Just like before, I want to take a moment to focus on real world multi-tasking. Instead of pitting a four core against a six core, this time it's a fair fight. Straight up four core Sandy Bridge vs. Ivy Bridge on identical systems at stock clocks.
What does multi-tasking mean in this context? I don't mean the architectural definition so much as the "work flow" definition.
For more information on my philosphy of multi-core multi-tasking, see "Gulftown vs. Sandy Bridge".
Let's start with something dramatic, a video comparing Sandy Bridge 3820 vs. Ivy Bridge 3770K in a VMware four OS shootout:
Windows 7 x64
Windows 8 Consumer Preview x64
Ubuntu 11.10 x64
MacOS 10.7.4 x64
Various settings and results are presented.
Now that we've seen that, let's take a closer look at the platforms we're using:
The Ivy Bridge Platform
Today you're going to see a LOT of Ivy Bridge commentary. You'll read the explanation of its feature set about a dozen times. Why don't I just cut to the chase? When this site is bigger, we'll go more into these architectural details.
For this round, I'm not using the full suite of benches I normally test GPUs with. There appears to be a possible issue with the HD 4000 implementation I have here, and I'm going to test with another board next week to provide a little more due diligence before making my final commentary on it.
Settings - Check ALL boxes, "Secondary Storage Suite", set to C drive.
Passmark Performance Test 7
Settings - "Run All Tests", and "Multitasking", both set to 600 seconds.
Settings - Performance, and Extreme
Prime 95 Temperature run
Full default with RealTemp GT measuring
The Test Platforms:
This is a pretty straight fair fight, consisting of two nearly identical test systems:
|Sandy Bridge 3820||Ivy Bridge 3770K|
|Motherboard||MSI X79MA G45||MSI Z77MA G45|
|CPU||Intel i7 "Sandy Bridge" 3820||Intel i7 "Ivy Bridge" 3770K|
|GPU||AMD HD Radeon 6870||AMD HD Radeon 6870|
|Memory||G.Skill Ripjaws Z F3-12800CL9Q-16GBZL 16GB||G.Skill Ripjaws Z F3-12800CL9Q-16GBZL 16GB|
|Cooler||Corsair H100||Corsair H100|
|Storage||Twin OCZ Vertex 3 MAX IOPS 120GB in RAID||Twin OCZ Vertex 3 MAX IOPS 120GB in RAID|
|PSU||OCZ ZT750||OCZ ZT750|
|OS||Windows 7 x64||Windows 7 x64|
The interesting anomaly here is that the Computation score for Ivy is off the charts. While other benchmarks disagree with this assessment, I think that there's something to this, and as you saw in my VMware video, Ivy does indeed trounce Sandy at VMware.
Passmark Performance Test 7
I see this sometimes. Some CPUs do not execute the "Drystone" test, thus skewing the results slightly. Everything else seems pretty close though, which backs up our other results. Remember that the above tests were all executed at once over a ten minute period.
Unlike PCmark, Geekbench indicates that both CPUs are nearly identical in performance. I feel like these benchmarks need a good version upgrade to catch up to what we need here. let's revisit this a little later . . .
Prime95 Temperature Testing
Prime95, in case you're not familiar, is a torture test. It forces the CPU to calculate at 100% on all cores until it dies or you stop it. If you have overclocked too aggressively, Prime95 will blue screen almost immediately. That did not happen here as these CPUs are stock clocks.
Interestingly, the Ivy Bridge CPU runs an averge of 5-10 degrees C hotter than Sandy Bridge. If you have super aggresssive cooling, such as the Corsair H100, this isn't really an issue, but it's something to note if this is important to you.
I find this part VERY interesting. Normally we don't use ATTO to test CPUs, but I had a feeling we'd get an interesting result here. Note that at the larger file sizes, our Vertex 3 RAID is seeing a fairly noticable jump in performance. We're finally able to crack into the 1GBs range. I can boot this system from BIOS screen to Desktop in 9 seconds flat. I might upload a video of that this week as well.
3Dmark 11 - Performance Setting ( With Radeon 6870 )
For gaming, it looks like a wash. There is literally no discernible difference between the two CPUs when using discrete graphics.
3Dmark 11 - Extreme Setting ( With Radeon 6870 )
Intel HD 4000 Graphics
But what about the integrated GPU? This is the part we've all been dying to know, and the results are very intresting, but at this point, slightly inconclusive. Let's start with the synthetic benches.
3Dmark 11 - Performance Setting ( With Intel HD 4000 Graphics )
Yikes. Talk about a plummet. The chart above are both Ivy Bridge, but this time the difference is comparing the Intel HD 4000 with the Radeon 6870. The situation looks no better when we go to Extreme:
3Dmark 11 - Exteme Setting ( With Intel HD 4000 Graphics )
Based on what Intel tells me, these results are right on the money, but when I run some real world games, the results are underwhelming. Is this an issue that we can correct with newer drivers or a different motherboard? Tune in next week to find out. For now, let's look at some real games:
So what can we take away from this? Before I get to that, have a look at this video, which will show you several games in a real life setting. The games were recorded using an external camera, so as not to put any additional tax on the system.
Can you play Skyrim on the HD 4000? Yes, but not well. This is sill a huge leap forward for integrated graphics, but if you're a core PC gamer, you're going to have a discrete GPU.
Very slight increase in heat over the Sandy Bridge of the same clocks.
The HD 4000 Graphics gave some anomalous results on the MSI board. I'm expecting another Z77 board next week, so look for that soon as well.
So what did we learn today?
If you're gunning for the best bang for the buck, AND you are a core gamer, look at a cheaper CPU and a discrete GPU. If you're a professional video producer, VMware user, or just want a fast computer with light gaming ability the Ivy Bridge released today will be perfect.
Personally, I can't wait for the six core Ivy Bridges! That should be a fun comparison.