Confused about Max Payne 3 performance (CPU bottleneck?)

[guskline]

Originally Posted by Rvenger

I am curious to see how an FX-8350 responds in this game since the i7-920 performs similar to the FX chip.

I have Max Payne 3 and the FX 8350 is being shipped to me as I type this. I'm replacing the 8150 in rig 3 below. I'll keep you posted.

[lehtv]

OT: guskline, it says "Asus 300" in your first rig, you probably mean the case, Antec 300?

[guskline]

lehtv, sorry. Corrected. Should read Antec300

[lehtv]

Originally Posted by lehtv

I'll try it out with stock CPU clock tomorrow.

I tried it with medium-high settings. Didn't make any real difference

    Medium-high, CPU 2.8GHz
    Framerate    GPU load (%)
Min        51            42
Avg        109            84
Max        183            99

Originally Posted by lehtv

        Medium-high    
    Framerate    GPU load (%)
Min        49            62
Avg        110            87
Max        197            99

[Termie]

Therefore I think you can conclude that there is no cpu bottleneck.

Like I said, I've found this game to be ultra-smooth even at low framerates.

Perhaps it was just vsync causing the problem.

[Jaydip]

Originally Posted by Termie

Therefore I think you can conclude that there is no cpu bottleneck.

Like I said, I've found this game to be ultra-smooth even at low framerates.

Perhaps it was just vsync causing the problem.

How can you concur that? from lehtv's data

CPU 2.8GHz GPU load 42%
CPU 3.36GHz GPU load 62%

GPU load is getting increased with cpu clock rates identifying a cpu bottleneck.

[Termie]

Originally Posted by Jaydip

How can you concur that? from lehtv's data

CPU 2.8GHz GPU load 42%
CPU 3.36GHz GPU load 62%

GPU load is getting increased with cpu clock rates identifying a cpu bottleneck.

You're looking at minimums. The GPU load was ~85% at both settings.

[Jaydip]

If the GPU load increase while increasing CPU clock its a classic CPU bottleneck situation.I understand your point but we need a frame rate graph @ stock to determine this.

[lehtv]

I spotted a sweet deal on an i7-3770K + P8Z77-V, and grabbed it. They've been used a couple months, warranty and receipts left. Similarly priced new components would be 3570K + Z77 Extreme4.

I'll receive the parts in a few days, this will solve the mystery once and for all

[Jaydip]

Cool

[blckgrffn]

Originally Posted by lehtv

I spotted a sweet deal on an i7-3770K + P8Z77-V, and grabbed it. They've been used a couple months, warranty and receipts left. Similarly priced new components would be 3570K + Z77 Extreme4.

I'll receive the parts in a few days, this will solve the mystery once and for all

Haha, excellent!

[Termie]

Originally Posted by lehtv

I spotted a sweet deal on an i7-3770K + P8Z77-V, and grabbed it. They've been used a couple months, warranty and receipts left. Similarly priced new components would be 3570K + Z77 Extreme4.

I'll receive the parts in a few days, this will solve the mystery once and for all

That's one way to handle it!

You'll like the upgrade. Plenty of advantages of IVB/z77 over the older architecture, and given that your OC was limited, you'll have a lot more headroom with the new setup.

All that being said, I still don't think you were CPU-limited in MP3!

[lehtv]

Yeah I think you're right. But I already know I'm somewhat CPU limited in Alan Wake (uses only two threads), and more so in Battlefield 3 conquest. It's not all for nuthin'

I'm hoping for a nice 4.5 with the 3770K while keeping my Mugen 2 quiet

[Rvenger]

Originally Posted by lehtv

Yeah I think you're right. But I already know I'm somewhat CPU limited in Alan Wake (uses only two threads), and more so in Battlefield 3 conquest. It's not all for nuthin'

I'm hoping for a nice 4.5 with the 3770K while keeping my Mugen 2 quiet

I hope you get a good chip because I'm running at 4.2ghz and I am getting 76c in IBT with a Havik 140. I can safely crank the speeds if I am running a benchmark, I just can't run IBT without hitting high 80s at 4.4.

[lehtv]

What temps do you get in Prime95 small FFT's, Rvenger?

[Deders]

Originally Posted by lehtv

I play at medium-high because the game's vsync sucks. If I increase settings any more, the game forces to 30fps occasionally, when without vsync it wouldn't even break below 60fps. It's as if the game anticipates a dip below 60fps. But at 30fps the game just looks stuttery, so I go as low on the settings as needed to avoid the 30fps dips. This is not adaptive vsync, it's just an artificial fps limiter. FWIW I don't really notice an appreciable difference in image quality between medium/high and high/ultra.

You can use D3DOverrider to enable triple buffering so that it doesn't lock on to 30FPS when using Vsync. Mine came with RivaTuner224MSIMOA2009Edition.

[Rvenger]

Originally Posted by lehtv

What temps do you get in Prime95 small FFT's, Rvenger?

68c/71c/75c/66c after 30 mins.

[lehtv]

OK. I just tested my fresh 3770K at stock settings, it stays under 60C in Prime95 at the lowest fan speed (Noiseblocker B12-3 @ 5V). CPU-Z reports 3.7GHz. I think this is promising, as I can shave off 7-8C by running the fan at full speed, and that gives me around 30C of long-term safe headroom to play with

edit: time to edit sig

[Rvenger]

Let us know what your 4.2 and 4.4ghz voltage and temps are. Looks like you might not have a hot running chip afterall.

[lehtv]

Originally Posted by Rvenger

Let us know what your 4.2 and 4.4ghz voltage and temps are. Looks like you might not have a hot running chip afterall.

I think I've found my stable 4.2 settings, but I've yet to run Prime for several hours and I haven't even tested LinX or IBT yet. Conversely my vCore could be a bit higher than needed.

All power saving settings, C-states etc. are on default/auto. LLC at 0%. Offset voltage at -0.030. BCLK 100.00, turbo multi 42.

I've got two front 120mm 1700RPM fans, one rear 120mm 1300RPM fan and 1700RPM 120mm CPU fan.

Door closed, all fans at the lowest speed (~50%): 73c, 72c, 71c, 75c

Door open, front fans at ~50% speed, rear and CPU fans at 100% speed: 64c, 68c, 65c, 66c

I'll test with 4.4 later on, I have a pretty good idea of what settings to use.

[lehtv]

Back on topic. Max Payne 3.

I tested the same sequence with 3770K @ 4.2. Well, almost the same sequence. I couldn't get past a bug right after a cut scene in the beginning of Chapter 5, I think this was due to having to reinstall Windows 7 and Max Payne 3 and then using my backed-up profile from the last install. I started from the second checkpoint this time around, so the sequence was a bit shorter. I adjusted the original result to make it as if I'd started from the second checkpoint.

Here are the GPU load and framerate results at medium-high settings:



11 fps higher average framerate, 5% higher average GPU load. These could be somewhat influenced by differences between the runs, the i7-920 run having taken a couple of minutes longer. Note that my graphics card is also running at PCIe 3.0 x16 now, instead of 2.0 x16 as it was previously.

i7 920 @ 3.36


i7-3770K @ 4.20


By the looks of the graphs, both framerate and GPU load fluctuate less with 3770K. So that's good.

[Termie]

Looks about right to me. My own upgrade from i7-860@3.5 to i7-3770k@4.4 showed about a 3-4% increase in games that were not CPU-limited, which I attributed to the platform upgrade (PCIe 3.0, basically). Here are my results: http://forums.anandtech.com/showpost...58&postcount=2. You had 10% here, which is pretty substantial, but as you say, the sequence was a bit different.

As I would have predicted, the much more important number is the minimum framerate, rather than average framerate or average GPU usage. That's the one number you are absolutely going to feel. As an experiment, I'd suggest you try the same sequence at a CPU speed of 3.4 (i.e. underclocked) to get a clock-for-clock comparison. My guess is that the minimums will still be much higher. If so, it would tend to prove that even users with an i7-920/860@4.0 stand to gain a lot from an upgrade. If not, it tends to show that if you're already that highly clocked, you won't feel the difference.

Very cool graphs, by the way. What did you use to create them?

[lehtv]

I was thinking of doing a lower clocked bench as well, actually. I'll try 33 * 102 MHz = 3366 MHz.

Very cool graphs, by the way. What did you use to create them?

Thanks. OpenOffice Calc.