APU results Just Cause 3

[DrMrLordX]

The 7870K with a discrete card should turn in basically the same results? Both 95W, and clocks are only 100-200 MHz apart, both steamroller (kv/gv).

Yes, especially if the iGPU is completely disabled.

 

[Madpacket]

I'm working on getting a copy for testing. Don't trust those videos, something with the hitching isn't right.

 

[Madpacket]

Okay finally snagged a Steam copy of JC3 for $25 CAD. Installing on my HTPC now but going to take a while at over 50GB install size sheesh. Should be done later tonight but if too late I'll test tomorrow night.

System I'll be performing the testing on:

AMD 7850K @ 4.0Ghz (4.2 turbo), 8GB 2400Mhz G.Skill TridentX RAM (running at 2133 due to weak APU Northbridge) with decent timings and 2GB dedicated to graphics and running at 860Mhz.

The motherboard is a Gigabyte mATX FM2+ A88X budget Microcenter combo board (forget model # but it's pretty budget oriented but solid so far), Intel 240GB 335 SSD for system storage. I'm only cooling the CPU with a Noctua NHL9A in a Silverstone ML03B case with two 80mm intake case fans but this setup has been working suprisingly well with mild overclocking. Oh and 430W Corsair Bronze modular power supply with a 16X DVD-ROM and a 6TB WD Red but it's half dead and needs an RMA, just listing everything for completion reasons.

I may also run at 2400Mhz memory speeds but have to loosen the timings a bit and up the NB voltage which adds more heat. I haven't adequately benchmarked which is faster but if 2400 ends up being quicker I'll sacrifice a little CPU speed to get the RAM working. Windows 10 OS up to date etc.

Anyways results should come in the next few days so we'll have independent review of how the game runs. Fun fun..

Oh, also grabbed the new Tomb Raider for 12 CAD through a business logic bypass in the Microsoft store, perhaps I'll bench that as well. It also uses the Denuvo copy left protection.

 

[DrMrLordX]

Nice, looking forward to seeing your results.

 

[xorbe]

Is that 8GB you got 1or 2 sticks of ram? I am guessing 2.

Yeah I could select 2400 and 2133, but it only ever showed 1866 in the bios / cpu-z after reboot. I guess my mini-itx board is a bit older.

I realized that I had accidentally run Crysis with just 512MB allocated for graphics. Super surprised it ran half-way decently on med/high. That game is getting pretty old now though.

 

[Madpacket]

Okay guys great news. As suspected the source video posted by the OP is complete bull****. The game runs perfect at 720P with zero hitching problems like the video shows.

I only had time to play through the first mission and a bit of the second (like in the source video) but the game is averaging around 35-45 FPS and dips occasionally to just under 30. It never reaches the 1-2 FPS or hitches when loading new textures as the video shows, it never becomes unplayable even during massive explosions with tons of smoke etc. The game is perfectly responsive and playable at least for entirely the first mission and the beginning of the second where you're blowing everything up.

The $350+ Intel chips with embedded DRAM are still faster but the CPU bumped 7850K at 2133 RAM (two sticks) speeds run it very well at 720P and low details. I'm sure I could turn up the resolution or details and still maintain a solid framerate.

As I said I didn't do extensive testing but the CPU usage averaged around 80% across 4 cores during the mission unlike the source Russian source shows the AMD CPU running full out.

The only thing that surprised me all little was the amount of DDR RAM the game uses. Of the 6GB or so left over (2GB dedicated to APU GPU) the game appears to use 4.5 to 5.5GB of it. Perhaps if I only allocate 1GB for video and 7GB for the OS it would speed up the loading timed which I found a tad slow.

Anyways that's my 45 minutes or so experience with JC3 and a 7850K.

tldr; Game runs just fine on 7850/7870 APU's, source RU video is simply incorrect on how the game runs on AMD APU's.

 

[DrMrLordX]

Did you try changing the amount of RAM dedicated to graphics from 2 Gb to 1 Gb? It should be interesting to see how that changes things (same settings).

 

[xorbe]

Perhaps ... it would speed up the loading timed which I found a tad slow.

Nah even with 2x8GB it took a while to load here on my A10-6700 <- with good SSD.
I could try the 1GB vram thing for kicks. Yeah, it stutters a bit with 1GB vram.
With 512MB vram, I saw 1-3 fps instances at 1280x720.
512MB vram + 1x8GB ram, it's truly unplayable (big 5 second pauses).
2GB vram + 1x8GB ram, it runs about 15 fps but steady.

Did they really get their 7870K running with DDR3-2666 ram?

 

[Madpacket]

Did you try changing the amount of RAM dedicated to graphics from 2 Gb to 1 Gb? It should be interesting to see how that changes things (same settings).

Nope, will test that tonight if I have time. I'll check with GPU-Z to see how much memory is being used at when forced to 1GB and if it has any effect on performance.

It would also be great to test at higher resolutions with more detail added as I suspect the 7870K would close the gap with the i7-5770C like in other games. Too bad I don't have a i7-5770C on hand to test with.

Did they really get their 7870K running with DDR3-2666 ram?

I doubt it. That's probably the speeds for the DDR4 Intel processors? Just a guess.

 

[Madpacket]

Interesting results in this game. game.gpu results

Iris pro destroys fastest AMD apu, while even HD530 is faster, especially minumums. Have a feeling the AMD apus are being held back by the cpu.

Of course, overall it just shows how inadequate APUs really are for any kind of serious gaming. Even at 720p only iris pro is able to average 30FPS.

You should probably update your thread to reflect how these APU's actually perform. Just sayin'

 

[frozentundra123456]

And what results from a reliable testing site would those be? If you have such results, feel free to link them.

 

[xorbe]

And what results from a reliable testing site would those be? If you have such results, feel free to link them.

You've got 2 posters here that confirmed we don't get the huge 1-2fps stutter-fest as shown in the fine article's video. It's not even worth discussing the article any longer, clearly something is up with their configuration that isn't disclosed.

Next up, we can discuss why 2/3rds of the Dell Inspiron 3656 "gamer edition" machines with CZ APUs come with just 1 stick of memory ...

 

[frozentundra123456]

Be that as it may, the poster I replied to was somewhat snidely implying that I was posting incorrect, misleading data; implying that I should "update" the thread. I only posted the results from a popular review site. If that site has more recent data, or another site has different data that explains the "anomaly" as I said, he is free to link it. I dont think I should be expected to "update" a thread with another posters data.

 

[AnandThenMan]

Be that as it may, the poster I replied to was somewhat snidely implying that I was posting incorrect, misleading data; implying that I should "update" the thread. I only posted the results from a popular review site. If that site has more recent data, or another site has different data that explains the "anomaly" as I said, he is free to link it. I dont think I should be expected to "update" a thread with another posters data.

Why not? The more info the better but it seems you've already made up your mind based on what you wrote here and don't want any data that contradicts.

Iris pro destroys fastest AMD apu, while even HD530 is faster, especially minumums. Have a feeling the AMD apus are being held back by the cpu.

Of course, overall it just shows how inadequate APUs really are for any kind of serious gaming. Even at 720p only iris pro is able to average 30FPS.

 

[xorbe]

Be that as it may, the poster I replied to was somewhat snidely implying that I was posting incorrect, misleading data; implying that I should "update" the thread. I only posted the results from a popular review site. If that site has more recent data, or another site has different data that explains the "anomaly" as I said, he is free to link it. I dont think I should be expected to "update" a thread with another posters data.

Quoting for posterity! :wub: We took the time to try to confirm for you, but never mind.

 

[Madpacket]

Be that as it may, the poster I replied to was somewhat snidely implying that I was posting incorrect, misleading data; implying that I should "update" the thread. I only posted the results from a popular review site. If that site has more recent data, or another site has different data that explains the "anomaly" as I said, he is free to link it. I dont think I should be expected to "update" a thread with another posters data.

I didn't mean to come off snidely, at least that wasn't my intention. But looking back at my post it does come off a bit rude. My apologies. I just wanted you to update the OP as the article is clearly wrong and we went through the effort to debunk their results. It's clear from our testing they've done something wrong.

I even bought the game to test this (wasn't really planning on it).

If you want me to record a video I can try to do that but I don't have anything but a smartphone to capture the footage so it'll be nasty.

 

[frozentundra123456]

Thank you madpacket for the nice reply. No need to post a video.

 

[Madpacket]

Thank you madpacket for the nice reply. No need to post a video.

No worries!

 

[DrMrLordX]

Nope, will test that tonight if I have time. I'll check with GPU-Z to see how much memory is being used at when forced to 1GB and if it has any effect on performance.

It would also be great to test at higher resolutions with more detail added as I suspect the 7870K would close the gap with the i7-5770C like in other games. Too bad I don't have a i7-5770C on hand to test with.

Okay, thanks. I'm not sure that increasing resolution will improve things much (it might just drive all the competitors into unplayable territory) but it might be interesting to see anyway.

I doubt it. That's probably the speeds for the DDR4 Intel processors? Just a guess.

Probably, though it is possible to run memory at those speeds on Kaveri, with some trickery. There were a few rare people that were able to do it with AMP/XMP profiles on the Crossblade Ranger which advertised support for the memory speed. I'm pretty sure activating the profiles just overclocks the bclk or something like that.