For the lulz... i5-4670k + HD4600 gaming performance

[BallaTheFeared]

I don't think anyone cares about the performance of the HD4600, but I was bored so I figured I'd see what it could do.

Feel free to mock it all you like, I hold no illusions that this combination makes any sense what so ever... It's a mismatch of performance in epic proportions.

System Setup:

i5-4670k @ 4.6GHz 1.155v, 4.5GHz Uncore 1.14v
HD4600 @ 1750MHz 1.225v
DDR3 1600 1.5v @ 2400 11-13-12-37 1T 1.7v
Mobo MSI Z87 GD65 Gaming
HX850 delivers the power
M4 SSD
F3 Spinpoint
WD Blue

Cooler is a Dark Knight II, with two Corsair SP 120mm PWN fans. Ram is cooled by the cpu cooler from my 1090T stock cooler :thumbsup:

First test I did just to see what would happen, it's BF3 on a 64 player map. I just mimicked AtenRa's settings to see what would happen, I wasn't excepting much.

http://www.youtube.com/watch?v=9M-pXhfg8sI

Actually got a hang here at the end, I bumped the voltage up slightly going forward, I haven't spent much time with the iGPU overclock forgive me :|


The second game I ran was Crysis 3. I got a bit more bold with my settings in this title, still at the low low resolution of 768 but it is what it is. However I did bump my settings up to Medium across the board, enabled FXAA, and used High for textures. Aside from the awful resolution, the game actually looked decent at these settings, probably better than what consoles got

http://www.youtube.com/watch?v=BvfV0inlbBk


It's actually kind of interesting to me to see what the HD4600 can pull with a 46% overclock. I was planning to do Metro 2033 Last Light. If I do I'll spend more time tweaking settings to get the best IQ with decent frame rates with it.

If anyone would like to see some different games, if I own them I'll see what I can do. I'm also going to try to find a tripod or something I can use to give a better view of the screen, not just the kill-a-watt and frame counter.

 

[Sweepr]

Very interesting. OCed HD4600 seems like a solid performer despite having only half the EUs of GT3 and no eDRAM. It would be nice if you could test Bioshock Infinite.

 

[NTMBK]

Cool post As much as they're undeniably slower, I find it more interesting to see what you can squeeze out of these things.

 

[SPBHM]

not bad for an intel IGP I guess, can you run 3dmark11 or Vantage with the OC?

I think you can overclock the IGP on locked CPUs right?

because even the cheapest i5 uses the HD4600, and i3s should probably also use the same,

 

[Durvelle27]

Nice. Could you try Far Cry 3 & Tomb Raider

 

[BallaTheFeared]

test Bioshock Infinite.

Your wish is my command!

1366x768 @ VeryLow

Per Scene Stats:
Scene Duration (seconds), Average FPS, Min FPS, Max FPS, Scene Name
32.48, 64.86, 30.62, 129.47, Welcome Center
7.12, 60.11, 31.36, 86.20, Scene Change: Disregard Performance In This Section
21.84, 70.18, 36.12, 96.85, Town Center
8.14, 69.32, 51.49, 110.94, Raffle
9.11, 88.98, 60.58, 135.25, Monument Island
3.03, 89.00, 48.06, 206.65, Benchmark Finished: Disregard Performance In This Section
81.73, 69.91, 30.62, 206.65, Overall

1366x768 @ Low

Per Scene Stats:
Scene Duration (seconds), Average FPS, Min FPS, Max FPS, Scene Name
32.68, 47.16, 25.69, 90.38, Welcome Center
6.67, 43.64, 20.33, 79.50, Scene Change: Disregard Performance In This Section
21.92, 48.17, 29.35, 79.89, Town Center
8.17, 43.79, 31.33, 63.37, Raffle
9.16, 59.04, 31.10, 80.69, Monument Island
3.05, 59.97, 54.33, 68.32, Benchmark Finished: Disregard Performance In This Section
81.66, 48.62, 20.33, 90.38, Overall

1366x768 @ Medium

Per Scene Stats:
Scene Duration (seconds), Average FPS, Min FPS, Max FPS, Scene Name
32.31, 39.91, 19.14, 138.95, Welcome Center
7.21, 36.53, 17.99, 63.05, Scene Change: Disregard Performance In This Section
21.73, 35.86, 12.15, 90.28, Town Center
8.25, 33.23, 8.07, 66.27, Raffle
9.19, 43.74, 23.25, 62.06, Monument Island
3.07, 43.69, 27.53, 48.91, Benchmark Finished: Disregard Performance In This Section
81.75, 38.44, 8.07, 138.95, Overall

1366x768 @ High

Per Scene Stats:
Scene Duration (seconds), Average FPS, Min FPS, Max FPS, Scene Name
32.46, 32.06, 14.52, 123.30, Welcome Center
7.23, 30.44, 11.16, 52.32, Scene Change: Disregard Performance In This Section
21.74, 32.86, 12.41, 57.04, Town Center
8.19, 30.25, 6.17, 60.08, Raffle
9.18, 40.76, 15.52, 63.59, Monument Island
3.06, 40.90, 26.64, 45.67, Benchmark Finished: Disregard Performance In This Section
81.86, 33.26, 6.17, 123.30, Overall

1366x768 @ VeryHigh

Per Scene Stats:
Scene Duration (seconds), Average FPS, Min FPS, Max FPS, Scene Name
32.77, 27.37, 12.12, 52.16, Welcome Center
6.79, 26.08, 7.42, 39.54, Scene Change: Disregard Performance In This Section
21.84, 26.97, 9.03, 49.73, Town Center
8.44, 24.94, 4.68, 51.19, Raffle
8.78, 34.48, 12.89, 62.44, Monument Island
3.60, 36.09, 24.22, 39.51, Benchmark Finished: Disregard Performance In This Section
82.21, 28.06, 4.68, 62.44, Overall


It seems to have issues in one area once the frame buffer gets too large. I will see what I can do as far as tweaking IQ settings and with actual gameplay when I wake up later today.

 

[Sweepr]

Thanks!

 

[BallaTheFeared]

try Far Cry 3 & Tomb Raider

I'll have to reinstall FC3, I'll do it when I wake up.

TR just did the built in benchmark for now, I'll do some actual gameplay (again when I wake up).

1366x768 Low

1366x768 Normal

1366x768 High

1366x768 Custom - High Present, 16x AF, dropped DoF and SSAO for Tessiliation

 

[Silverforce11]

What does it look like at 1080p?

 

[BallaTheFeared]

1080p "Normal" preset

 

[Durvelle27]

I'll have to reinstall FC3, I'll do it when I wake up.

TR just did the built in benchmark for now, I'll do some actual gameplay (again when I wake up).

1366x768 Low

1366x768 Normal

1366x768 High

1366x768 Custom - High Present, 16x AF, dropped DoF and SSAO for Tessiliation

Wow that's pretty good and ok will be waiting

 

[cytg111]

low res of course, still very playable, GT3 will definitely hold its own.

 

[AtenRa]

First test I did just to see what would happen, it's BF3 on a 64 player map. I just mimicked AtenRa's settings to see what would happen, I wasn't excepting much.

http://www.youtube.com/watch?v=9M-pXhfg8sI

Is that on Seine Crossing map ???

 

[BallaTheFeared]

No, I only have the limited edition BF3, nothing but B2K for me. I never got into it so I don't have any of the newer maps.

It was Noshahr Canals full sever, 64/64... There weren't a lot of options at the time :\

 

[monstercameron]

so, it is safe to assume that balla... wont be trolling apu threads anymore?

also, you should have bought the celeron g2020, and a 7750, much better gaming than a 4670...

 

[SPBHM]

also, you should have bought the celeron g2020, and a 7750, much better gaming than a 4670...

well, I'm pretty sure the TR and Bioshock performance would be MUCH higher with a Pentium G2020 + 7750.

but I thought it was obvious enough that this is not his main gaming GPU

 

[monstercameron]

well, I'm pretty sure the TR and Bioshock performance would be MUCH higher with a Pentium G2020 + 7750.

but I thought it was obvious enough that this is not his main gaming GPU

he lists his gpus in his sig but he did invite us to mock him. That was similar to the sentiments he posts in apu gaming threads

 

[Gikaseixas]

This is great news for people who want to build itx based pc's without the added consumption or heat from a dedicated GPU.
That also proves that you can indeed play with APUs or Intel equivalent at a moderate settings.

Can you test Diablo 3 and SC2 Balla?

 

[BallaTheFeared]

Borderlands 2

1366x768 - Max settings (save PhysX)

Area used was the first boss fight, with the guy on the cannon and his brother. Normally I use PhysX so I was probably over-estimating the amount of effects that would take place, but I did take a blast from the cannon to the face which dropped the fps down. After that I moved forward towards the mountain which has the first real boss, used a 3.5 minute fraps time to try to keep the video short so I only made it inside the gate before ending the video.

http://www.youtube.com/watch?v=OdfAYwEGFcg


FRAPS read-out:

Frames, Time (ms), Min, Max, Avg
8034, 210000, 23, 47, 38.257

Can you test Diablo 3 and SC2 Balla?

Diablo 1080p - Textures on High, Physics on High, Shadows on Medium (not smooth), Clutter Density on Medium (didn't really make a difference), AA enabled.

I did Diablo 3 for you, but I'm not sure how to go about benching SC2. Should I use single player? A custom game? 1v1 or 4v4, what area would I capture, a big battle towards the middle/end or other? Heh, not sure exactly how to go about it.

For Diablo 3 there is a freeze when loading new textures, I had to be careful because even though I don't really ever play anymore, I don't want to lose any of my hardcore characters, but at the same time I don't want to just kill the first level zombies.

I decided to use ML4 Hell difficulty on my 2H Whirlwind bard, I figured the best area to test would have to be Act III, on the bridge before Siegebreaker, so that was what I recorded.

http://www.youtube.com/watch?v=qZIoLEsKQ5s

FRAPS read-out:

Frames, Time (ms), Min, Max, Avg
7463, 210000, 12, 53, 35.538

I'm not sure if the freeze is because the frame buffer or because of the HDD, or what but it wasn't to the point where I felt I was at risk of losing my character.








Not sure if I should use high res, or high settings at low res. If I was to use this it would have to be used on a coach with a TV where I sat far away. At which point resolution wouldn't mean quite as much, though a lot of these chips will also come in low res laptops. As a desktop user though there are too many sacrifices needed, but considering my last encounter with Intel graphics was with GMA about four years ago or so it it couldn't play hardly anything I am pleasantly surprised at the progress being made to tighten up the low end performance sector.

 

[BallaTheFeared]

Metro Last Light

1366x768 - High Quality // SSAA Off // 16xAF // Motion Blur Low // Tessellation Normal

http://www.youtube.com/watch?v=W_78HOlW3R4

Frames, Time (ms), Min, Max, Avg
9189, 210000, 34, 60, 43.757

 

[Ventanni]

Definitely a potent iGPU. I used my HD4000 for a little bit while I figured out where my 285 was lol.

 

[Rvenger]

Nice work Balla.

 

[Silverforce11]

Is the scaling for the iGPU OC good? I can't get a sense since numbers are everywhere, perhaps just text with the results is good enough.

If it makes sense, I may build some mITX rigs for people with the HD4600, I've gone with A10s for a few already.

 

[BallaTheFeared]

Scaling is quite good, but I have no idea how good other chips will overclock. Or how much other factors are at play, such as the RAM and cpuNB speed.

It looks like at stock vs the A10 it's anywhere from a few percent slower to 50% slower (maybe more?).


If you look at some of Anands test, like Tomb Raider:

The HD4600 is only about 23% behind, if the scaling is anything close to decent you could easily surpass the A10 in this test, similarly in several others it should as well. While in some of the worst case where it's behind like 50% like in Bioshock, all you can hope for is to close the gap somewhat. It also strikes me as odd there is so much disparity between titles with Intel, that says to me their drivers are not as good as they could be, more so than to say the hardware isn't up to par.

Luckily Anand states they used the IG benchmark for this one, with a 46% gpu overclock, the 2400MHz ram, the overclocked cpu, I actually gained roughly 43% performance over the stock HD4600 in their test.

While it's still slower than the A10 in that title, you can still see the scaling is quite nice. Also the gap has closed up considerably from a rather large 50% performance advantage for the A10, to just a mere 6% advantage.

Obviously games where it's much closer to the A10 are going to show the HD4600 actually pulling away.

Though again I have to state clearly, my i5-4670k is pretty unique, if 10% or so hit 4.6GHz at 1.2v, than doing it at 1.155v means it's pretty special. How or if that translates in anyway to the iGPU I have no idea, I've never overclocked one before or actually used one my SB system was P67 and my i3-540 was P55.

 

[Silverforce11]

I don't normally see such massive OC on the Intel iGPU, which is why this thread is informative.

Generally A10s have 15-20% OC headroom on the iGPU, not that great.

If its the norm and HD4600 is a crazy good OC, it makes sense to build a mITX rig with Intel, you get the best CPU performance and acceptable iGPU performance, for a bit $$ extra, I think its worth it.