Upgrade from 1070 to 2070 and performance lost.

[krisboats]

Hey guys, so i took the plunge and picked up an msi 2070 super gaming x trio to replace my zotac 1070 amp extreme. I've booted safe mode, run DDU, powered off, fitted new card and loaded up windows again. After installing the latest drivers i started benching and i'm seeing mixed results.

Heaven benchmark saw an increase from 188.1 fps and 4739 score to 250.6 fps and 6313 score.

Seemed pretty good.

Valley took a hit in performance from 133.6 fps 5591 score (max fps 221.4) and even after overclocking the new card slightly it only hit 137.5fps 5754 and max fps of 211.6

Rise of the tomb raider saw and increase from 54fps average to 72fps average at the same settings, apex legends saw a 18fps avg increase, 7 fps minimum frame loss and 3fps max fps gain.

Destiny 2 has gone down in performance from avg of 74fps in the tower to 68fps.

I've got a 4790k at 4.5ghz, 16gb ram and running 3440x1440.



Anyone got any ideas as atm i'm kinda bummed on spending so much to take a step backwards?

 

[krisboats]

CPU Bottlenecked. A Quadcore with 8 threads would have been bottlenecking your 1070 too

see all the research i'd done beforehand had prepared me for a few fps difference between my cpu and a newer one when upgrading the card, but nothing suggested i'd get less performance. That can't be the issue?

 

[krisboats]

Are you running with afterburner open? without looking at the clocks/temps/etc (incl. CPU) while running the games, it's hard to say. It might very well be the CPU getting ran ragged, esp. as nvidia uses extra cpu power very well in dx11 to boost their gpus. This used to be a big deal way back when and ppl didn't recommend AMD for ppl with weaker CPUs because they required more CPU grunt.

Simplest test you can do is to install Shadow of the tomb raider demo from steam and run the benchmark. That will show you very clearly, what kind of a CPU bottleneck you have, in their graphs.

brilliant, cheers. got it downloading now.

 

[krisboats]

According to intel XTU charts, while gaming in destiny 2 at the tower... it's running roughly 60-70% cpu utilisation.

 

[krisboats]

Thats my run through

 

[krisboats]

I notice ur not on latest 1909 build on win 10. Doubt it will help updating but there is a newer update if u wanted to try it.

I gave that a bash and it made no difference

I just dont get why its doing worse than the 1070. I went from gtx980 to gtx1070 with my 2500k and saw a good increase in everything.

I even tried taking the 2070 out and just sliding the 1070 back in, no driver changes or anything and average frames in destiny 2 went back up by 5-10fps. Slide the 2070 back in and it goes down again.

 

[krisboats]

I changed something, i dunno what. But was flying through some missons in destiny and pulling between 120-140fps at 3440x1440 and around 170fps at 1920x1080 windowed at high settings. Changing to low settings seemed to make around 20fps difference so not a gpu issue and gpu is now pulling 100% instead of barely scraping 60% usage. Still getting low fps in the tower but it's not everywhere now at least

 

[krisboats]

I don't think you would see a massive upgrade from a 1070 to a 2070. A 2070 is what 1080 level of performance?

Same pc, same benchmark, same settings. 1070 vs 2070 super. not sure why the minimum frames are lower though, couldn't tell they had happened during the benchmark.

 

[krisboats]

Urgh, i hate reinstalling windows

fresh install of windows. make sure you have latest BIOS, latest chipset drivers, etc, etc.



you wot m8?

he has went from a 1070 to equivalent of a 1080ti or faster.

so he has went up a considerable amount.

Noooo, i hate reinstalling windows

What is your PSU rating?

650w

Cpu pulls 150 ish, gpu while overclocked is 370 ish

 

[krisboats]

370W? My 2080Ti running at 2100/8000 draws less than 300W.

Just how OC’d is your 2070S!?!

I meant 270, my bad Its around 220 as standard.

Just FYI I don't know what cooling you are running but I got more out of my 2070S perf wise by under volting… might be worth playing with. Power draw usually stays below 90% but clocks are higher due to more stable boost. Stock but with power and thermal slider maxed it would drop down to mid-low 1900's even with fans at max during gaming while pulling around 112%... undervolted it will happily sit at 2010-2025 all day long.

While I don't think it would explain the problem here it may not be helping you get the most out of your card.

Running power limit slider on full, core voltage on +2 and core clock is +90 and memory is +700. It seems quite happy in ark at a constant 2055mhz at 60 degrees with 30% fan speed and boosts up to 2090 in other demanding games.

 

[krisboats]

Feel free to continue looking for a way to fix things online then. Doing a clean install when browsing on phone or iPad can be done in an hour or so. Much quicker imo.

Its more the hassle of re-installing programs, getting settings put back into place etc etc.

 

[krisboats]

Same pc, same benchmark, same settings. 1070 vs 2070 super. not sure why the minimum frames are lower though, couldn't tell they had happened during the benchmark.

Well, bit of a thread revival but new 5600x, ddr4 at 3600 and a fresh windows install later. Same 2070super



I think my 4790k and ddr3 was holding me back a bit in ROTR and it gets much better scores now. I doubt anyone cares but i'm happy now and thought the thread could do with a conclusion.

 

[krisboats]

My 4790k’s soldiering on thanks. Had a huge iq uplift in everything I play on 60hz screen moving from a 1070 to 3070 between 1440 and 4K

ACC
RDR2
MSFS
Cp2077
DCS

MSFS and CP are still mainly GPU limited.

I know things would improve with the 5900x I’m thinking about but in no urgent hurry

In high fps (tbf you probably wont notice on 60hz screen) games i was having stuttering, fps drops little micro pauses and the whole thing felt remarkably unsmooth. Going to the 5600x and all of that is gone. I just get way more stable frames now. Rendering clips for youtube is literally half the time now which is insane.

You can see if you compare my SOFTR benches it's a 10%+ boost to mins, max, avg, 95% across the board. Using my 120hz monitor, going from 90fps with drops and stuttering to 110-120fps and smooth feels like a night and day difference. Heck, with the 4790k ROTR had a min frame at one point of 6fps. the 5600x in the same section had a min of 41. thats bonkers.

 

[krisboats]

In high fps (tbf you probably wont notice on 60hz screen) games i was having stuttering, fps drops little micro pauses and the whole thing felt remarkably unsmooth. Going to the 5600x and all of that is gone. I just get way more stable frames now. Rendering clips for youtube is literally half the time now which is insane.

I would say it was that AMD gaming ram bottlenecking the intel system

Not sure if sarcasm or not but the tech specs were:

Capacity: 16 GB / 2 x 8 GB (Dual Channel)
Clock: 2400 MHz
Latency: CL 11-12-12-31
Voltage: 1.65 V

Tighter timings than corsair vengeance pro that runs at 11-13-13-31.

 

[krisboats]

Not a CPU bottleneck something else wrong - I went from a GTX1070 to a GTX3070 with a 4820K and saw a huge uplift in everything, then swapped to a 1650 V2 which has helped in games like Cyberpunk.

4 core / 8 thread still holds up in most games except a small smattering of newer ones.

You will see a big improvement, but you won't see AS BIG of an improvement as the card is truly capable of. So swapping from a gtx1070 to a 3070 you'd be getting 80% of the 3070's fps instead of 100%.

 

[krisboats]

Let’s put it like this....

you have a 4790k and a 1070.

someone offers you a full platform upgrade to say a 5900x or a 3070...which one would you take assuming you couldn’t sell it?

it’s a complete no brainer. The extra £800+ outlay of the new platform would be nearly double the GPU outlay for much less gaming benefit. Even if you just said 5600x it’s still £500+.

I just don’t see the essential nature of moving from a 4790k. Very desirable yes, necessary and worth the money...not really otherwise I would have done it by now.

Averages will go up sure, but lows wont improve much. 6fps lows vs 40fps lows on my very same gpu due to swapping the cpu out. Thats a 6.5x increase on the lows (1070 to 2070s) as well as 10-15% increase in averages compared to a less than 30% difference on the averages (going 2070s to 3070) as it wont be hitting the full 30% benchmarks sites say it will if i kept a 4790k. So i could keep the 4790k, get an estimated 20-25% avg boost and still have bad lows and stuttering, or i could keep the 2070s and remove the lows and stuttering and get a 10-15% boost to averages. Not to mention the overall system speed benefits, nvme support, halving the encoding times etc.

Thats the point i was at and with that choice, the gaming difference between a system upgrade (£300 cpu, £60 ram and £150 mobo) vs a 3070 (£600-650) is a much closer call than you're making it out to be.

 

[krisboats]

my 3070 was £469

Yeah but i can't buy one of those now so it's £600 to be comparable.

 

[krisboats]

What you're missing out here though is a 4790k will not be bottlenecking a 1070. So he could sink £500 into a CPU upgrade and he'd hit the same framerate limits.

Going from 1070 - 3070 should drastically improve the average framerate. You wont see the same improvement by pairing a 1070 with a 5900x

I'm not going from a 1070 to a 3070, i'd be going from a 2070s to a 3070.

 

[krisboats]

Decided to try running valley again. Left screen was the very very highest i was able to OC the 2070s to while using the 4790k. Took a good 10-15 runs to get there, starting the benchmark from a cold card etc. New score for the ryzen has it performing much more in line with results other people were getting.

 

[krisboats]

I doubt the OP will be bottlenecked by the 4790K @ 3440x1440. The gpu will be the bottleneck at such a resolution.

The [email protected] is faster than a 2400G and very close to a Ryzen 3600. I should know because I recently moved from a [email protected] to a Ryzen 3600.
The overclocked 4790K is still a pretty decent cpu so there is barely any difference between the two except when games use the extra cores. The 3600 may gain a bit more with a better cooler though.

Here's the benchmarks for Heaven on both my old and new systems.

Spoiler: Heaven results - Extreme preset 1600x900

[email protected] + Dark Rock Pro 3 Cooler


Ryzen 3600 Stock + stock cooler

OP is getting a much higher score of 250fps so I assume he is using a different setting. I know the 5700XT is close to a 2070S in normal games.

I benched at 1920x1080 without the 8xAA. Everything else was highest setting. I didnt realise i'd missed off the 8xAA with the old cpu so just used the same settings to make it a fair comparison. I stick to the 1920x1080 as ive got at least one bench from each iteration of upgrades so i can see how its improved along the way.

I do get higher fps in games with the new cpu, but its the smoothness in the low end that feels most noticeable. Ive been able to squeeze a bit more out of the gpu on the new setup, not sure if its the pcie4 thats helping or what. I'm honestly not too up to date with new OC'ing methods got some reading to do