Sign in with
Sign up | Sign in
Phenom II 955 Versus Core i7 920: Gaming Value Compared
By ,
1. Introduction: A Little Background

We recently published a review of Cyberpower's Gamer Dragon, a Phenom II 955-based gaming PC equipped with DDR3 RAM and a pair of Radeon HD 4890s in CrossFire. The Cyberpower Gamer Dragon was delivered to us with a nice 3.6 GHz overclock, and when we considered what we should compare this system to, our first thought was our recent $1,300 System Builder Marathon (SBM) machine. With an Intel Core i7-920 and two GeForce GTX 260s in SLI, the cost of the components for the two builds was similar.

The comparison also addressed one of the primary complaints we received when we chose the i7 over the Phenom II for our SBM build: that the Phenom II would have been a better choice for a gaming rig, because its lower price tag would have enabled us to add better graphics cards with the budget available. Since the graphics subsystem is often the limiting factor for game performance, a cheaper AMD processor coupled with a higher-powered graphics card would have offered a better fit for gaming, according to the reader feedback.

On paper, two Radeon HD 4890s best a pair of GeForce GTX 260s, so I was satisfied that the scenario would make for a great comparison of Phenom II versus Core i7 gaming value. I merrily proceeded to take benchmarks, record the findings, and form conclusions based on the data. It seemed like the perfect opportunity to investigate the matter, and after all was said and done, my conclusion was that the Core i7-920 beat the Phenom II by a fairly substantial margin. When a budget allows for a powerful dual-graphics card setup, the Core i7 appeared to be the ideal choice, while the Phenom II was a better pick for machines in a lower price range, which the Core i7 can't touch. 

Well, the Cyberpower Gamer Dragon article probably generated more forum feedback than any other piece I've written. Normally this would be a good thing, but this time my testing and subsequent conclusion left a lot of people upset. According to a number of folks, there were three main sins I committed in my review of the Phenom II machine:

1. Different Graphics Card Manufacturers

This ties in directly with concern number two below. As both of the test systems used different brands of graphics cards (one sporting Radeon GPUs and the other with GeForce GPUs), the results depended on games that demonstrated a performance preference for one of these architectures.

2. Selection of Game Benchmarks

We use four game benchmarks in our SBM series, which we selected using a number of factors, including popularity, significance, and variety. Those games are Crysis, Far Cry 2, World in Conflict, and Stalker: Clear Sky. Keep in mind, games are only a small part of the PC performance that we test in our SBMs, so up until this point we haven't noticed any problems with this selection of titles.

However, as mentioned, we pit Radeons against GeForces, opening ourselves up to any preference a specific game title might have for either architecture. It has also been suggested that our game selections are overly CPU-dependent, thus giving the Core i7 an edge.

3. The Cyberpower Gamer Dragon May Not Represent Ideal Phenom II Performance

It has also been suggested that our Cyberpower Gamer Dragon test sample wasn't working up to snuff. Concerns have been raised that the memory performance results as tested by the SiSoft Sandra benchmark were lower than they should have been. The Gamer Dragon also uses the Gigabyte GA-MA790XT-U4DP motherboard, which is equipped with AMD's 790X chipset limited to one x16 and one x8 PCI Express (PCIe) 2.0 slot. The AMD 790FX would better represent ideal Phenom II performance, as it has two full PCIe x16 slots--one for each of the graphics cards in CrossFire.

I do believe these three points cover most people's concerns with the Cyberpower article, and since these issues deserve to be investigated further, we came up with a more even-handed plan to use in this follow-up review. Instead of looking at the Cyberpower Gamer Dragon, we'll concentrate specifically on Intel Core i7 versus AMD Phenom II gaming value using our own builds.

2. Phenom II Versus Core i7-920: Competing System Cost Analysis

With these potential issues in mind, how do we set up a fair comparison between similarly priced Phenom II and Core i7 systems?

To truly realize the cost difference, we need to list and price the PC components that are linked to either the Phenom II or Core i7 CPU.

For instance, the case, power supply, hard drives, optical drives, etc. are generic and can be used for any kind of CPU. The components that are specific to an AMD or Intel platform would be the CPU, motherboard, RAM, and CPU cooler.

Let's start with our Core i7 build. In this case, we will use the same one we set up for the last SBM:

Core i7: Core System Components
Motherboard

DFI LANParty Jr X58-T3H6 Micro-ATX
Intel X58/ICH10R, LGA1366

$199.99

Processor

Intel Core i7-920
Four Cores, 2.66 GHz, 8 MB Cache

$279.99

Memory

G.Skill 10666CL7T 6GBPK
Triple-channel memory kit, 3 x 2 GB

$94.99

CPU Cooler

Xigmatek Dark Knight S1283

$39.98

 

Total Cost

$614.95


Why did we choose these parts for the i7 system? For starters, the Core i7-920, which retails for less than $300, is the cheapest Core i7 CPU by far. The next step up, the Core i7-940, costs about twice that for only a couple hundred more megahertz. Intel's Core i7-920 is also known to be a fantastic overclocker, so it's an easy choice.

The motherboard we chose is DFI's X58-T3H6 because we needed a MicroATX board to fit in our previous SBM system, but any quality X58-based motherboard--a requirement of the Core i7 CPU--will be in this price range. The cheapest Core i7 board we could find was the MSI X58 for $170, which is $30 cheaper than the DFI board we chose.

For RAM, we chose the most cost-effective module we could find with CAS 7 timings, G.Skill's PC3-10666. Because the Core i7 benefits from running in triple-channel mode, we needed a triple-channel kit to squeeze the most performance out of it. For under $100, the kit isn't a huge expense.

Finally, a good Core i7 build needs a solid cooler, and Xigmatek's Dark Knight will do the job for $40.

So, the base cost of our Core i7 build comes to about $615. Let's see how much a Phenom II build will run us:

Phenom II: Core System Components
Motherboard

ASUS M4A79T Deluxe
AMD 790FX, AM3

$199.99

Processor

Phenom II X4 955 Black Edition
Four Cores, 3.2 GHz, 6 MB Cache

$245.00

Memory

G.Skill 10666CL7T 6GBPK
Dual-channel memory kit, 2 x 2 GB

$64.99

CPU Cooler

Zerotherm NV120

$49.99

 

Total Cost

$539.97


A little more thought went into the AMD system. Why didn't we go with a cheaper DDR2 motherboard, less expensive RAM, and a lower-end Phenom II processor? We'll tell you, but first we have to consider the price of the components we've chosen.

At $540, the Phenom II X4 955 system's core components cost about $75 less than the Core i7 system's core components. This represents the price difference between a pair of Radeon HD 4870 and Radeon HD 4890 cards. With this in mind, let's consider what we could've done differently.

First of all, the processor we chose was the Phenom II X4 955 Black Edition. There are a number of good Phenom II processors from which to choose, ranging from the cheaper AM2+ models that use DDR2 to the triple-core X3 models that can use either DDR2 or DDR3. These CPUs offer a low price and reasonable performance (especially for games, since many game engines tend not to use more than three CPU cores, anyway).

If that is the case, why did we go with the most expensive Phenom II CPU for this build? The decision hinged on this: what could we get with the extra cost savings? We already have enough money left over in our budget  to upgrade from a pair of Radeon HD 4870s to two Radeon HD 4890 cards in the Phenom II system. If we go for a cheaper CPU, saving another $100 or even $200 doesn't allow us to upgrade to a more powerful graphics solution because the next step up is a pair of Radeon HD 4870 X2 cards. These would cost $400 more, which is impossible with our budget.

The rest of the components fall into place from there. The Asus M4A79T Deluxe motherboard is certainly not the cheapest model, but it does sport the 790FX chipset and dual full-speed PCIe x16 slots for the CrossFire graphics cards. Cheaper motherboards with the 790FX chipset can be found for as low as $135, such as the DFI LANParty DK 790FX-M2RS, which is about $45 less expensive than our test setup. Since the cheapest X58 motherboards are about $30 less expensive than the one we selected, we're still in the same range here.

The choice of RAM is a notable deviation from the Core i7 components as well. Consider that a Phenom II system requires paired modules to operate in dual-channel mode, while the Core i7 requires one more module for triple-channel mode. While two 2 GB sticks of DDR3 RAM will leave us 2 GB short in comparison to the Core i7 system, we know from experience and testing that the real-world difference between 4 GB and 6 GB is almost negligible in most situations. Additionally, the cost savings from less RAM allows us to allocate more cash toward the Phenom II-based system's graphics cards.

Finally, the CPU cooler. Zerotherm's NV 120 is a nice unit, but most CPU coolers worth their salt will be in the same price range, so insert your favorite flavor of cooler if you disagree with the model selected.

We did have some forum users mention that we might want to spec out an even cheaper AM2+ Phenom II system and use the savings toward a solid state drive (SSD) instead of a graphics upgrade, but we're not sure that would provide enough of a tangible performance benefit to offset losing the 790FX chipset and its dual PCIe x16 graphics slots, the faster Phenom II X4 955 processor, and the speedy DDR3 RAM. It's certainly an option, though, and if you're simply looking to build an even cheaper gaming system, the AM2+ Phenom IIs are a very good choice.

Now that we understand what we're building, let's have a final look at all of the components in our test systems.

3. Test Systems And Benchmark Setup

First off, we'd like to thank Long and the crew at MemoryExpress and Wayne at Computer Boulevard in Winnipeg, Canada, who helped us with some last-minute test-equipment requirements; we couldn't have finished this review without their assistance.


Phenom II System
Core i7 System
CPU

AMD Phenom II X4 955 Black Edition
3.2 GHz, 1.35v, FSB-200, 2000 MHz HT link
Overclock: 3.706 GHz, FSB-218,
2,180 MHz HT link

Intel Core i7-920 2.66 GHz,
133 MHz Bclk, 1.36 V (load)
Overclock: 3.44 GHz at 1.296 V (load),
172 MHz Bclk

Motherboard

ASUS M4A79T Deluxe,
BIOS 1503, AMD790FX 

DFI LANParty Jr X58-T3H6 Micro-ATX
Intel X58/ICH10R, LGA1366

Networking
Onboard Gigabit LAN controller

Onboard Gigabit LAN controller

Memory

G.Skill 10666CL7T 4.0 GB DDR3-1064
  2x 2.0 GB, CAS 8-8-8-19, 1.56 V
Overclock: DDR3-1454 at 1.675 V, CAS 8-8-8-19

G.Skill 10666CL7T 6.0 GB DDR3-1064
3 x 2.0 GB, CAS 8-8-8-19, 1.56 V
Overclock: DDR3-1378 at 1.56 V, CAS 8-8-8-19

Graphics

2 x HIS Radeon HD 4890 in CrossFire
850 MHz GPU, 2,150 MHz RAM, 1 GB Per Card

2 x HIS Radeon HD 4870 in CrossFire
750 MHz GPU, 1,800 MHz RAM, 1 GB Per Card

Hard Drive

Seagate Barracuda ST31500341AS
1.5 TB, 7,200 RPM, 16 MB Cache, SATA 3.0 Gb/s

Western Digital Caviar Black 640GB
7,200 RPM, 32 MB Cache SATA 3.0 Gb/s

Power

PC Power & Cooling Silencer 750 Quad S75QB
750 W, ATX12V 2.2, 80-Plus Certified

PC Power & Cooling Silencer 750 Quad S75QB
750 W, ATX12V 2.2, 80-Plus Certified

Software and Drivers
Operating System
Microsoft Windows Vista 64-bit 6.0.6001, SP1
DirectX versionDirectX 10
Graphics Drivers

Nvidia GeForce 185.85, ATI Catalyst 9.6


Notice that we had to use different hard drives in our test systems. This was unavoidable, as we were running concurrent benchmarks and we didn't have two of the same model drives on hand. However, SiSoft Sandra demonstrated that the drives performed very similarly to each other and we're confident that that the disparity did not cause any performance variations.

Our Phenom II overclock went very smoothly and, using identical timings as the Intel Core i7 overclock, we managed to get the Phenom II 955 to 3.7 GHz with little trouble. With a slight CPU core voltage increase to 1.42 V, the memory voltage increased to 1.675 V, and the northbridge voltage increased to 1.2 V. We set the CPU multiplier to 17x and the front side bus (FSB) to 218 MHz, resulting in an overclocked CPU speed of 3,706 MHz. This is about 100 MHz faster than the overclocked Cyberpower Gamer Dragon system we tested. From what we've seen, Phenom II 955s can get to 3.6-3.7 GHz with little trouble and minor voltage and heat increases, but they hit a wall soon after.

We have also benchmarked many more games this time around to satisfy folks who didn't think we had enough of a cross-section in the Cyberpower article. Here are the games and their settings:

Benchmark Configuration

3D Games

Crysis

Patch 1.2.1, DirectX 10, 64-bit executable, benchmark tool
Test Set 1: High Quality, No AA
Test Set 2: Very High Quality, No AA

Far Cry 2

DirectX 10, in-game benchmark
Test Set 1: Very High Quality, No AA
Test Set 2: Ultra High Quality, 4x AA

Stalker: Clear Sky

Average of 4 segments "A-Tested Object"
Test Set 1: High Preset, DirectX 10, EFDL, no MSAA
Test Set 2: High Preset, DirectX 10, EFDL, 4x MSAA

World In Conflict

Patch 1009, DirectX 10, timedemo
Test 1: Very High Details, No AA/No AF
Test 2: Very High Details 4x AA/16x AF

Tom Clancy's H.A.W.X.

DirectX 10.1, in-game benchmark
Test Set 1: Highest Settings, No AA
Test Set 2: Highest Settings, 8x AA

Fallout 3

Custom THG Benchmark
Test Set 1: Ultra High Quality, No MSAA, No AF
Test Set 2: Ultra High Quality, 4x MSAA, 15x AF

Left 4 Dead

Custom THG Benchmark
Test Set 1: Highest Settings, no MSAA
Test Set 2: Highest Settings, 8x MSAA, 16x AF

Prototype

Custom THG Benchmark
Test Set 1: High Quality, No MSAA, No AF
Test Set 2: High Quality, 4x MSAA

4. Synthetic Benchmarks

The Sandra tests show us a few things: first of all, it looks like the Cyberpower CPU was performing as it should for a Phenom II X4 955 overclocked to 3.6 GHz. The results are close to our new Phenom II system when overclocked to 3.7 GHz.

However, the memory bandwidth of the Cyberpower Gamer Dragon we previously reviewed does look low compared to our Phenom II test system. The new test system's latencies were lower, but it's still a little out of whack. It will be interesting to see if this results in different benchmark numbers.

5. Game Benchmarks: Crysis

Let's start with Crysis. In the Cyberpower review, the game was somewhat dominated by the Intel Core i7 system and its dual GeForce GTX 260 graphics cards:

Right out of the gate, we're seeing a huge difference between these new results and the Cyberpower Gamer Dragon results. Here, while the Core i7 system is showing a slight advantage at lower resolutions when overclocked, the Phenom II system is holding its own where it counts at 1920x1200.

And, when anti-aliasing (AA) is applied...

...the Phenom II system with its Radeon HD 4890 cards even wins by a few frames.

6. Game Benchmarks: Far Cry 2

Here, the Core i7 shows a solid lead over the Phenom II system, especially when overclocked. Both systems also offer perfectly playable performance, even when clocked at stock speeds.

With AA enabled the situation doesn't change much. Far Cry 2 appears to favor the Core i7 architecture, especially when overclocked. But as we mentioned before, both systems are perfectly playable at stock speeds.

7. Game Benchmarks: World In Conflict

World in Conflict shows a notable preference for the Core i7, especially when overclocked. The Phenom II delivers smooth frame rates, though. 

With 4x AA and 16x anisotropic filtering (AF) added, the margin is lowered, but the Core i7 system still demonstrates better numbers. However, we do have to mention that the Phenom II is still playable and smooth.

8. Game Benchmarks: Stalker Clear Sky

This game likes the overclocked Core i7 CPU at lower resolutions, but at 1920x1200, the results are very close.

With 4x AA applied, the results are still very tight. The Phenom II is a couple frames per second (FPS) ahead at 1920x1200.

9. Game Benchmarks: Tom Clancy's H.A.W.X.

The HAWX benchmark shows a lead in favor of the Core i7 system. Both systems will provide extremely smooth game play in this CPU-limited title. It should be noted that GeForce cards don't handle the ambient occlusion (AO) setting in this game nearly as well as Radeons do.

Even with 8x AA added, the slowest system is achieving almost 90 FPS. The Core i7 is generating higher numbers.

10. Game Benchmarks: Fallout 3

As seen in World in Conflict and HAWX, Fallout 3 is showing an advantage for the Core i7 system, but it's somewhat meaningless as far as real-world playability is concerned. All of the systems are delivering high frame rates in this CPU-limited title.

Fallout 3 seems totally unconcerned with any extra workload that 4x AA demands, demonstrating a notable CPU limitation rather than a GPU limitation.

11. Game Benchmarks: Left 4 Dead

The Left 4 Dead results indicate a tie, with the Phenom II showing a tiny lead at stock speeds, and the Core i7 squaring things up when overclocked. Once again, all systems are running the game smoothly.

Adding AA is, unsurprisingly, free in this CPU-limited title.

12. Game Benchmarks: Prototype

Prototype is demonstrating the same proclivities seen in the past few games: a CPU limitation, with the Core i7 taking a lead when overclocked.

There is nothing new to see here, as AA is a freebie in this CPU-limited game when using graphics cards as powerful as the Radeon HD 4870 in CrossFire.

13. Conclusion

Well, I'll start off by apologizing to AMD and our readers for some of the conclusions I drew from the Cyberpower Gamer Dragon review.

I deserve a good measure of blame for rationalizing the difference that we saw when comparing a machine with Radeon HD 4890s against a machine with GeForce GTX 260s. While I had considered this when I was benchmarking, I had also convinced myself that the power of the Radeon HD 4890s would compensate for any game engine preferences. I was also mistaken about the dramatic extent that Crysis favored the GeForce GTX 260 over the Radeon HD 4890. This represents one quarter of the games in in our condensed gaming benchmark suite used in the SBM series, and I will definitely push to have the benchmark suite revised to better handle this disparity.

It also appears that our overclocked Cyberpower Gamer Dragon sample was not performing as well as it should have been, with our Phenom II test system routinely outperforming it at stock speeds. It was odd that the synthetic CPU benchmarks demonstrated that the processor was working as expected, but the synthetic memory benchmark showed an anomalously low result for the Gamer Dragon. While it's true that our Phenom II test system had the advantage of slightly faster RAM and the 790FX chipset with its dual PCIe x16 slots, the game performance results were notably below what our Phenom II system demonstrated.

In any case, I am more than happy to revise my previous conclusion and say that a Phenom II can make a fine gaming rig at the same price as a Core i7 system. The money saved by choosing the Phenom II setup instead of a Core i7 system can be applied to the graphics subsystem, helping the Phenom II stay competitive in gaming environments. Across all of the games and all of the resolutions we tested, the Phenom II system delivered frame rates just as playable as the Core i7 system.

However, at the end of the day, I'd go with the Core i7 system if someone were to hand me enough money for one of these builds. Why? Primarily because I do a lot more than game on my PC. AMD's got game, for sure, but if I have to make a choice between gaming and everything including gaming, I find it difficult to rationalize going with the Phenom II. In all of the game benchmarks we took, the Core i7 rarely lost by a notable margin, but there were quite a few instances where it pulled ahead, especially when overclocked. On the other hand, in our application testing as demonstrated in previous articles, the Core i7 leads the Phenom II by a more substantial margin. And there are secondary factors as well. Note that our Core i7 overclock was conservative compared to the Phenom II overclock, giving us more potential with which to play. Also, the Core i7 offers the flexibility to use both Nvidia and AMD GPUs in multi-card configurations, wheres AMD CPUs are either limited to Radeons in CrossFire for multi-card setups or GeForces in SLI if you go with an Nvidia chipset--but never both.

Does this mean we recommend staying away from AMD CPUs? Not at all. But AMD's offerings shine below the Core i7's price segment, especially when budgetary constraints mean that buying a Core i7 requires sacrificing a balanced graphics card. We can build a killer Phenom II system for hundreds of dollars less than a Core i7 if we go with an AM2+/DDR2 Phenom II or the Phenom II X3. Also, even more money can be saved if you're willing to settle for a motherboard with one of AMD's more value-oriented chipsets. If this kind of low-cost system is overclocked, we should see gaming results very similar to the ones we saw in our Phenom II X4 955 tests today.

This is where the AMD option makes the most sense: not by going toe-to-toe against the Core i7 at higher prices, but by offering similar gaming utility at a much lower cost. This is also where AMD pulls a rabbit out of its hat, by serving up an alternative for budget gamers who want to build a cost-effective gaming rig.