Prey running the GTX 1080Ti on 11 CPUs (more to come)

We have already published a Prey review on the GPU side. Now it’s time to deliver for you guys, our CPU analysis. There has been plenty of time dedicated for those tests but, at the end of the day, it turns out that we’re really proud to present this content to you.

Before start…

“(…) We miss Skylake / Kabylake family too… but in the next couple weeks we’ll have them here(…)”

For a financial reason, we can’t own 11 CPUs. What we precisely have here at the tests lab are 3 CPUs from each CPU platform or family. What we do to test all those 11 CPUs is to downcore ou turn off some of their resources like SMT or HT. To be fairly honest to you, let’s take a dive in every CPU condition for this test.

a) Haswell Family: Core i7 5960x fully enabled counts like 5960x (thx master of the obvious); disabling two of its cores, we got an i7 5930k; disabling more two cores, an i7 4790k and; turning Hyper Thread we got i5 4690k. (All of them at 4,4GHz)

b) Ryzen Family: Ryzen 7 1800x is in our hands representing himself on the CPU chart; disabling 2 cores we got 1600x; two cores more (or shoul I say less?! Whatever… you got the point) 1500x and; disabling SMT, we got a “brand new” or antecipated but not launched Ryzen 3 1300x. (All clocked at 4,0GHz)

c) Vishera Family: FX 8350 makes its true show and with two core (or one module) less, gives place to FX 6350, and one module less, to 4350. (All at 4,2GHz)

Test Benchs

a) Haswell:

ASUS X99 ROG Strix Gaming 2011-v3

16GB Geil EVO X at 3200MHz

Corsair H100i Cooling Solution

EVGA 1300G2

b) Ryzen:

MSI X370 Xpower Gaming TITANIUM (AGESA 1.0.0.4a)

16GB Corsair Vengeance at 3000MHz

Cooler Master MasterLiquid 240

NZXT Hale V2 700w

c) Vishera

ASUS 990FX TUF Sabertooth R3,0

8GB Kingston at 1600MHz

AMD Wraith Cooler

Corsair HX 750i

Some points about this…

1. Yes! We miss Skylake / Kabylake family too… but in the next couple weeks we’ll have them here (i7 7700k, (- 2 cores = 7600k and so on…)
2. Yes! We should’ve oced Vishera even further, but we still don’t have proper cooling solution.
3. Yes! It would be better if all the processors in fact were tested but… we do the way we can.
4. OK! We’d like more 8GB DDR3 for Vishera

“(…)the game code demands more than 8GB of RAM, excluding any possibility of RAM bottlenecking the Vishera Family(…)”

Finally, the results

“(…)4K… “where there’s no CPU bottlenecking”, they have said(…)”

1920x1080 resolution tested on all CPUs (Colored is AVG, Grey is 1% Low, and Black is 0,1% Low)

In 1080p resolution, it’s seenable all the scale that the game does with more cores in any architecture. Despite Ryzen being under an i7 4690k, don’t you remember that we are talking about a CPU with 4.4GHz clock vs a 4.0GHz clock. The game, probably demands a lot from clock (what is not bad at all) but it also takes advantage of more cores (great point). Let’s see what happens changing the aspect ratio.

Ultrawide resolution tested in all CPUs (Colored is AVG, Grey is 1% Low, and Black is 0,1% Low)

In UltraWide, same behavor, but with a less of GPU performance. It seem that the game puts all cores to work even if the IPC or Core Clock cant handle huge amount of FPS.

Quad HD tested in all CPUs (Colored is AVG, Grey is 1% Low, and Black is 0,1% Low)

Quad HD has another behavior… Vishera family remains on its ~90FPS wall, but the difference between Ryzen and Haswell has been lowered for the GPU bound scenario. An important point must be said: even with 5930k giving less average FPS, its 0,1% and 1% lows are better than its less cores brothers.

4K tested in all CPUs (Colored is AVG, Grey is 1% Low, and Black is 0,1% Low)

4K… “where there’s no CPU bottlenecking”, they have said… honestly…? The CPU itself almost won’t make a true gameplay experience to change, but the Vishera family is not recommended. They deliver not only above 60FPS experience (Core i5 does that) as they deliver terrible percentile with awful frampacing (that the core i5 does not)! Not recommended at all!

Please, don’t let yourself to be fooled by the average FPS. PAY ATTENTION on percentiles.

That’s not over… lets analyse the system behavior while running Prey

“(…) To be honest, even if you get “only” a Ryzen 1300x you will be pretty well served with GPU performance(…)”

Threads usage on Core i7 5960x at 4,4GHz

Now we can show you, how the game properly get advantage of every exclusive resource of the processor. We have a 5960x using 11 cores with more than 10% of average usage. Also, we have the first two cores been quite fully loaded assuring IPC gains. And even than, in much cases like the i5 4690k, when the core counting doesn’t help, there is the Core Clock to be the counter balance for it.

System Usage (Colored is GPU, Grey is Max Core and Black is CPU) (Baixo is Low, Medio is Medium, Alto is High and Muito Alto is Very High)

From the system usage perspective, we can see that only at 1440p High that the GPU is starting to be a bottleneck. In other words, for the GTX 1080Ti to be really on its maximum power, you should consider its performance in 1440p High and forth. But even in those cases, we have a high Main Thread, or Core 0 (as I call) over the 70% load. Thinking this as a thread that disputes “Tasks space” with its paired thread (every physical core corresponds to two logical threads) its something to be concerned that you have more than one single thread occupied (use the thread figure to find out). It’s a good thing to see that Prey uses lots of threads, and, despite the other game engines, CryEngine does NOT demmand lots of threads, it takes ADVANTAGE of it. So, if you have a Core i5, you shall not got into trouble.

To be honest, even if you get “only” a Ryzen 1300x you will be pretty well served with GPU performance (which assure you with more than 100 FPS). The bad point remains with the FX’s. They suffered in many aspects.

Memory Approach (Colored is RAM, Grey is Dynamic RAM and Black is VRAM) (Baixo is Low, Medio is Medium, Alto is High and Muito Alto is Very High)

If you just passed your eyes through the memory behavior chart, you’ll realize that we don’t need any GPU to have more than 4GB of VRAM (unless you’re talking about 4K medium and up). You realize too, that in any of those cases, the game code demands more than 8GB of RAM, excluding any possibility of RAM bottlenecking the Vishera Family. Dynamic RAM is obviously unnecessary in any case.

An optmized game, thx god! Well… Not this time

Then, yes, we can call it an optimized game.

At this point of the review, the game should be prized as an optimized game. But that would be truth if, and only if, the game was not so much filled with bugs everywhere. We are talking about texture popping, shadow vanishing and missbehaviors with screen resolution and aspects. I won’t mention the conflicts fullscreen exclusiveness, when some application make the game crashes or freeze because they have some screen priorities that conflicts with the Denuvo (or any other system DRM).

Shadows vanishing

Texture Popping

We are pretty sure that all those screenshots are happening in every and all GPUs we have tested (22). It’s a game problem. If I have to guess, I’d say its some culling and clipping problem. Probably this could be fixed in next patch updates. Then, yes, we can call it an optimized game.

See you there…

My best reguards,

Marcus Sarmanho

CEO

PC Facts