For this quarter’s SBM, we built a gaming PC with a GeForce 980 paired with a Core i5. Can it keep up with our i7-based builds from previous SBMs?
System Builder Marathon Q2 2015
Here are links to each of the five articles in this quarter’s System Builder Marathon (we’ll update them as each story is published). And remember, these systems are all being given away at the end of the marathon.
To enter the giveaway, please fill out this SurveyGizmo form, and be sure to read the complete rules before entering!
$1600 Gaming PC
So this is my first foray in tackling the System Builders Marathon. I’ve built plenty of machines over the years, but my background is mostly IT and my emphasis was usually to build with a good CPU. When I was offered the chance to work on this quarter's System Builder Marathone, I was told that the SBM would be a totally different experience. They were right, but with a little help from Thomas who answered my questions regarding our System Builder Marathon rules, I was ready to spend some money.
With Paul and Don not available for this quarter’s System Builder Marathon, I knew I had to step into some pretty big shoes. With $3200 at our disposal, Thomas and I were tasked to build two $1600 machines each. One of the two machines was to be ATX-based, while the other would be mini-ITX. To change things up a bit, Thomas would focus on building out a CPU-driven machine and I would focus on a graphics-heavy build.
Oh the fun we had!
Is $1600 an unrealistic target? I’m pretty sure that's not an over-the-top price for a great machine, even if some of us would have to collect the pieces over a period of time (or save up a bit longer for the big payoff). Regardless, the budget Thomas and I were given is pretty fair, especially for the level of performance we were aiming for.
By the way, I’m simply dubbing this first machine “Big Build”. It’s a tower-based system running Windows 8 and housing an ATX-sized motherboard, which should leave plenty of room for any future upgrades or expansion.
Here’s what I chose for Big Build:
- Platform Cost: $1348
- Total Hardware Cost: $1499
- Complete System Price: $1599
In order to accomplish the goals I set forth for Big Build, I had to choose a graphics card first, and I went with a brand I’m familiar with. We use Gigabyte’s Windforce-cooled GeForce GTX 970 in our reference systems, so I was pretty cranked to upgrade to the latest/greatest in the Windforce family. Once I got my graphics card settled, I moved on and picked a CPU and motherboard. After all that, I just picked the pieces that made sense and fit into the budget.
In keeping within the spirit of the GPU vs. CPU challenge, I focused a huge chunk of my budget on the platform's 3D potential. In fact, just over one-third of this first machine’s System Builder Marathon funding went to graphics, followed by the CPU.
After choosing the two key components for Big Build (and spending more than half of my budget), the next thing I needed to focus on was memory and storage. I wanted to keep things interesting, so I bounced around a bit to fit in at least one well-known goodie (the SSD) along with one brand I’m not too familiar with (RAM) and a couple of obvious accessories (hard drive and DVD).
For storage, I’m going with a nice tiered subsystem sensitive to my budget limitations, including an SSD for the operating system and a HDD for user files. In this case, I’m throwing in a 240GB SanDisk Extreme Pro SSD for the system drive, and a 1TB Seagate Barracuda for the music, movies and documents.
Once I figured out all of the big components, I needed to build out the backbone and chassis to put it all in. The motherboard was priority one in this group, followed by a power supply and a nice big case.
CPU & RAM Cooling
There were two key components I wanted to emphasize with my cooling: the CPU and memory modules. The DIYPC case I picked has five 120mm fans. Originally, the top fans were blowing air out of the case, but since they were positioned right above the CPU cooler and the RAM’s heat sinks, it made sense to flip them both over and have their air blown downwards. If I had a decent piece of plastic board, the right tools and the time, I would have loved to build an air channel that would separate the airflow between the CPU cooler’s fan and the rear fan, but I didn’t. In this configuration, the rear fan will not only push out hot air from the PC, but it’s also going to get some of the fresh air meant for the CPU.
One other possibility to maximize CPU and RAM cooling would be to position the CPU cooler fan on the other side of its heat sink, letting the front-top fan blow fresh air over the CPU cooler and RAM at the same time. If that were possible, I could have kept the rear-top fan in its original position where it'd help exhaust hot air out, along with the rear fan. Ultimately, this wasn’t possible since the Zalman fan was too thick and wouldn’t fit between the CPU cooler’s fins and the RAM's heat sinks.
Storage & Graphics Card
Next, I wanted to address where to put the storage drives. I expected that the hard drive and SSD would generate some heat, so I wanted to make sure I did something to lessen their thermal footprint.
In the “Big Build’s Overview” picture, you’ll notice an empty space between the motherboard and the the front panel’s fans. That space was originally occupied by two hard drive cage. The middle one is removable and slips right out. The bottom hard drive cage, unfortunately, is riveted into the case’s frame and cannot be removed without some heavy-duty modification. The good thing, though, is that the hard drive trays inside the bottom cage are removable. As a result, in order to maximize the amount of fresh air going into the front of the case, I emptied out that entire section and put the hard drive and SSD inside the hard drive cage at the top of the case where they’ll both sit right below the DVD writer.
There were two advantages to moving and isolating the storage drives into the top hard drive cage. The first was that by not using the lower cages, I can get cool air in from the front of the cage and direct it towards the GPU. A second benefit is that I have one less set of power cables to run from the power supply to the lower cage area. Now I only have a single SATA power cable that goes from the power supply to the top hard drive cage, and by having just that one cable, airflow towards the back of Big Build is going to be a little less obstructed.
Overall, the design of the DIYPC case isn’t too bad. There’s plenty of space to work with. However, there were a couple of issues I had to work around.
First, the top hard drive cage holding the Samsung DVD writer, Seagate HDD and SanDisk SSD was too close to the motherboard’s main power harness and memory slots. Not only did I have to unplug the power harness and remove the memory modules to install the three storage devices, but I still had to use some of my old desktop support yoga skills to get the SSD into the cage, as some of the soldered parts on the motherboard stuck out and got in the way.
The second thing I didn’t care much about the DIYPC case were those lower hard drive cages that I ended up removing. Running the cables to the drives when I initially put the machine together was awkward. Because of the way the mounting holes were positioned, the SSD would only go in rear-first, while the hard drive went in front-first. At that early point in the build, the way the cables were laid out didn’t work for me, so moving the drives up into the top fixed the minor aesthetic problem I was having and helped with airflow.
Graphics Card Drivers
When I started working on Big Build, I initially used a single Acer S277HK 4K display. Armed with the CDs that came with the ASRock motherboard and Gigabyte GeForce GTX 980, I installed the necessary drivers, and then updated them over the network. Everything seemed fine until the day I moved the Big Build to its open test area in the office. As I plugged in three new 4K displays into the Gigabyte card’s DisplayPort outputs, just one of the interfaces worked. The second and third ones weren’t putting out any signal at all.
After spending some time swapping cables around and verifying that the monitors weren’t misconfigured or defective, I uninstalled the Nvidia drivers and software, only to find that the native drivers were working fine as all three displays came up and successfully displayed the Windows 8 desktop. Eventually, I was able to re-install the latest Nvidia drivers (version 353.05) again and got all three 4K monitors running.
After digging around in the OC Tweaker UI and getting mixed results with the ASRock Z97 Extreme6’s “Optimized CPU OC” settings, I went ahead and manually edited the overclock settings. I initially set the multiplier option to 42, and then started a conservative climb to 44. As I went along, I tested different multiplier/voltage settings by using Prime95 to generate a workload and RealTemp to see how hot the CPU was getting. Once I was able to settle on a combination that kept the CPU peaking in the low 90-degree C range, I made sure that I was happy with the 3.5GHz Core i5-4690K running at 4.4GHz and 1.24V by tacking on an additional workload – in this case, our Tom’s Hardware 3ds benchmark.
For the memory, I just ended up using the OC Tweaker’s XMP profile to set my DRAM. The time to get Big Build ready for the quarter's SBM was getting shorter, so I had to put off the memory overclocking and re-prioritize. At that point, I had bigger fish to fry…
Next, I needed to overclock the Gigabyte GeForce GTX 980 graphics card using Gigabyte’s OC Guru UI. This took a little tinkering. I just so happened to start the tests with high memory and GPU clocks, and from there I slowly dropped clock rate until I found something stable in 3DMark’s Fire Strike benchmark. It took a few tries, but I ended up with the memory clock set to 8000 MT/s and the GPU base clock at 1365MHz. Lastly, I set the GPU fans to manual and pushed them to their fastest speeds. After one last run of Fire Strike, I started the automated benchmarks and headed home for a nap.
In its overclocked state, Big Build ran the gaming benchmarks pretty smoothly. There were a couple of points when I was testing with Grid 2 that Windows crashed, so I dropped the memory to 7900 MT/s and the GPU 1350MHz. After that, the benchmarks went fine and I was ready to move on to the thermal and power tests. Similar to what I did when I was setting up the host processor the night before, I used Prime95 to stress the machine. Suddenly, I started getting one blue screen after another. It didn’t seem like a complete hardware failure since the machine didn’t immediately reboot and go into POST; it would just blue-screen and tell me that that it would eventually reboot itself.
I suspected a combination of factors. Going back to the default CPU clock rate ultimately fixed the problem, and raising and lowering the GPU’s overclock settings didn’t seem to keep the machine running at the higher CPU clock setting. I was able to stabilize Big Build’s power and thermal testing by running the core speed at 4.2GHz and dropping the voltage down to 1.235V. Afterwards, I got my power and thermal numbers and put Big Build back into its prior 4.4GHz/1.24V configuration and reran the graphics benchmarks to confirm stability.
I also want to add some extra notes about getting past the problem I had with the power/thermal testing. I was able to stretch the amount of time between Windows crashing on me by disabling Nvidia’s Surround feature. I also reinstalled the Nvidia drivers after one particular crash and was able to run stably for a little while longer. Lastly, and interestingly enough, I noticed the CPU wasn’t overheating when the OS blue-screened; RealTemp usually reported that the processor was in the mid to high 80 degree-range whenever it crashed.
How We Test
I covered some of the points on how we tested Big Build in the prior Overclocking section. Aside from those notes, we purchased the parts, assembled the machines and ran the tests ourselves. I set up Big Build and the other machine in our Los Angeles test lab where I had access to the tools and space needed for this quarter’s System Builders Marathon. One key factor that I find important for working in the lab is having a climate-controlled environment to run the machines in, especially when testing the overclocking. Since most of my work was done at night, I had the A/C to myself and set the temperature to about 73 degrees F. With both machines fired up around mid-afternoon, the room's temperature rose to about 78 degrees F.
Using the standard Tom’s Hardware image and software, I benchmarked the non-overclocked PC configuration first, then set the overclocking on that machine and ran the tests one more time.
Aside from the variety of tools and hardware needed, we also introduced 4K monitors to the test scenario. We used a couple of 27” Viewsonic VP2780-4K displays sandwiching a 28” Acer XB280HK. Since Big Build’s Gigabyte GTX 980 graphics card had three DisplayPort outputs, we used them all for testing.
The images we use center on Windows 8 and are pre-loaded with all of the System Builder Marathon software. We simply write a fresh image to the machine's SSD, update the firmware, install the drivers and configure third-party tuning apps.
|Q2 $1600 Gaming PC||Q1 $1750 Performance PC||Q4 2014 $1600 PC|
|Intel Core i5-4690K: 3.5GHz, Four Physical Cores O/C to 4.2-4.4GHz, 1.24V||Intel Core i7-4790K: 4.0-4.4GHz, Four Physical Cores O/C to 4.6-4.8GHz, +20mV||Intel Core i7-4790K: 4.0-4.4GHz, Four Physical Cores O/C to 4.6GHz, 1.26V|
|Gigabyte GeForce GTX 980: 1178MHz GPU, GDDR5-7010 O/C to 1335MHz, GDDR5-8000||2x PNY GeForce GTX 970: 1178MHz GPU, GDDR5-7012 O/C to 1328MHz, GDDR5-7312||PNY GeForce GTX 980: 1216MHz GPU, GDDR5-7012 O/C to 1456MHz, GDDR5-7972|
|16GB Team Extreme DDR3-2400 CAS 10-12-12-31, Applied XMP Profile||16GB G.Skill DDR3-1866 CAS 10-11-10-28, O/C to DDR3-2133 CL 11-12-11-24, 1.6V||8GB G.Skill DDR3-2133 CAS 9-11-10-28, O/C to DDR3-2400 CL 10-12-12-28, 1.6V|
|ASRock Z97 Extreme6: LGA 1150, Intel Z97 Express Stock 100MHz BCLK||Gigabyte Z97X-Gaming 5:|
LGA 1150, Intel Z97 Express
Stock 100MHz BCLK
|Biostar Hi-Fi Z97WE:|
LGA 1150, Intel Z97 Express
Stock 100MHz BCLK
|Case||DIYPC Adventurer-9601G ||Corsair Graphite 230T||Thermaltake Chaser A31|
|CPU Cooler||Zalman CNPS10X Optima||Corsair H100i Closed-Loop||Phanteks PH-TC14PE 140mm|
|Hard Drive||SanDisk Extreme Pro 240GB SATA 6Gb/s SSD||Crucial MX100 256GB SATA 6Gb/s SSD||Plextor M6S PX-256M6S 256GB SATA 6Gb/s SSD|
|Power||Corsair CSM Series CS750M: 750W, 80 PLUS Gold ||Rosewill Capstone-750: 750W, 80 PLUS Gold||Rosewill Capstone-750-M: 750W, 80 PLUS Gold|
|OS||Microsoft Windows 8 Pro x64||Microsoft Windows 8 Pro x64||Microsoft Windows 8 Pro x64|
|Graphics||Nvidia GeForce 353.06||Nvidia GeForce 347.25||Nvidia GeForce 344.75|
|Chipset||Intel INF 188.8.131.527||Intel INF 184.108.40.2066||Intel INF 220.127.116.116|
For starters, the overclocked numbers generated in 3DMark show the machine did fairly well out the gate, especially if you notice the graphics score. Compared to Thomas’ other machines, Big Build couldn’t catch up to Q1 2015’s dual GeForce GTX 970-based rig, but then again, two GM104 GPUs might have something to do with that. On the other hand, Q4 2014’s $1600 Performance PC was as close as I could get to an even comparison. Putting both $1600 rigs up against each other, Big Build may have scored lower, but I was about to squeeze out a better score spread between the baseline and overclock configurations.
The Sandra results show what happens when you compromise and go with Core i5 when the other kids go i7. Still, I’m really proud of the smaller gap in Big Build’s scores.
We ran into an issue gaming at 4800x900 on the 4K monitors. Basically, the resolution wasn’t available, even though 1600x900 was. Because of that, we had to exclude it from our comparison.
Overall, Big Build excelled at 1600x900 and 1920x1080 across the four games we benchmark. This wasn't the case at 5760x1080, though. Grid 2 is the exception; it was fairly smooth at the High Quality setting.
Looking at the charts from a historic perspective, the overclocked Big Build kept up with Thomas’ past creations, including his dual-card configuration from Q1 2015. The only trouncing I really received in the game testing came from Battlefield 4's Ultra quality preset, where most of Thomas’ builds outright beat me (though in one case, just by 3.2 FPS).
Media, Productivity And Compression
We have a few more examples of what happens when you put a Core i5 up against an i7. For the scoring, lower numbers are better. And since I’m not seeing any numbers higher than mine, I definitely know my place in this section. In the end though, larger gaps between the baseline and overclock show what a difference time and patience can make when you’re trying to bump up performance with an overclock.
Power, Heat, Efficiency And Value
With all of the equipment inside the DIYPC case, Big Build drew quite a bit of power just sitting idle. But once the GPU and CPU received a load, draw from the wall increased 5x. Thomas' experience in this series is apparent when you look at his results from Q4 2014 and Q1 2015, where power consumption went up by 10x under load compared to the idle measurements.
Even though I was looking forward to the power and thermal tests (mostly because they were the last ones I had to run), I spent quite a while trying to mitigate the system crashing whenever I started running Prime95. As explained earlier, I ultimately ended up dropping the CPU to 4.2GHZ and adjusting the GPU to lower settings as well. Other than the temperature and power tests, all of this build's other benchmarks were run with the CPU at 4.4GHz.
Also a note about the results in the Temperatures Above Ambient graph. The fans on the GPU were set to manual and configured to go to 100% during the overclock testing. Also, the Gigabyte GeForce GTX 980 I used got an unobstructed flow of air from the front of the case because of the hard drive cages I removed.
By going with a Core i5, I discovered that it's possible to build a great gaming rig for $1600, though overclocking it (at least in this case) is a must. In its baseline configuration, Big Build isn’t going to win any awards. However, it does hammer home the point that finding time and patience for tuning can yield a rewarding performance increase. I would have had to go over-budget by $100 or $150 to score a higher-end Core i7 CPU.
Looking at the Overall Performance chart, the task that I was originally assigned – building a gaming PC – seems to have been accomplished, even if my take on the system needs to be overclocked for an optimal experience. Our Tom's Hardware technical career guidance counselors deem the overclocked Big Build to be a decent graphics workstation, too. Interestingly enough, we also see the SanDisk Pro SSD thrusted into the spotlight as its storage performance misses the 100% target by just 1%. So, if you’re a gamer or a creative content producer, a machine like Big Build may just fit in with your career plans if you don’t mind a little overclocking.