same voltage levels. The voltage plays a large part of the CPU?s power consumption, and the fact that the top 3 states all ran at the same voltage seemed strange to us. Increasing voltage on the CPU allows for greater overclocks, but results in diminishing returns: the more you raise the voltage, the less the increase in voltage allows you to increase the CPU?s clockspeed. Since the FX 9590 is a factory overclocked FX 8350, we figured this was a result of a lazy overclock with no regards to power consumption. We asked AMD about it, and it turned out we were partially right.
Since the silicon on all processors have some variation, AMD decided that it would be safest to operate the FX 9590 CPUs at a higher voltage to make sure all FX 9590 CPUs that rolled off of factory lines would be able to run at the high frequency of 4.7 GHz without issue, as having too low of a voltage would cause system instability. Rather than overclocking laziness, it is more of a caution issue to make sure all FX 9590 CPUs run stably. In general it?s preferable to see different voltages at every P-state to allow a processor to really take advantage of the P-states? power saving features, but AMD did say that they weren?t concerned with the FX 9590?s power consumption, so that wasn?t really a consideration in this case.
The iBuyPower AMD FX Ultimate PC is a great computer. Our version had seriously beefed up specifications, and a price point to match. The system flew through most of our benchmarks, only limited by the star of the show: AMD?s FX 9590 CPU. We won?t fault them too much for it since the CPU bottleneck was the purpose of the build, but it was clear that the FX 9590 CPU should be paired with more comparable components to make a truly balanced PC.
Whether the FX 9590 is a better choice over an i7-4820K or i7-4770K is up to the buyer, the idea of having an 8-core CPU running at 4.7 GHz is definitely an exciting one. As long as one is not concerned with power consumption, the FX 9590 is a solid performer, and one that really shines in multithreaded applications. For the AMD fan, there?s no better alternative: the FX 9590 is as good as it gets, and pairing it with dual HD 7990 GPUs and AMD?s RG2133 Gamer Series RAM yields a very potent PC, especially for playing at resolutions higher than 1080p. However, if you?re not loyal to one brand, and you insist on having dual HD 7990 GPUs (or any $1000+ GPU solution), we recommend looking at other, more expensive CPUs to make sure you?re squeezing as much performance out of those GPUs as possible.