3080ti or new CPU to overcome FPS issues

[danielwwm]

Hi,

 

Just after a bit of advice. I have a 2080ti with a 2600k clocked @ 4.5 Ghz running at 3440 x 1440 resolution. The main issue I have been facing is with A10C and F18C when I enable the TGP and especially the Mav cameras. FPS just tanks from 80s to low 50s and the stutter is noticeable, even with G-Sync. So much so if I am coming round for a second pass on ground targets in a A10c, i exit the Mav Cam back to TGP and go back to Mav cam when i am straight and level at the target.

 

My question is would a CPU upgrade to the upcoming 10XXX intel/4XXX Ryzen or 3080ti be better suited to overcome this problem. I know Vulkan will address this issue somewhat with multi-core support, but what do you guys think? Just making plans for my next upgrade.

 

I have already dropped the MFD to 512 res.

 

Thanks

[panton41]

My vote is the CPU. I'm assuming you mean the Core i7-2600K, which is close to a decade old and obviously your weakest part compared to the graphics card.

[derneuemann]

CPU +1

 

 

Now 9700k OC, with fast Ram

 

 

 

For Future, we will see (Vulkan)

[danielwwm]

Thanks guys. CPU it is. Hopefully the Ryzen 4xxx series will have a core clock on par with intel and the additional cores will take care of Vulkan when that comes online.

 

I’ve been trying to look for an ultra wide 2080ti with a relatively new CPU benchmark video with TGP or Mav on MFD to see what I should expect in terms of the FPS with a CPU upgrade but its surprisingly hard to find. Most out there are flying around with DSMS/HSI and non-TGP/Mav Mfds.I just wanted to see what the minimum FPS were.

 

To really kill the FPS in an A10C have the TGP in left MFD and MAV on right MFD.:(

[Wali763]

Thanks guys. CPU it is. Hopefully the Ryzen 4xxx series will have a core clock on par with intel and the additional cores will take care of Vulkan when that comes online. (

 

Mhz matter less than IPC. AMD has better IPC so 4,2Ghz of AMD (Zen 2) are equivalent to say 4,8Ghz of Intels latest gen (coffelake). A i7-2600 (for its poorer IPC) maybe needed to clock 6Ghz to reach same singlethread-performance.

 

I guess next gen AMD (Zen 3) will have even better IPC, so Intel could loose its 'fastest-gaming-CPU' - crown later this year.

[Hoirtel]

Mhz matter less than IPC. AMD has better IPC so 4,2Ghz of AMD (Zen 2) are equivalent to say 4,8Ghz of Intels latest gen (coffelake). A i7-2600 (for its poorer IPC) maybe needed to clock 6Ghz to reach same singlethread-performance.

 

I guess next gen AMD (Zen 3) will have even better IPC, so Intel could loose its 'fastest-gaming-CPU' - crown later this year.

 

Yeah gap between intel and ryzen is now minimal and may be nothing once all next gens are out. Trouble is there is so little hardware testing specifically for flight sims.

 

Worth noting though - since X-plane switched to Vulkan it still likes to use the CPU in the same way, its a graphics API so achieves much better performance through the way it utilises the GPU and not CPU cores. Check out some of Michael Brown's videos on xplane and its current vulkan beta, he discusses this better than I can.

[danielwwm]

I am looking forward to getting an AMD chip that would take the best gaming CPU mantle since the FX-51.

 

Thanks for the tip on Michael Brown’s video, Hoirtel. Vulkan made a huge difference for me on Xplane 11. Hopefully will see the same for DCS.

[Taintid]

Definitely a CPU bottleneck going on. I have a 2080ti, i7-7700K @ 4K and do not have this problem you speak of.

[danielwwm]

Definitely a CPU bottleneck going on. I have a 2080ti, i7-7700K @ 4K and do not have this problem you speak of.

 

Thanks Taintid

[derneuemann]

Mhz matter less than IPC. AMD has better IPC so 4,2Ghz of AMD (Zen 2) are equivalent to say 4,8Ghz of Intels latest gen (coffelake). A i7-2600 (for its poorer IPC) maybe needed to clock 6Ghz to reach same singlethread-performance.

 

I guess next gen AMD (Zen 3) will have even better IPC, so Intel could loose its 'fastest-gaming-CPU' - crown later this year.

 

 

IPC in what, some applications ok, but not in games.

 

There are nice tests that show that Zen 2 is slower in games than the old Skylake, coffeelake architecture, at the same clock speed!

[Dudikoff]

There are nice tests that show that Zen 2 is slower in games than the old Skylake, coffeelake architecture, at the same clock speed!

 

"Old" makes it sound like the later Intel CPU's are actually based on a different micro-architecture.

[dis80786]

Not so fast! I have an OC'd 2080Ti FE and a 5960x OC'd to 4.6GHZ and (in VR) I still return around 40FPS in VR with at least some visual acuity preserved.

 

Now, your choice depends somewhat on what you do in DCS. For example, if you fly the P-51D in the Blue Nosed Bastards of Bodney campaign, they have not yet created the CPU that can cope with calculating the wake vortices or bomb trajectories of a massed B-17 formation no matter what GPU you have it all slows down to 8 FPS. However, if you like to bimble around Dubai with no enemies at 4k you'll be GPU limited no matter what hardware you have.

 

Given you and I have the same GPU (more or less) cpu.userbenchmark states that my system is a mere 11% faster than yours, or that you are 29% slower than a 9900K (all at stock). Given that you are clearly good at overclocking, and that 29% will be far less

 

My point is simply that you should consider closely what you are about to splash out your hard-earned bucks on, and the role you set for that system: all that money for an improvement of a mere third (at stock) may not be worth it when you could just buy a new GPU.

 

For example, whilst you currently have an 4C/8T processor, it has been demonstrated elsewhere (Gamer's Nexus et al on YouTube) that 6 cores is the minimum for today's gaming demands, and modern games benefit from up to 8 cores (more than that is, apparently, overkill). However, DCS is not a modern game, and it may well be that a 2600K is sufficient for the workload DCS demands of it; after all, DCS utilises two parallel cores at most. In fact, if you turn of hyperthreading and allow those 4C to access their full quota of cache etc, it actually improves performance in games (and only in games). In applications like DCS, it is clock speed which matters above all else, and the high the clock speed the more data that one core can crunch. You could go out and buy a 16C/32T Ryzen 9 and actually lose performance in DCS vs your 2600K if the clock speed is considerably lower!

 

Enough. My advice is: do nothing. Wait until the 3080TI is released, tested to destruction, and we have hard, empirical benchmarks to go on. Also consider you RAM: OCing your memory so it has the lowest possible latency (freq/timings) also gives decent performance gains. Then, if a 3080Ti doesn't look like it'll do much to help you out, then decision is made: new system.

[panton41]

For the question of Intel versus Ryzen, I'll just mention what I read in a similar argument on a tech news forum - Intel might have marginally higher single-thread, but Ryzen is curb-stomping them on multithreaded performance. The thing is, games have been heavily multithreaded for years, and there's plenty of day-to-day productivity tasks benefit more from multithreaded performance than single-threaded.

 

FWIW, the article was on Ars Technica and about the newest MacBook Pro and the ARM-based A-series in the iPadPro was running rings around the Intel chip in the Mac. It's pretty sad how far they've fallen when a passively-cooled thermally and power constrained tablet chip can outperform the latest actively cooled 45-watt Intel laptop chip.

[dis80786]

For the question of Intel versus Ryzen, I'll just mention what I read in a similar argument on a tech news forum - Intel might have marginally higher single-thread, but Ryzen is curb-stomping them on multithreaded performance. The thing is, games have been heavily multithreaded for years, and there's plenty of day-to-day productivity tasks benefit more from multithreaded performance than single-threaded.

 

FWIW, the article was on Ars Technica and about the newest MacBook Pro and the ARM-based A-series in the iPadPro was running rings around the Intel chip in the Mac. It's pretty sad how far they've fallen when a passively-cooled thermally and power constrained tablet chip can outperform the latest actively cooled 45-watt Intel laptop chip.

 

Very true, but as I said above, DCS depends on single core freq and not number of cores. DCS just does not use all those cores, and so Intel still reigns here (and only here) with it's better single core perf. Most other games, yes; DCS no. Throwing cores at DCS just doesn't work.

 

Finally, Ars Technica shouldn't be surprised that software designed by Apple to run on the CPU's that power it's own products to be better optimised for those CPUs than vs everything else at given tasks. That's not apples with apples (hahaha).

[danielwwm]

Not so fast! I have an OC'd 2080Ti FE and a 5960x OC'd to 4.6GHZ and (in VR) I still return around 40FPS in VR with at least some visual acuity preserved.

 

Now, your choice depends somewhat on what you do in DCS. For example, if you fly the P-51D in the Blue Nosed Bastards of Bodney campaign, they have not yet created the CPU that can cope with calculating the wake vortices or bomb trajectories of a massed B-17 formation no matter what GPU you have it all slows down to 8 FPS. However, if you like to bimble around Dubai with no enemies at 4k you'll be GPU limited no matter what hardware you have.

 

Given you and I have the same GPU (more or less) cpu.userbenchmark states that my system is a mere 11% faster than yours, or that you are 29% slower than a 9900K (all at stock). Given that you are clearly good at overclocking, and that 29% will be far less

 

My point is simply that you should consider closely what you are about to splash out your hard-earned bucks on, and the role you set for that system: all that money for an improvement of a mere third (at stock) may not be worth it when you could just buy a new GPU.

 

For example, whilst you currently have an 4C/8T processor, it has been demonstrated elsewhere (Gamer's Nexus et al on YouTube) that 6 cores is the minimum for today's gaming demands, and modern games benefit from up to 8 cores (more than that is, apparently, overkill). However, DCS is not a modern game, and it may well be that a 2600K is sufficient for the workload DCS demands of it; after all, DCS utilises two parallel cores at most. In fact, if you turn of hyperthreading and allow those 4C to access their full quota of cache etc, it actually improves performance in games (and only in games). In applications like DCS, it is clock speed which matters above all else, and the high the clock speed the more data that one core can crunch. You could go out and buy a 16C/32T Ryzen 9 and actually lose performance in DCS vs your 2600K if the clock speed is considerably lower!

 

Enough. My advice is: do nothing. Wait until the 3080TI is released, tested to destruction, and we have hard, empirical benchmarks to go on. Also consider you RAM: OCing your memory so it has the lowest possible latency (freq/timings) also gives decent performance gains. Then, if a 3080Ti doesn't look like it'll do much to help you out, then decision is made: new system.

 

Thanks for the great advice Dis80786. I am not too fussed about the max fps but just making sure the min fps drops are reduced. From what I have found it is usually the TGP and MAV cameras on the A10c that slows it down. I also noticed when I played mission 1 of the Eastern Friendship campaign by Sedlo, when JTAC directs you for CAS on enemy units, my FPS really crawls down to mid 30s. I am unsure if its the units or the JTAC comms but something is slowing it down.

 

Well I have to wait for 4 to 5 months anyway for either the new Intel/Ryzen and/or the 3080ti.

 

The Ampere reveal at GTC will be on in a few hours! Can’t wait!

[David OC]

This is easy to see using MSI afterburner.

 

GPU workload (A-10C tgp and mavericks)

CPU workload (Comes down to a single core eventually) not overall. That would be 20% and one core at 100% etc. It's the one core that hits the limit to feed the GPU more.

 

That chip would limit the 2080ti by possibly up to 30% and only 20 perhaps for DCS. My chip limits my 1080ti by maybe 10% etc. Unless I ring the neck out of it at 5G's

It's not all ghz's. It' overall throughput of the system too.

 

Still a very impressive chip that has held up very very well for single core.

 

 

My goal for DCS is to hold 60fps (locked) max settings in all / any situations. This smooths out trackir also, which my system can at 1440p. So still very happy with it.

Example

Edited May 14, 2020 by David OC

[Dudikoff]

Still a very impressive chip that has held up very very well for single core.

 

Just to point out that it has held up even longer since it's pretty much exactly the same CPU architecture as the one before it (Skylake) and all the later ones, except that the latter increased the core count finally thanks to some competition from AMD.

 

So, if you upgraded it with a newer Intel, I guess your gains wouldn't really be noticeable in DCS since you're already running it at higher clock (which is the other way Intel gained slight performance increase over these same architecture generations).

Edited May 14, 2020 by Dudikoff

[derneuemann]

"Old" makes it sound like the later Intel CPU's are actually based on a different micro-architecture.

 

 

 

 

Of course, that is not what is meant.

Intel hasn't changed the architecture since Skylake. It is nevertheless the case that ZEN2 per core is not faster than Skylake and everything Intel has brought so far.

[Waxer]

I have quite a lot of experience of building my own systems. Here are a few comments:

 

1) I feel that the guy arguing that IPC gains mean a 4.2GHz AMD CPU will outperform a 4.8GHz Intel CPU on single threaded performance due to IPC gains is flat out wrong. IPC gains do mean that later generation CPUs are generally faster than earlier CPUs at the same clock speed, but not to the extent that it would make up 600MHz of clock speed.

 

2) AMD CPUs are already efficiently overclocked by AMD with their clever software, meaning there is very little extra performance to be bought by end user overclocking. You can do it, but gains are small and you pay the price in noise, power consumption and heat fatigue on your system. (Transistors degrade if run at high temperatures).

 

In contrast, historically Intel CPUs can usually be overclocked effectively with big performance gains if you have a capable motherboard, power supply and cooling solution. For example my CPU should turbo at 4.0GHz, but I have it running 24/7 at 4.8GHz while at low operating temperatures.

 

3) If your focus is - as you say - reducing low 1% framerates even more than boosting average fps (which is a very sensible goal btw), then the following applies:

 

a) Make sure your system cooling is keeping temperatures under control. If you have temperature spikes this could cause CPU throttling. Keep it cool and allow your CPU to keep running flat out. How? Lots of air flow in and out of the case. Personally I have an overkill dual D5 pump custom water-cooling rig. That works, but it is expensive. (Great fun though!)

 

b) Make sure you never hit RAM capacity. Chart RAM usage. If you are maxing out 16GB, go for 32GB.

 

c) Get fast RAM and don't ignore latency. You need to balance the two to get a well optimised system. Currently DDR4 3600 / CL16 is decent. But look at secondary timings too so look for 16-16-16-36... so manufacturers sell 16-18-18-38 as CL16 memory... LOL jokers. You can go more crazy than that but it is diminishing returns.

 

4) Because a system has background tasks you want a multicore system for sure. But there is little point adding more and more cores that sit idle, or even worse generate heat and make core density higher making things more difficult to cool. CPU manufacturers LOVE to sell you more cores because they find it hard to increase clock speeds from where they are now and they want you to spend money. If I were building a new system now I would shoot for 8 cores as that is what the next generation of gaming boxes is going to have so software will be optimised for that many cores. That is assuming your main use for the computer is gaming. If productivity tasks requiring lots of cores then Threadrip away and get as many cores as you can afford. But that is a different use case.

[danielwwm]

Thanks for the great tips. I will run MSI afterburner with TGP and MAV on.

 

Your comments also made me realise my RAM is also a choke point. Since I am on Sandy bridge, my RAM is only DDR3.:cry:

 

Well not long now as Pre-orders for Comet Lake has already popped here in Melbourne and shipping May the 21st. Now the agonising decision whether to wait for the Ryzen, which is rumoured to be October or move ahead.

[Waxer]

Yes DDR3 bottleneck. I am on DDR4, so probably I will wait to build another system until I can get DDR5. I am looking at Comet Lake, but unlikely to buy unless I have a problem with my current rig before then. I might well get a 3080 Ti however, but I will be a late mover as I want an aftermarket GPU with a high bios power cap.