I recently found myself without a gaming PC, but I had a GTX980 and an old LGA775 machine which had a PCIe slot. I figured it was time for some experimentation.
I pulled out the hard drive cage along with the old 250GB platter drive that was in there, which made room for the Gigabyte G1 Gaming card and my SSD.
I then turned to the power supply, which turned out to be a 300W unit providing a grand total of 180W shared across two 12v rails. This simply wouldn't do. I turned to my Corsair TX750M, but its EPS12v power plug was strictly an 8pin arrangement and the old 775 board would only accept a lone 4pin plug. My solution was to run everything but the GPU off the old PSU and piggy-back the 750W to take up the slack. It worked brilliantly, if not looking great in the process.
Click for bigger pics.
Then it was time for some benchmarks, which I would then repeat on my new rig later.
First I installed Windows 10 x64, nVidia's 358.91 drivers, a few browsers, Crysis 3, Tomb Raider and 3DMark Firestrike. The Windows install was done from a high perforamance USB2.0 flash drive, whereas the 3DMark and game installs were all done from SSD to SSD using ISO images and Windows 10's native ISO mount functionality.
Fairly unsurprising results, aside from that Windows install time. That was a combination of two main stages of the install; the initial file decompression and install, and then the "Setting Up a Few Things" process after the first restart. I honestly have no idea why the new rig took so long; perhaps a reader can provide some insight?
It should be noted here that the performance improvements were probably affect just as much by the difference in SATA generations as much as they were by the CPU or ram.
The Windows 8.1 install that was on the 250GB hard drive was absolutely abysmal. Boot times were in the order of 3 to 5 minutes, plus another 2 or 3 minutes before the machine was usable after entering the password. After a while the machine did settle down and performance improved, but honestly not what I would call easily usable.
On the other hand, Windows 10 was brilliant on the old rig. Boot times were just as fast as I'm used to on any modern machine, and the user experience was perfect. Only on very rare occasions did I see a slight lag after clicking on something, but it wasn't enough to be annoying. I was very impressed. There didn't seem to be any detriment to running a cpu from five generations back. It was only when I hit Youtube and watched videos at 4k downscaled to 1080p that I noticed the odd skipped frame and laggy response to skipping back & forth within the video.
So I fired up Peacekeeper, a straight forward browser benchmark tool. This revealed some very interesting information.
Look at Firefox go. I'm a Chrome guy myself, but poor old Microsoft still haven't got it right apparently and their Edge browser is, at best, the same as IE.
Before you laugh off Edge completely, take a look at this:
This represents how much faster each browser ran on the i5 than it did on the Core2, presented in the same order as they are on the previous graph. Not only is Firefox the fastest on an old processor, but it gains the smallest improvement when handed a faster procesor. This points to it being the most efficient, getting the most done for the least CPU time.
The Microsoft offerings are quite the opposite; give them a slow procesor and they chug along at between hal;f and two thirds the speed of their rivals. Give them a fast CPU and they suddenly take off, gobbling up CPU time in the process.
All of that said, none of the eight user experiences were what I would call slow on the old rig. Even the power-hungry Edge browser was (almost) always smooth, with the ancient CPU and painfully slow ram not holding up the show.
But can it run Crysis?
The coloured graph is the new machine, the grey line is the old Core2 machine.
Here's a detailed explanation of frame times to those who don't know.
Crib notes: each dot on this graph represents a frame and how long it took to display. A lower line means an overall higher FPS. A flatter graph means a more stable frame rate. High values on the right represent stuttering. The higher the value the longer the stutter. The more dots that are up there, the more often the stutter happens.
When you're benchmarking a system like this, which potentially has a serious bottleneck feeding the GTX980, frame times will paint a much more accurate picture than a simple FPS measure.
As you can see, not only does the new machine average a nearly 50% higher frame rate overall, but the old machine has around 10% of its frames hanging around on the screen for over 33ms, which equates to around 30fps.
Again, the old Core2 machine puts in a commendable effort. Averaging 36fps in something as demanding as Crysis 3 on maximum detail is no mean feat.
Now for something a little easier; Tomb Raider.
First the canned benchmark:
...and then some real-world gameplay:
Wow. Not only is the old machine keeping within 4fps of the new rig but the frame time curves are just about parallel, showing that the rig isn't really tripping over its own shoelaces, it's just generally running a little slower.
I initially scoffed at both Crysis 3 and Tomb Raider's minimum CPU and ram requirements, but if this test is anything to go by then they were spot on.
Ok, for those resisting the frame time view of things, dust off your monocle and see below:
...and finally, what would any good benchmark article be without Firestrike?
Granted, the 4570T is not a beast, but those scores are much closer than I ever would have guessed.
In conclusion, my belief that your CPU and ram make very little difference to gaming has been upheld. This is a bit of an extreme case, but if you're sitting with an 1151 or 1155 rig at the moment, your next CUD move should probably still be a GPU.
In fact, if you have a strong 775 rig you're probably still fine if you're not obsessed with maintaining a solid 60fps. LGA775 motherboard, CPU and ram combos go on carb for way under R800 all the time; just think, you could build a R2500 box off Carbonite that could comfortably run Crysis 3.
Info:
GPU: Gigabyte GTX980 G1 Gaming
SSD: Samsung 840 Evo 250GB
Core2 Rig:
Intel DQ35MP-E Motherboard
Intel Q6600 Core2Quad 2.4GHz CPU
4 x 1GB Samsung DDR2 800Mhz ram
i5 Rig:
MSI H97 Gaming 3 Motherbaord
Intel Core i5 4570T 2.9GHz CPU
4 x 4GB Corsair Vengeance Pro DDR3 ram @1600Mhz
I pulled out the hard drive cage along with the old 250GB platter drive that was in there, which made room for the Gigabyte G1 Gaming card and my SSD.
I then turned to the power supply, which turned out to be a 300W unit providing a grand total of 180W shared across two 12v rails. This simply wouldn't do. I turned to my Corsair TX750M, but its EPS12v power plug was strictly an 8pin arrangement and the old 775 board would only accept a lone 4pin plug. My solution was to run everything but the GPU off the old PSU and piggy-back the 750W to take up the slack. It worked brilliantly, if not looking great in the process.
Click for bigger pics.
Then it was time for some benchmarks, which I would then repeat on my new rig later.
First I installed Windows 10 x64, nVidia's 358.91 drivers, a few browsers, Crysis 3, Tomb Raider and 3DMark Firestrike. The Windows install was done from a high perforamance USB2.0 flash drive, whereas the 3DMark and game installs were all done from SSD to SSD using ISO images and Windows 10's native ISO mount functionality.
Fairly unsurprising results, aside from that Windows install time. That was a combination of two main stages of the install; the initial file decompression and install, and then the "Setting Up a Few Things" process after the first restart. I honestly have no idea why the new rig took so long; perhaps a reader can provide some insight?
It should be noted here that the performance improvements were probably affect just as much by the difference in SATA generations as much as they were by the CPU or ram.
The Windows 8.1 install that was on the 250GB hard drive was absolutely abysmal. Boot times were in the order of 3 to 5 minutes, plus another 2 or 3 minutes before the machine was usable after entering the password. After a while the machine did settle down and performance improved, but honestly not what I would call easily usable.
On the other hand, Windows 10 was brilliant on the old rig. Boot times were just as fast as I'm used to on any modern machine, and the user experience was perfect. Only on very rare occasions did I see a slight lag after clicking on something, but it wasn't enough to be annoying. I was very impressed. There didn't seem to be any detriment to running a cpu from five generations back. It was only when I hit Youtube and watched videos at 4k downscaled to 1080p that I noticed the odd skipped frame and laggy response to skipping back & forth within the video.
So I fired up Peacekeeper, a straight forward browser benchmark tool. This revealed some very interesting information.
Look at Firefox go. I'm a Chrome guy myself, but poor old Microsoft still haven't got it right apparently and their Edge browser is, at best, the same as IE.
Before you laugh off Edge completely, take a look at this:
This represents how much faster each browser ran on the i5 than it did on the Core2, presented in the same order as they are on the previous graph. Not only is Firefox the fastest on an old processor, but it gains the smallest improvement when handed a faster procesor. This points to it being the most efficient, getting the most done for the least CPU time.
The Microsoft offerings are quite the opposite; give them a slow procesor and they chug along at between hal;f and two thirds the speed of their rivals. Give them a fast CPU and they suddenly take off, gobbling up CPU time in the process.
All of that said, none of the eight user experiences were what I would call slow on the old rig. Even the power-hungry Edge browser was (almost) always smooth, with the ancient CPU and painfully slow ram not holding up the show.
But can it run Crysis?
The coloured graph is the new machine, the grey line is the old Core2 machine.
Here's a detailed explanation of frame times to those who don't know.
Crib notes: each dot on this graph represents a frame and how long it took to display. A lower line means an overall higher FPS. A flatter graph means a more stable frame rate. High values on the right represent stuttering. The higher the value the longer the stutter. The more dots that are up there, the more often the stutter happens.
When you're benchmarking a system like this, which potentially has a serious bottleneck feeding the GTX980, frame times will paint a much more accurate picture than a simple FPS measure.
As you can see, not only does the new machine average a nearly 50% higher frame rate overall, but the old machine has around 10% of its frames hanging around on the screen for over 33ms, which equates to around 30fps.
Again, the old Core2 machine puts in a commendable effort. Averaging 36fps in something as demanding as Crysis 3 on maximum detail is no mean feat.
Now for something a little easier; Tomb Raider.
First the canned benchmark:
...and then some real-world gameplay:
Wow. Not only is the old machine keeping within 4fps of the new rig but the frame time curves are just about parallel, showing that the rig isn't really tripping over its own shoelaces, it's just generally running a little slower.
I initially scoffed at both Crysis 3 and Tomb Raider's minimum CPU and ram requirements, but if this test is anything to go by then they were spot on.
Ok, for those resisting the frame time view of things, dust off your monocle and see below:
...and finally, what would any good benchmark article be without Firestrike?
Granted, the 4570T is not a beast, but those scores are much closer than I ever would have guessed.
In conclusion, my belief that your CPU and ram make very little difference to gaming has been upheld. This is a bit of an extreme case, but if you're sitting with an 1151 or 1155 rig at the moment, your next CUD move should probably still be a GPU.
In fact, if you have a strong 775 rig you're probably still fine if you're not obsessed with maintaining a solid 60fps. LGA775 motherboard, CPU and ram combos go on carb for way under R800 all the time; just think, you could build a R2500 box off Carbonite that could comfortably run Crysis 3.
Info:
GPU: Gigabyte GTX980 G1 Gaming
SSD: Samsung 840 Evo 250GB
Core2 Rig:
Intel DQ35MP-E Motherboard
Intel Q6600 Core2Quad 2.4GHz CPU
4 x 1GB Samsung DDR2 800Mhz ram
i5 Rig:
MSI H97 Gaming 3 Motherbaord
Intel Core i5 4570T 2.9GHz CPU
4 x 4GB Corsair Vengeance Pro DDR3 ram @1600Mhz
Last edited: