- Oct 9, 2006
- 2,055
- 1
- 20,160
Hello everyone!
The goal of this review is to find out if the upgrade from an i7-970 to an i7-6850k was worth it.
For the past 6 years I’ve been running on a 6-core Intel Core i7-970 overclocked to 4GHz on a Gigabyte GA-X58A-UD5 mainboard and 12GB DDR3-1600 RAM. The X58/i7-970 combo has been truly amazing. But for the past year, the question I’ve been trying to answer is – does it make sense to upgrade from X58 to X99, especially with an overclocked 6-core? I couldn't find any meaningful comparisons in the online forums. Shame on Intel for making this such a hard one to answer. Shame on AMD for not providing any meaningful competition until, we hope, Zen finally pulls out all the stops. In 6 years, performance has not improved enough to justify frequent upgrades while the graphics card industry has doubled performance in one short year. For example, my preliminary research puts a 4GHz i7-6850k just 50% ahead of a 4GHz i7-970 in overall performance. Since I’ve been playing WoW and ESO, single-threaded performance has become more important to me and that’s one area where the i7-970 fails compared to modern CPUs. Combine that with a relatively ancient SATA2 controller (don’t even bring up the Marvel SATA3 controller…) and 12GB memory limit (24GB is unavailable at the latencies I want, mixing two 12GB kits is a bad idea), and the desire to max out the 144fps limit of my monitor, it may finally be time to upgrade.
I could go with a faster single-core-performance Quad Core, but I’m a nut for high-end hardware and I’ve been really happy with the 6-core’s ability to punch through multi-threaded tasks, so I’ve decided to bite the bullet and go with X99 and the six-core i7-6850k complete with 40 PCI-E lanes that I probably won’t use. Because there are so few reviews between the two platforms, I’ve decided to justify some of the cost by providing a review of the Core i7-970 and Core i7-6850k at 4GHz.
Here is the list of components. The 970 is water-cooled with a NZXT Kraken X61 and I will be keeping that water-cooler for the new CPU. I will also be swapping out the case from a TT Level 10 GT to a TT Core X71, and upgrading the OS/programs storage from a 2x128GB C300 RAID-0 SSD array to a single 512GB 950 Pro M.2 PCI-E 3.0 x4 SSD. Because the X71 case is custom-loop oriented, cooling performance may change a bit between cases but hopefully the new motherboard will be able to cope with potentially higher temperatures when I lose the side intake fan.
Old Hardware:
- Core i7-970 @ 4GHz (1.35v)
- Gigabyte GA-X58A-UD5 Rev.2.0
- G.Skill 12GB DDR3-1600 (CAS 8)
- Gigabyte GTX 1080 Windforce 8G OC
- Thermaltake Level 10 GT modified to fit the Kraken on top
- 2x 128GB Crucial C300 SSD in RAID-0
- Windows 10 Professional x64
New Hardware:
- Core i7-6850k @ 4GHz (1.26v)
- Gigabyte X99-Ultra Gaming G1
- G.Skill 32GB DDR4-3200 (CAS 14)
- Gigabyte GTX 1080 Windforce 8G OC
- Thermaltake Core X71
- Samsung 950 Pro M.2 SSD
- Windows 10 Professional x64
Testing Software and a Kill-a-Watt
- Cinebench R15
- 3DMark Firestrike, Firestrike Extreme, Firestrike Ultra
- SuperPi 1.90 WP
- Passmark Performance Test 8
- WoW – (StormWind City)
- ESO – Elden Root
- SiSoftware Sandra 2016
- NZXT CAM
- CoreTemp 1.3
- Prime95
- Kill-a-Watt power meter
Since I’m more concerned with single-core performance for this upgrade, I chose to focus the testing and benchmarks more on the per-core performance but do have a few multi-threaded results in there too, because why not?
A word on overclocking: The i7-970 is overvolted to stabilize the 4GHz core clock. The i7-6850k, when c-states and Speedstep are disabled, automatically set the multiplier to 40x bringing it to 4GHz on stock voltage. I should be seeing some power and heat reductions because of this.
Cinebench R15
Core i7-970 (MT): 901cb
Core i7-970 (ST): 119cb
OpenGL: 103.24fps
Core i7-6850k (MT): 1229cb (+36%)
Core i7-6850k (ST): 169cb (+42%)
OpenGL: 179.65fps (+74%)
A bit of a surprise here – I didn’t expect that much of a difference in OpenGL GPU performance but there it is. Also interesting is how the 42% increase in single-core performance doesn’t exactly scale in multi-threaded performance.
3DMark Firestrike/Extreme/Ultra
Core i7-970
Combined: 16925/9610/5189
Physics: 13524/13450/13484
Graphics: 22166/10510/5195
Core i7-6850k
Combined: 18852/10132/5380 (+11.38%/+5.4%/+3.68%)
Physics: 17915/17877/17899 (+~33%)
Graphics: 22684/10802/5312 (+2.34%/+2.78%/+2.25%)
The Physics improvements are clear but not the 50% improvement that I was expecting. What’s this? A slight increase in graphics score? Read on to find out why.
SuperPI 1.90 WP (single-thread)
Core i7-970: 9m 27.484s
Core i7-6850k: 7m 59.562s (+22% if my math is correct)
Again, a bit of a surprise here on the single-core performance scale. Was really hoping for more!
Passmark 8
Core i7-970
CPU: 10645
CPU Single Threaded: 1695
CPU Physics: 720
2D Graphics: 687
3D Graphics: 10318
Memory: 2136
Core i7-6850k
CPU: 15695 (+47.4%)
CPU Single Threaded: 2329 (+37.4%)
CPU Physics: 1096 (+52.2%)
2D Graphics: 952 (+38.6%)
3D Graphics: 10366 (+.4%)
Memory: 3177 (+48.7%)
Increases across the board but only the CPU Physics score hit my 50% goal. I was happy to see a nearly 49% memory performance increase considering the 3200mhz CAS14 would be equal to DDR3-1600 at CAS7, which would only be slightly faster than my old RAM minus the extra channel. So adding a channel plus lower latency does have its benefits.
SiSoftware Sandra 2016
Processor Arithmetic
Core i7-970: 133.1 GOPS
Core i7-6850k: 176.35 GOPS (+32.5%)
Processor Multi-Media
Core i7-970: 205.2 MPix/s
Core i7-6850k: 526.36 MPix/s (+156.5%)
Memory Bandwidth
Core i7-970: 23GB/s
Core i7-6850k: 50.47GB/s (+119.4%)
Cache Bandwidth
Core i7-970: 99.13GB/s
Core i7-6850k: 233GB/s (+135%)
Video Shader Compute
Core i7-970: 1.43GPix/s
Core i7-6850k: 1.45GPix/s
Video Memory Bandwidth
Core i7-970: 35.09GB/s
Core i7-6850k: 50.34GB/s (+44%)
Video Memory Bandwidth – Interface Transfer Bandwidth
Core i7-970: 5.65GB/s
Core i7-6850k: 11GB/s (+95%)
Wow! Synthetic benchmarks really paint a rosy picture for the 6850k but so far we haven’t seen overall performance even come close to these figures. I added the Video Memory Bandwidth and the Interface Transfer Bandwidth subtest to try and answer why we saw slight improvements in 3D graphics performance in the previous tests. It turns out that the GTX 1080 may actually saturate the PCI-E 2.0 bus by quite a bit, and giving it some room to breathe gives us a little more performance.
Gaming
WoW – StormWind City (PvP sewers)
Core i7-970: 33-50fps
Core i7-6850k: 80-100fps (+100-242%)
ESO Elden Root
Core i7-970: 23fps
Core i7-6850k: 63fps (+174%)
ESO Elden Root has always been a consistent 23fps for me so those are results that I am very happy to see. WoW was more difficult to get right but I’m confident that those numbers are legit. Well beyond the 50% I was shooting for!
Temperatures
GPU Load 3DMark Firestrike Extreme Stress Test – 10 minutes
Core i7-970
GPU Temp: 69C
GPU Fan Speed: 2700 RPM
CPU Load: 23%
CPU Temp: 44C
Kraken X61 Fan Speed: 30%
Core i7-6850k
GPU Temp: 65C
GPU Fan Speed: 2550 RPM
CPU Load: 16%
CPU Temp: 35C
Kraken X61 Fan Speed: 26%
CPU Load Prime95 Small FFT – 10 minutes
Core i7-970
CPU Temp: 64C
Kraken X61 Fan Speed: 34%
Core i7-6850k
CPU Temp: 69C
Kraken X61 Fan Speed: 35%
Combined – Prime95 and 3DMark Firestrike Extreme – 10 minutes
Core i7-970
CPU Temp: 66C
Kraken X61 Fan Speed: 39%
GPU Temp: 69C
GPU Fan Speed: 2700 RPM
Core i7-6850k
CPU Temp: 71C
Kraken X61 Fan Speed: 34%
GPU Temp: 66C
GPU Fan Speed: 2572 RPM
A little background story, the Kraken X61 was pre-set to Silent mode and adjusted fan speed based on liquid temperature. Lower voltage gives the 6850k an advantage at idle, but under load it is unable stay as cool as the overvolted 970. What you’ll notice here is that the fan speed for the 6850k is lower than the 970 under load, while core temps remain higher. In fact, one of my cores hit 81C at one point. I ended up reapplying GELID GC-Extreme thermal paste and using a bit less than before but the results were no better. The fan speed is the result of a lower liquid temperature, so we have an issue with heat transfer from the core to the water block. My guess is that the 14nm package simply has a harder time dissipating the heat over a larger area vs the 970. Setting the Kraken to Performance mode reduces core temps to low 60C. This may be the setting I use if I overclock. Another good thing is that the GPU temp actually dropped after moving to the Core X71 case.
Power – Kill-a-Watt (whole strip measurement)
Core i7-970
Idle: 244 Watts
CPU Load: 384 Watts
GPU Load: 466 Watts
Combined Load: 580 Watts
Core i7-6850k
Idle: 210 Watts
CPU Load: 330 Watts
GPU Load: 455 Watts
Combined Load: 550 Watts
The whole power strip, including my two monitors, 5.1 speakers, and tower, are plugged in to the Kill-a-Watt, so what you’re seeing is the power draw of my entire gaming platform. My original power tests actually put Idle and CPU Load results higher than the i7-970, but I decided to re-test and got the posted results to my delight.
Conclusion: Great synthetic results but disappointing "real-world" benchmark improvements abound. However, around 150% improvement in MINIMUM FPS in CPU-bound games, a decent reduction in power usage, and nearly 5x improvement in storage performance (2,500mb/s sequential vs 550mb/s), I am happy with the X99 and 6850k. The 970 is no slouch though – in multi-threaded games it will still hold its own. I was also happy to see that there was an improvement moving to PCI-E 3.0 in interface and memory bandwidth confirming that PCI-E 2.0 may no longer be suitable for high-end graphics cards. While the overall performance improvements were small in my tests save for the synthetics, the gap will only grow from here on out. Was the $1,300 upgrade worth it? For me, yes.
The goal of this review is to find out if the upgrade from an i7-970 to an i7-6850k was worth it.
For the past 6 years I’ve been running on a 6-core Intel Core i7-970 overclocked to 4GHz on a Gigabyte GA-X58A-UD5 mainboard and 12GB DDR3-1600 RAM. The X58/i7-970 combo has been truly amazing. But for the past year, the question I’ve been trying to answer is – does it make sense to upgrade from X58 to X99, especially with an overclocked 6-core? I couldn't find any meaningful comparisons in the online forums. Shame on Intel for making this such a hard one to answer. Shame on AMD for not providing any meaningful competition until, we hope, Zen finally pulls out all the stops. In 6 years, performance has not improved enough to justify frequent upgrades while the graphics card industry has doubled performance in one short year. For example, my preliminary research puts a 4GHz i7-6850k just 50% ahead of a 4GHz i7-970 in overall performance. Since I’ve been playing WoW and ESO, single-threaded performance has become more important to me and that’s one area where the i7-970 fails compared to modern CPUs. Combine that with a relatively ancient SATA2 controller (don’t even bring up the Marvel SATA3 controller…) and 12GB memory limit (24GB is unavailable at the latencies I want, mixing two 12GB kits is a bad idea), and the desire to max out the 144fps limit of my monitor, it may finally be time to upgrade.
I could go with a faster single-core-performance Quad Core, but I’m a nut for high-end hardware and I’ve been really happy with the 6-core’s ability to punch through multi-threaded tasks, so I’ve decided to bite the bullet and go with X99 and the six-core i7-6850k complete with 40 PCI-E lanes that I probably won’t use. Because there are so few reviews between the two platforms, I’ve decided to justify some of the cost by providing a review of the Core i7-970 and Core i7-6850k at 4GHz.
Here is the list of components. The 970 is water-cooled with a NZXT Kraken X61 and I will be keeping that water-cooler for the new CPU. I will also be swapping out the case from a TT Level 10 GT to a TT Core X71, and upgrading the OS/programs storage from a 2x128GB C300 RAID-0 SSD array to a single 512GB 950 Pro M.2 PCI-E 3.0 x4 SSD. Because the X71 case is custom-loop oriented, cooling performance may change a bit between cases but hopefully the new motherboard will be able to cope with potentially higher temperatures when I lose the side intake fan.
Old Hardware:
- Core i7-970 @ 4GHz (1.35v)
- Gigabyte GA-X58A-UD5 Rev.2.0
- G.Skill 12GB DDR3-1600 (CAS 8)
- Gigabyte GTX 1080 Windforce 8G OC
- Thermaltake Level 10 GT modified to fit the Kraken on top
- 2x 128GB Crucial C300 SSD in RAID-0
- Windows 10 Professional x64
New Hardware:
- Core i7-6850k @ 4GHz (1.26v)
- Gigabyte X99-Ultra Gaming G1
- G.Skill 32GB DDR4-3200 (CAS 14)
- Gigabyte GTX 1080 Windforce 8G OC
- Thermaltake Core X71
- Samsung 950 Pro M.2 SSD
- Windows 10 Professional x64
Testing Software and a Kill-a-Watt
- Cinebench R15
- 3DMark Firestrike, Firestrike Extreme, Firestrike Ultra
- SuperPi 1.90 WP
- Passmark Performance Test 8
- WoW – (StormWind City)
- ESO – Elden Root
- SiSoftware Sandra 2016
- NZXT CAM
- CoreTemp 1.3
- Prime95
- Kill-a-Watt power meter
Since I’m more concerned with single-core performance for this upgrade, I chose to focus the testing and benchmarks more on the per-core performance but do have a few multi-threaded results in there too, because why not?
A word on overclocking: The i7-970 is overvolted to stabilize the 4GHz core clock. The i7-6850k, when c-states and Speedstep are disabled, automatically set the multiplier to 40x bringing it to 4GHz on stock voltage. I should be seeing some power and heat reductions because of this.
Cinebench R15
Core i7-970 (MT): 901cb
Core i7-970 (ST): 119cb
OpenGL: 103.24fps
Core i7-6850k (MT): 1229cb (+36%)
Core i7-6850k (ST): 169cb (+42%)
OpenGL: 179.65fps (+74%)
A bit of a surprise here – I didn’t expect that much of a difference in OpenGL GPU performance but there it is. Also interesting is how the 42% increase in single-core performance doesn’t exactly scale in multi-threaded performance.
3DMark Firestrike/Extreme/Ultra
Core i7-970
Combined: 16925/9610/5189
Physics: 13524/13450/13484
Graphics: 22166/10510/5195
Core i7-6850k
Combined: 18852/10132/5380 (+11.38%/+5.4%/+3.68%)
Physics: 17915/17877/17899 (+~33%)
Graphics: 22684/10802/5312 (+2.34%/+2.78%/+2.25%)
The Physics improvements are clear but not the 50% improvement that I was expecting. What’s this? A slight increase in graphics score? Read on to find out why.
SuperPI 1.90 WP (single-thread)
Core i7-970: 9m 27.484s
Core i7-6850k: 7m 59.562s (+22% if my math is correct)
Again, a bit of a surprise here on the single-core performance scale. Was really hoping for more!
Passmark 8
Core i7-970
CPU: 10645
CPU Single Threaded: 1695
CPU Physics: 720
2D Graphics: 687
3D Graphics: 10318
Memory: 2136
Core i7-6850k
CPU: 15695 (+47.4%)
CPU Single Threaded: 2329 (+37.4%)
CPU Physics: 1096 (+52.2%)
2D Graphics: 952 (+38.6%)
3D Graphics: 10366 (+.4%)
Memory: 3177 (+48.7%)
Increases across the board but only the CPU Physics score hit my 50% goal. I was happy to see a nearly 49% memory performance increase considering the 3200mhz CAS14 would be equal to DDR3-1600 at CAS7, which would only be slightly faster than my old RAM minus the extra channel. So adding a channel plus lower latency does have its benefits.
SiSoftware Sandra 2016
Processor Arithmetic
Core i7-970: 133.1 GOPS
Core i7-6850k: 176.35 GOPS (+32.5%)
Processor Multi-Media
Core i7-970: 205.2 MPix/s
Core i7-6850k: 526.36 MPix/s (+156.5%)
Memory Bandwidth
Core i7-970: 23GB/s
Core i7-6850k: 50.47GB/s (+119.4%)
Cache Bandwidth
Core i7-970: 99.13GB/s
Core i7-6850k: 233GB/s (+135%)
Video Shader Compute
Core i7-970: 1.43GPix/s
Core i7-6850k: 1.45GPix/s
Video Memory Bandwidth
Core i7-970: 35.09GB/s
Core i7-6850k: 50.34GB/s (+44%)
Video Memory Bandwidth – Interface Transfer Bandwidth
Core i7-970: 5.65GB/s
Core i7-6850k: 11GB/s (+95%)
Wow! Synthetic benchmarks really paint a rosy picture for the 6850k but so far we haven’t seen overall performance even come close to these figures. I added the Video Memory Bandwidth and the Interface Transfer Bandwidth subtest to try and answer why we saw slight improvements in 3D graphics performance in the previous tests. It turns out that the GTX 1080 may actually saturate the PCI-E 2.0 bus by quite a bit, and giving it some room to breathe gives us a little more performance.
Gaming
WoW – StormWind City (PvP sewers)
Core i7-970: 33-50fps
Core i7-6850k: 80-100fps (+100-242%)
ESO Elden Root
Core i7-970: 23fps
Core i7-6850k: 63fps (+174%)
ESO Elden Root has always been a consistent 23fps for me so those are results that I am very happy to see. WoW was more difficult to get right but I’m confident that those numbers are legit. Well beyond the 50% I was shooting for!
Temperatures
GPU Load 3DMark Firestrike Extreme Stress Test – 10 minutes
Core i7-970
GPU Temp: 69C
GPU Fan Speed: 2700 RPM
CPU Load: 23%
CPU Temp: 44C
Kraken X61 Fan Speed: 30%
Core i7-6850k
GPU Temp: 65C
GPU Fan Speed: 2550 RPM
CPU Load: 16%
CPU Temp: 35C
Kraken X61 Fan Speed: 26%
CPU Load Prime95 Small FFT – 10 minutes
Core i7-970
CPU Temp: 64C
Kraken X61 Fan Speed: 34%
Core i7-6850k
CPU Temp: 69C
Kraken X61 Fan Speed: 35%
Combined – Prime95 and 3DMark Firestrike Extreme – 10 minutes
Core i7-970
CPU Temp: 66C
Kraken X61 Fan Speed: 39%
GPU Temp: 69C
GPU Fan Speed: 2700 RPM
Core i7-6850k
CPU Temp: 71C
Kraken X61 Fan Speed: 34%
GPU Temp: 66C
GPU Fan Speed: 2572 RPM
A little background story, the Kraken X61 was pre-set to Silent mode and adjusted fan speed based on liquid temperature. Lower voltage gives the 6850k an advantage at idle, but under load it is unable stay as cool as the overvolted 970. What you’ll notice here is that the fan speed for the 6850k is lower than the 970 under load, while core temps remain higher. In fact, one of my cores hit 81C at one point. I ended up reapplying GELID GC-Extreme thermal paste and using a bit less than before but the results were no better. The fan speed is the result of a lower liquid temperature, so we have an issue with heat transfer from the core to the water block. My guess is that the 14nm package simply has a harder time dissipating the heat over a larger area vs the 970. Setting the Kraken to Performance mode reduces core temps to low 60C. This may be the setting I use if I overclock. Another good thing is that the GPU temp actually dropped after moving to the Core X71 case.
Power – Kill-a-Watt (whole strip measurement)
Core i7-970
Idle: 244 Watts
CPU Load: 384 Watts
GPU Load: 466 Watts
Combined Load: 580 Watts
Core i7-6850k
Idle: 210 Watts
CPU Load: 330 Watts
GPU Load: 455 Watts
Combined Load: 550 Watts
The whole power strip, including my two monitors, 5.1 speakers, and tower, are plugged in to the Kill-a-Watt, so what you’re seeing is the power draw of my entire gaming platform. My original power tests actually put Idle and CPU Load results higher than the i7-970, but I decided to re-test and got the posted results to my delight.
Conclusion: Great synthetic results but disappointing "real-world" benchmark improvements abound. However, around 150% improvement in MINIMUM FPS in CPU-bound games, a decent reduction in power usage, and nearly 5x improvement in storage performance (2,500mb/s sequential vs 550mb/s), I am happy with the X99 and 6850k. The 970 is no slouch though – in multi-threaded games it will still hold its own. I was also happy to see that there was an improvement moving to PCI-E 3.0 in interface and memory bandwidth confirming that PCI-E 2.0 may no longer be suitable for high-end graphics cards. While the overall performance improvements were small in my tests save for the synthetics, the gap will only grow from here on out. Was the $1,300 upgrade worth it? For me, yes.
hope your new system is rockin'
Facebook
Twitter
Instagram