2016 Hardware Benchmark - DCS World 1.5.x

That's why I'm mainly looking at the R5 1600x. It might boast better IPC with that same 4GHz clock.

May I humbly ask what makes you think so regarding 1600X? As far as I understand it will have exactly the same cores as R7, however with two of them disabled. IPC wise it should be totally identical to R7 with the exception that the clock speeds will run a bit higher. From the many different sources already we know that the upper frequency limit or Ryzen is around ~4.1Ghz, so there's not going to be too much headroom for overclocking. Price-wise, of course the 1600X should look quite good.

I am now banging my head against the purchasing decision: 1700 vs 7700K. I am going to use it *mostly* for gaming (read: DCS), so 7700K wins hands off here. However, once I do the next upgrade in n years, this CPU will end up in my home server where 1700 is more preferable.

In that case, 7700k hands down, even when Ed decides to multicore Edge I'm almost certain it won't need more than 8 threads.

My case however is different I play star citizen and do work on my PC, so this lackluster single thread performance on the Ryzen is causing me some grief while trying to decide

Bitmaster, you mean clock, not IPC. IPC = instructions per cycle. Nothing is getting anywhere near that much IPC. According to my calculation based off Wikipedia (*cringe*) data, the 1800x has 5.2 IPC per thread or 10.4 IPC per core. I could find no data on the 7700k but the 4770k gets 8.57 IPC per physical core. Regardless, total processing power is IPC*clock speed (over simplified, but gives a rough idea).

 

Nitpicking aside, I thought Hiromachi bought a 1700x. Or is our Hiromachi different from the one on the BoS forums? In regards to minimums, I noticed that too in a lot of the posted benchmarks. I agree that they are more important. That said, DCS loves its single thread performance and I don't think any of the R7's are made to complete with the I7-7700k in that regard. That's why I'm mainly looking at the R5 1600x. It might boast better IPC with that same 4GHz clock.

Kind of, IPC derives from the slices available. A 1 Hz CPU has 1 Hz/clocking and 1 cycle. I think cycle, slice and clock are 3 names for 1 thing.

How many instructions a cpu can process in 1 timeslice = IPC, that is correct.

The IPC gain from SB to KL might be 15%, guessed from my own benching.

Basically, at the same clock speed the KL gains 200 Points more in Passmark, the most !

Tetsed with 4.8 and 5 GHz, Passmark ver. 9.0

As I dont have the SB CPU anymore I have to tell you those numbers out of memory. There are pics in the forum somewhere from that rig if needed.

Anyway, the real deal is how it flies DCS. We all already know it's a monster in any multicore app.

I hope OP would make DCS 2.0 benchmark for new cards such as 1080, 1070, 1060, RX 480. Old DCS engine loved NVIDIA. How about DCS 2.0 ? Is that still the case ?

I'm planning to buy a new card but still waiting Normandy map to see I can justify the upgrade. Currently happy with MSI GTX 960 4Gb.

An article I just saw on the Ryzen cpu. Thought some of the others here might find it an interesting read...

https://arstechnica.com/gadgets/2017/03/intel-still-beats-ryzen-at-games-but-how-much-does-it-matter/

Sent from my E6653 using Tapatalk

Afaik no one yet posted DCS results for Ryzen.

I am waiting on that too... as mITX and Ryzen seem a perfect combo for a living room PC and NAS, the one that does it all + occasional gaming at 65w and fair priced !

I would favour a 1600 or 1700, maybe with X, maybe not. 65w is a deal for a 24/7 runner ( downclock too to save energy as NAS )

I was almost convinced after weeks upon weeks of reading about Ram & Gaming that I was overthinking it all and would be just fine with getting a lower frequency Ram,, until I reviewed the test results here. Ram prices have been on the rise for a few months and will be some time before they get back to reasonable prices. I am about to build a new rig and the only parts I have not purchased yet is the Ram. I had no idea a few months ago that the prices were going to rise so fast. Anyway, I had convinced myself to buy this:

G.SKILL 32GB (2 x 16GB) TridentZ Series DDR4 PC4-25600 3200MHZ 288-Pin F4-3200C15D-32GTZ 15-15-15-35 Voltage 1.35V

https://www.amazon.com/s/ref=nb_sb_noss?url=search-alias%3Daps&field-keywords=3200C15D-32GTZ+

When I 1st put it on the radar, it was $266, now its $350. I can wait till the end of the year if I have to for price drops if need be but was wondering something that I could not extract from the tests. And that is, ok, say I go with 3200, should I then worry about CL of 14, 15 or 16? There are substantial price differences, especially with tighter timings. I opted for going with 32 gig as opposed to 16 because of one of the posts here that someone stated they were monitoring their ram usage for DCS for NTTR and had occasionally went over 16 gig. So right now, my head is spinning. Any suggestions would be appreciated. I am running a 30” monitor for Main screen and a 19” touch screen for Helios and currently average about 30+ FPS on my current system.

New PC Specs=

Intel® - Core™ i7-6700K 4.0GHz Processor - Silver BX80662I76700K (planning on overclocking)

CPU Liquid Cooler Corsair Hydro Series H115i Extreme Performance

GIGABYTE - ATX Motherboard 3866MHz (Socket LGA1151) GA-Z170X-Gaming 7 (rev. 1.0)

Samsung 850 EVO 500GB 2.5-Inch SATA III Internal SSD (MZ-75E500B/AM)

SeaSonic Platinum SS-860XP2 860W ATX12V / EPS12V SLI Ready CrossFire Ready 80 PLUS PLATINUM

EVGA GTX 980 TI Hybrid GPU

I monitor RAM all the time and it never goes below 16GB if you have 32GB. When I had 16GB it always took like 11-13GB max, there wasnt just more to grab. This counts for the whole system, incl each and every Cache and OS and TiR, TS, SRS, BF-Gadget, ..etc... but everything not needed is OFF.

I bought 32GB with the mindset "to be furtureproof" and as soon as I had it, with the new rig, I saw it is actually used to some extend in DCS. It sometimes swells to 22GB RAM total, tho with the last patches it settles around 16-18GB, always ( online ).

The speed matters and the Latency values, don't trade one for the other, good values on both is the goal. Afaik all or most of those recent fast DDR4 are based on the Samsung-B_die and you can have a look at google for the details and specs for that kind of chip. 3866 would be awesome at CL15, CL16 much more likely to find as the fastet one can still afford and CL18 will be the main stream 3866. When I bought mine there where mine with CL15 and same model with CL18 for like 50€ less. I favoured latency over cashback and took the CL15. Dont regret it.

I wouldnt press the RAM to the limit of the board, my limit is 3866 as well and 3600XMP was the only one I got stable, rock stable. Any higher speed or tighter latency will eventually drop a core in p95 etc... or worse...depending on how much you push it. I am sure with a 4166 board I had good chances to get 3866 at same latency with just some more Volts, say 1.45 instead of 1.40V vDIMM.

I blame my board to be the limit here, and I have FOUR modules, not an overclockers dream setup, 2 dimms oc a lot better than 4 do !

But till today I am unaware of 16GB DDR4 modules faster than 3200 or 3400. The fastest 16GB I had was 3200 and my new 3600er set only comes as 4x8, 2x16 is not available from any vendor last time I checked. That forces you to 4 modules for 32GB and little oc headroom or 2 modules with max 16GB but better oc headroom.

I would get 32GB and buy the speed you want to run, dont oc them if you oc ypur CPU high as well. There are certain limits when mixing those oc things together.

Very informative response, many thanks. Yeah, so it seems I should not make a sacrifice and go below 3200 and try and stick around a CL 15 with tight timings,,, so basically, the ram I am looking at would sort of be the sweet spot it seems.

Yes, ypour choice is OK, 3200@CL15 in a 2x16GB pack is great. The bigger the modules the looser the timings. With 16GB it will be hard to make it CL14 or better. With 8GB or even 4GB only per module you could go CL12 I guess. Given you would want to spend the extra extra price on top.

Replacing a 6 year old intel CPU with an AMD Ryzen it yielded 15% more performance, but when the GFX was changed it got me a whopping 68% difference.

Note: the 1080Ti test was conducted with a DCS 2.1.0.6002 version (had sent off the older GFX before the switch), and that might explain why downtown was higher than Nellis (for reasons that escape me). SAince this version is more demanding I can only conclude that >70% performance difference would be a more realistic value.

are you able to make last test with 2500k also?

No. I no longer have the 2500K. And it would probably bottleneck the Ti so hard it wouldnt make a good comparison between the 2 cards. Neither would anyone go that route.

I have a 1080ti using a 2600k, and yeah, the 1080ti is bottlenecked hard in DCS...

with my GTX780, visibility and textures on high, everything else off or low, cockpit 512 and 2xmsaa i got 50-80 fps....

using the settings below i get between 70-140 fps in DCS with my 1080ti in 1.5

Thanks for this :)

I'm having hard time figuring out why I can't get proper FPS on 2.1 (significantly less the I get on 1.5.x).

after going over your findings, I starting to think that my i7-2600K is bottlenecking my 1080 seeing under 30FPS in 2.1.0 and 45+ in 1.5.6 (i'm running in VR)

anyone has suggestions? or should I start putting money aside for the 8700K?

Replacing a 6 year old intel CPU with an AMD Ryzen it yielded 15% more performance, but when the GFX was changed it got me a whopping 68% difference.

 

...

Hello,

1440p results ?

I have a 1080ti using a 2600k, and yeah, the 1080ti is bottlenecked hard in DCS...

 

with my GTX780, visibility and textures on high, everything else off or low, cockpit 512 and 2xmsaa i got 50-80 fps....

 

using the settings below i get between 70-140 fps in DCS with my 1080ti in 1.5

Bottlenecked how? You have almost double the FPS...

Hello,

 

1440p results ?

yes, both old and new machine for closest comparison.

Thanks for this :)

 

I'm having hard time figuring out why I can't get proper FPS on 2.1 (significantly less the I get on 1.5.x).

 

after going over your findings, I starting to think that my i7-2600K is bottlenecking my 1080 seeing under 30FPS in 2.1.0 and 45+ in 1.5.6 (i'm running in VR)

 

anyone has suggestions? or should I start putting money aside for the 8700K?

the 8th gen Intel CPU's have been postponed, from this summer to who-knows-when. Likely early next year.

If you dont want to wait you can probably get away with a R5 1600X build for now. Yes, you'll lose single thread performance (about 20%) but as I demoed the GFX is now more important, and that 1080 will be fantastic (it's screaming for a modern CPU). However if you can wait that 8700K 6 core will be faster overall. tough decision. :)

Bottlenecked how? You have almost double the FPS...

yeah but my gpu isn't maxing out, if i had a faster cpu, i'd get even higher fps.

Thanks for this :)

 

I'm having hard time figuring out why I can't get proper FPS on 2.1 (significantly less the I get on 1.5.x).

 

after going over your findings, I starting to think that my i7-2600K is bottlenecking my 1080 seeing under 30FPS in 2.1.0 and 45+ in 1.5.6 (i'm running in VR)

 

anyone has suggestions? or should I start putting money aside for the 8700K?

I've recently been there done that, first upgraded to a 1080 and didn't see the FPS move, upgraded to a 7700, and wow what a difference.

I went from:

i5 2500K @ 4.2GHz with GTX 770 2GB

To:

Ryzen 1700 @ 3.75GHz with RX 480 8GB

£800 upgrade for basically nothing. DCS is better optimised for Intel as I experienced virtually no increase in FPS. If anything it seems worse in the Caucasus map.

howie, the R7 should give you better performance.

Heres my testing:

Also check this article:

https://www.techspot.com/review/1450-core-i7-vs-ryzen-5-hexa-core/page9.html

-q9u1rZXO0E

So considering everything I exposed here, I believe you may be graphics bottlenecked. That 700 may not be enough anymore.

I remember a thread a while back expressing some serious frustration regarding DCS with AMD GPU's. Maybe things will change when AMD releases Vega sometime in the near future.

we are a couple of weeks away, literally, no joke.