My own 3770K vs. R7 1700 games comparisons

[steve258]

*** 14/03/2017 Added SMT off results***

Made the jump from 3770k to 1700 couple days ago. So far the gaming results are better than I expected, average framerate is generally just a few fps difference from my 3770k at 4.7Ghz. This is with my 1700 running at 3.9ghz ram at 2933 cl15. So basically the gaming performance remains the same but I also now have a huge multithreaded advantage.

R7 1700 3.9ghz 16Gb DDR4 2933 CL15
Gigabyte AB350 Gaming 3
GTX 1070 at 1947 core 4000 mem

3770K was running at 4.7Ghz with 16Gb DDR3 at 1866mhz CL10, Same GTX1070 of course.

Here are the game results that I tested, all at max possible settings at 1080p and 2x smaa/2x TXAA where applicable. I also included stock 1700 stock results for the games that I've benched. Hope this helps:

Heaven Benchmark:

3770K 4.7: min. 9.2 avg. 122.8 max: 270
1700 3.85 (forgot to rerun at 3.9): min. 32.6 avg. 121.7 max: 257.4

R7 1700 3.0 2133 cl15: avg. 115.8

Rise of the Tomb Raider:

3770K 4.7:
Mountain Peak 100.96 (min. 48.09 max 168.12)
Syria 68.82 (min. 26.42 max 99.84)
Geothermal 74.91 (min. 50.90 max 108.61)
Overall 82.32

1700 3.9:
Mountain Peak 101.32 (min. 49.11 max 172.23)
Syria 70.50 (min. 25.42 max 92)
Geothermal 71.32 (min. 41.37 max 99.17)
Overall 81.67

1700 3.9 SMT off:
Mountain Peak 102.67 (min. 38.07 max 144.98)
Syria 69.71 (min. 27.64 max 86.47)
Geothermal 72.05 (min. 35.61 max 93.35)
Overall 82.30

R7 1700 3.0 2133 cl15: Overall 69.49

Far Cry Primal:

3770K 4.7: min. 63 avg. 74 max: 81
1700 3.9 : min. 57 avg. 72 max: 83

1700 3.9 SMT off: min. 60 avg. 73 max: 84

R7 1700 3.0 2133 cl15: avg. 63

Rainbow Six Siege:

3770K 4.7: Overall min. 62.8 avg. 90.4 max: 139.9
1700 3.9 : Overall min. 63.6 avg. 91.3 max: 131.3

1700 3.9 SMT off: Overall min. 65.6 avg. 91.2 max: 128.1

Dirt Rally:

3770K 4.7: min. 71.98 avg. 94.66 max: 127.31
1700 3.9 : min. 72 avg. 93.89 max: 125.52

1700 3.9 SMT off: min. 71.94 avg. 93.39 max: 123.50

R7 1700 3.0 2133 cl15: avg. 92.81



As you can see from the above games, the average frames in these situations are within negligible level. So the conclusion is same as others have said really, that if you mostly play older titles that favours 4 cores or less, especially when you are running 144hz or higher refresh rate where the max fps really does count. Then the older gen i7 still have the raw power to really drive the fps. if you play newer titles at higher resolution when GPU matters more then you'll find the rysen, overclocked at least, will not dissapoint. But it has to be overclocked to really shine.

Sorry no 640x480 benches

***14/03/2017*** With a relatively modest GTX1070 it seems the SMT off performance increase in games is minimal at best, not really worth turning off SMT for. Bar the most powerful GPUs in SLI/Xfire configuration I doubt in most scenarios you will notice any difference.

 

[steve258]

Have you been able to test anything to show that huge multithreading advantage you mention....?

Hi

Nothing outside of synthetic benches unfortunately, realbench shows huge improvement in heavy multitasking and encoding scores:

3770K

R7 1700:

Nothing else I use so far that really stresses all cores, apart from the odd handbrake works. Although I had a quick go at watch dogs 2 yesterday and I was surprised to find that CPU usage hits 90%+ across all 16 threads

3770K is still very capable no doubt but rysen has been the most exciting thing in a long while and got me itching again I was expecting gaming performance to be worse tbh from all the reviews so I was pleasantly surprised that it keeps up this well.

 

[steve258]

Holding its own against a heavily overclocked 3770k is somewhat comforting but also not stellar considering the generational differences and, perhaps, my expectations. Then again I felt that regarding the incremental changes from my 3770k to the 6700k

True, True. But if you think about it, 4.7 vs 3.9, the IPC for these zen chips are pretty darn impressive, the only problem they have is that they don't clock very well. But that will improve with next iterations. I won't be surprised to see zen+ with 10% increase in IPC and oc to 4.4/4.5Ghz quite comfortably All speculations of course.

Yes I know it's only GTX1070 and if you slap in a 1080TI/pascal TitanXXX SLI then the gap will widen. But in these particular situations, with graphics card that a lot more of us will likely own, and for those of us who are still hanging onto SB/IB rigs, it's a genuinly viable upgrade. And the gaming performance certainly isn't all doom and gloom like you see in the reviews.

 

[steve258]

I have the same board as you. . if its not too much to ask, could you explain the voltage options on the board? it seems you have to manually enter how much you want/allow the board to go over the base 1.2v?

Yes that's the gist of it. The offset value is the amount in addition to your CPU default vcore, which varies between CPUs. So an offset value of +0.130 on a default 1.2v CPU will bring the vcore upto 1.330v.

 

[steve258]

Awesome thanks, before i have a play around with it, is there an option to disable the boost, like you had with the old FX processors? or if i change the multiplier to read 3.6ghz will it just sit at that frequency?

Yes it's possible, although the whole point of offset voltage is so that the additional voltage is only applied when CPU is under boost. On this board though the offset voltage acts pretty much like fixed voltage. It's always good to disable power saving and core boost features when ocing so the clock stays the same at all time. Just go in to advanced CPU menu and disable cool and quiet and C6 features.

Interesting results but you missed 1 thing. You forget to take notes of temps!

I want you to included temps so we can see how both 3770K and Ryzen 7 1700 doing with temps at stock and OC.

What cooler do you use?

I could think of try OC my 3770K first time soon after read all delid tools reviews then I can have first experience with 3770K then months later maybe I could do same with 7700K to get lower temps if I find 7700K stock die paste temp too high.

I benchmarked Rise of Tomb Raider with stock 3770K, 16GB DDR3 and GTX 1070 and got overall score of 71.21 fps.

Sorry it was meant to be a quick test before swaping so didn't properly measure temp and such. My 3770K ran HOT so had to delid mine (with a good ol razor blade, no such luxury as delid tool back then ) and even then I had to get rid of the IHS all together. With a little bit of ghetto mod I managed to do a direct core contact with my Megahalems cooler and that kept the temp in check at 4.7ghz, topping out 80-85c 1.32v during prime stress test.

My 1700 is still on stock cooler so not really a fair comparison of temp. Prolimatech customer support is absolutely useless. Not a single word from them regarding the AM4 conversion kit despite multiple attempts to contact

 

[steve258]

I can't tell if those results are gpu bound or they are genuinely near identical performance.

Assuming the latter that puts gaming IPC about 20% better than Ivybridge. Problem is that games don't scale very well in terms of CPU clockspeed or real ipc gains.

Was the 1700 tested with SMT off? Windows currently treats logical cores as physical cores.

Most likely GPU limited more than other. Like I said earlier I realise that with more powerful GPU the gap will widen surely. These tests are done with SMT on

OP can you compare some of the cpu loads for us? That would be really helpful.

Sorry the 3770K system has been dismantled

 

[steve258]

Updated OP with SMT off results for all games

 

[steve258]

When the next generation of gfx cards arrive maybe they will shift things back to cpu limitations and then ryzen will look worse? Too much hype with ryzen. No wonder they left the multipliers unlocked the cpus dont even clock that high.
Zen2 will be interesting though.

All true of course, but only if you use next gen gpu to play this gen games . Only real reason why most people upgrade gfx is for higher resolution monitors or because new games demand it. So for the most part gpu will still be the bigger limiting factor.