i3-540@4.2ghz vs i5-3470@3.2ghz (3.6ghz turbo)

[bangoman 2]

Everything else being equal (except RAM - the Ivy Bridge CPU has 1600mhz RAM attached to the mobo), which would be better for FSX?

 

Current setup:

- The i3-540 is my main desktop machine running with a GTX460 connected to a 1080p monitor overclocked to 4.2ghz (lower than before as the CPU is a couple of years old)

- The i5-3470 is currently a headless server running multiple virtual machines with 16gb of 1600mhz RAM with a Gigabyte B75 mobo.

 

The i3-540 (running on 2C/2T, HT disabled) has a higher out and out frequency but does the "lower" clocked IB CPU outperform it's older brother at 3.2/3.6 turbo. It's on a Gigabyte B75 motherboard which isn't supposed to be able to OC but I have to look closely into whether the BCLK can somehow be changed. As I understand it, FSX doesn't take much advantage of the i5's 2 extra cores.

 

Intel comparo: http://ark.intel.com/compare/46473,68316

 

I'm currently running a GTX460 but have just picked up a GTX670 for cheap (CAD310) as well as three secondhand Dell monitors 1600x1200 (yes 4:3!) for a triple monitor setup. I was just going to drop it into the i3-540 machine but thought I'd see what you guys thought about dropping the i5-3470 in there.

 

Would my Ivy Bridge i5 yield significant gains? (No further purchases are allowed for this scenario - I'm waiting for Haswell)

[dazz 69]

I've always been a fan of those I3's, too bad Intel has restricted OC in the newest I3 lineup. But...

 

Going by the numbers, the I5 has about a 20% higher IPC, so that's like an I3 at 3.8 - 3.9GHz give or take.

With 4 cores you also have the option to run an affinity mask that will keep the main scheduler from sharing CPU cycles with the fibers (as Jesús Altuve aka bojote tought us all) hence improving smoothness.

 

It's true that the two extra cores won't help with FPS too much. In my tests it was about a 10 - 15%, but that's enough, coupled with the higher IPC, to beat the I3. With two extra cores you will probably get better texture loading too.

 

All in all, I would pick the Ivy bridge I5 for FSX

[bangoman 2]

Thanks for your answer, dazz, I remember our threads when Clarkdale was out and I was experimenting with the i3-540, which has to be the best bang for buck technological thing I've ever bought (been building PC's since the 486/DX2 lol).

 

I think I'll end up trying both - the Clarkdale first just to see how the bottleneck is. I guess the IB CPU/mobo combo also has PCI 3.0 - maybe the extra CPU <-> GPU might have benefit, who knows.

 

So much of the analysis here focuses on the K-series SB/IB CPU's (with good reason), it will be interesting to see what a "regular" IB CPU can do.

 

(Having said all of that, the i3-540 definitely won't do the job that the i5-3470 is doing right now...)

[SAAB340 28]

How fast could you run your i3-540 with HT on? That combined with AffinityMask=9 would give you the other threads(fibers) offloaded as well and that way give you higher Avg FPS.

FPS wise i think they will be equal or possibly slightly higher on your i3 depending on how fast you can use it with HT on. But the i5 will give you faster texture loading and load times no doubt.

[bangoman 2]

Hi, I've never tried OC'ing with HT turned on - all the guides say turn it off. I've never really benchmarked or really taken notice of HT. I'm already of the arbitrary opinion that the 2 extra threads of HT don't contribute anything useful.

 

Other considerations/things going through my mind:

- Faster (1600mhz) RAM with the IB CPU - I remember a thread around here describing the difference fast RAM makes.

- PCI 3.0 - ??? better bandwidth between CPU and GPU.

- Higher IPC for IB. What if I could OC the BCLK to get high 3's on the i5-3470 (like 3.8ghz). How would that change the comparison?

 

I don't have much spare time at the moment otherwise I would try both and report back.

[SAAB340 28]

I'm already of the arbitrary opinion that the 2 extra threads of HT don't contribute anything useful.

Well they could. In particular with a Dual-core with HT using AffinityMask=9. Read more about it here. But it also depends on how much activating HT limits your OC.

 

- Faster (1600mhz) RAM with the IB CPU - I remember a thread around here describing the difference fast RAM makes.

It will help a little bit. A few %. But not a huge amount.

 

- PCI 3.0 - ??? better bandwidth between CPU and GPU.

It will only be of benefit at high FPS with a lot of autogen. And if you use the 1/2 rate v-sync tweak (and I would strongly recommend people to do that) it's a non issue.

 

- Higher IPC for IB. What if I could OC the BCLK to get high 3's on the i5-3470 (like 3.8ghz). How would that change the comparison?

Only a % or two to gain there as well. Nothing that will make a big change.

 

The i5 Ivy will give you better load times and texture loading at an FPS that probably only can be differentiated with a benchmark from the i3. So if you can spare the time to move your FSX installation over to the i5 you'll actually notice the improved load times and possible better texture loading. I doubt you'll se much difference in FPS with the naked eye.

[bangoman 2]

Thanks for the comprehensive reply!

 

Sorry, my time comment was more relating to doing some proper benchmarks with the two setups, and not having the time to do repeated clean installs etc..

 

I've almost convinced myself to move the i5 into my desktop PC - esp when I throw Photoshop and Adobe Premiere into the mix.

[bangoman 2]

The answer to the question nobody asked

 

So I had a chance to benchmark the two setups. Sorry for the non std pictures, I used Google docs. IB > OC Clarkdale. Swapping the mobo/CPU didn't play nicely with Win7 so a fresh install was done for the i5-3470. And therefore a fresh install of FSX for each setup and the FSXMark instructions followed to the letter.

 

Overall I'm pretty happy with the IB setup. B75 mobo, i5-3470 CPU and 16gb DDR3-1600 RAM was about $350 including taxes and shipping. Of course if buying specifically for FSX, I would've stumped up the extra $100 or so all in for a suitable K-series CPU and motherboard.

 

 

 

 

 

 

[SAAB340 28]

The answer to the question nobody asked

Thanks for posting your results. IvyBridge is certainly very competent compared to Lynnfield. The difference was a lot more than I expected.

Did you have HT activated on the overclocked results as well with the i3-540 and did you use the AM=5 that the tweaking tool assigns or did you use the non standard (but better) AM=9?

 

My guess is that you used AM=5...?

[bangoman 2]

Thanks for posting your results. IvyBridge is certainly very competent compared to Lynnfield. The difference was a lot more than I expected.

Did you have HT activated on the overclocked results as well with the i3-540 and did you use the AM=5 that the tweaking tool assigns or did you use the non standard (but better) AM=9?

 

My guess is that you used AM=5...?

 

This is Clarkdale i3-540 not Lynnfield.

I just followed the FSXMark11 instructions to the letter. Excuse my ignorance, I'll have to look up AM=5/AM=9 lol

HT is off for the 4140mhz OC'd i3-540. I can't get much past 3800mhz with HT on.

What is neat is that the whole system fits into a Silverstone SG02, making for a great LAN box (no LAN gaming with FSX though!). It doesn't need a big cooler as I'm not doing any overclocking.

[SAAB340 28]

This is Clarkdale i3-540 not Lynnfield.

I just followed the FSXMark11 instructions to the letter. Excuse my ignorance, I'll have to look up AM=5/AM=9 lol

HT is off for the 4140mhz OC'd i3-540. I can't get much past 3800mhz with HT on.

What is neat is that the whole system fits into a Silverstone SG02, making for a great LAN box (no LAN gaming with FSX though!). It doesn't need a big cooler as I'm not doing any overclocking.

Sorry, my bad. Clarkdale it is.

 

Having just dual-core or dual-core+HT AM=5 normally generate a big difference between the benchmark runs as the other threads (or fibers as many like to call them) end up jumping between the physical core the Main Thread that produces the FPS resides on and the other physical core that has a texture loader on it. With HTon and AM=9, the other threads tend to stay on the same physical core the texture loader is on resulting in a consistent better FPS.

 

As you can see you only 1-2.7% difference between the runs on the Ivy Setup while the Clarkdale setup is varying 3.8-11.9% between runs. But even with HTon and AM=9 the FPS would only be a little over or at the fastest benchmark runs you have so it wouldn't close the FPS gap by too much.