3700X vs 9900K, that is the question...

[cucker tarlson]

It is. Other tests show consistent deltas between two sets of two processors. Yours show consistent deltas between two processors, but wildly different deltas for two other processors, of the same set of four. This is an anomaly.

thinking that just because a 720p test shows XX% difference between two,three or five cpus means any other run that shows a wider/lesser gap is incorrect is just wrong.

 

[Papahyooie]

thinking that just because a 720p test shows XX% difference between two,three or five cpus means any other run that shows a wider/lesser gap is incorrect is just wrong.

That's not how deltas work lol. When the gap is consistent across two testing methodologies, except for a specific result, that is anomaly.

 

[biffzinker]

What’s this about SSD performance?

Seems to be fine for me.

 

[btarunr]

If you don't overclock, have you considered a combination of i9-9900 (non-K) with a cheap B365 motherboard?

https://www.newegg.com/intel-core-i9-9th-gen-core-i9-9900/p/N82E16819118023 ($440 here)

https://www.newegg.com/p/N82E168131...B365&cm_re=ASRock_B365-_-13-157-863-_-Product ($90 here).

 

[dgianstefani]

What a silly suggestion. The point of Intel is to overclock. If you're not going to overclock buy a Ryzen, they're about equivalent to Intel at stock, but the Intels gain 10/20% by overclocking whereas the Ryzens are already running maxed out.

 

[Vayra86]

you're confusing the very thing you're arguing about.
your point was about replicablility factor,the consistency in carrying out the tests.not the location vs resolution difference that I'm trying to get across.just cause your data with 720p could be more consistent does not mean it's any more relevant than location testing.I mean it could be for a reviewer maybe,but not for the end user of a cpu.

Please don't think I don't get your point.I do,though I think you exagerrate things.I just think that more consistent methodology (720p benchmark) does not matter for me as much as testing a cpu heavy location at more standard resolution would.You end up with a consitent bag of nothing.Which result will the end user ever be closer to seeing with their own eyes while playing?

You're both right here, to be honest... the idea behind the test TPU does and the one you see elsewhere focusing on specific game events/locations is different.

- TPU wants to show the maximum potential FPS in game X on CPU Y, and how each processor ranks in that maxed out scenario - uncolored, 'scientific' value. A window into the maxed potential with regards to high FPS gaming. Keep in mind those CS GO players used to run 4:3 at low res and super low settings... They like this test and it serves a purpose in their buying decision - and rightly so, I might add. It is also here you see Ryzen's improvement over last gen in all of its glory.

- Specific game location tests serve a night and day different purpose; they show us the worst case scenario for a CPU. Does it run maxed out? Yes, but it also runs into a badly optimized, or heavy in asset/code part of the game. This gives us a window into the worst frame drops per game per CPU. It also tells you mostly just thát: how does it work out in that specific game. This is interesting for people who play mostly that specific type of games. TW3's Novigrad or AC: O's towns, same type of game & same behavior on CPUs.

Different purpose, and I think both approaches are valuable to readers. I also have to say the 720p testing fits well in the rest of the review for TPU because there is zero focus on minimum FPS. Its all about averages, ballpark idea of relative performance. Not a deep dive.

 

[cucker tarlson]

What a silly suggestion. The point of Intel is to overclock. If you're not going to overclock buy a Ryzen, they're about equivalent to Intel at stock, but the Intels gain 10/20% by overclocking whereas the Ryzens are already running maxed out.

stock vs. stock the K skus look meh compared to ryzen on price terms,non-k will be cheaper for the cpu and cheaper for the mobo while the performance hit is not that big.
it's just another option,and tbh it's not silly at all.

You're both right here, to be honest... the idea behind the test TPU does and the one you see elsewhere focusing on specific game events/locations is different.

- TPU wants to show the maximum potential FPS in game X on CPU Y, and how each processor ranks in that maxed out scenario - uncolored, 'scientific' value. A window into the maxed potential with regards to high FPS gaming. Keep in mind those CS GO players used to run 4:3 at low res and super low settings... They like this test and it serves a purpose in their buying decision - and rightly so, I might add. It is also here you see Ryzen's improvement over last gen in all of its glory.

- Specific game location tests serve a night and day different purpose; they show us the worst case scenario for a CPU. Does it run maxed out? Yes, but it also runs into a badly optimized, or heavy in asset/code part of the game. This gives us a window into the worst frame drops per game per CPU. It also tells you mostly just thát: how does it work out in that specific game. This is interesting for people who play mostly that specific type of games. TW3's Novigrad or AC: O's towns, same type of game & same behavior on CPUs.

Different purpose, and I think both approaches are valuable to readers. I also have to say the 720p testing fits well in the rest of the review for TPU because there is zero focus on minimum FPS. Its all about averages, ballpark idea of relative performance. Not a deep dive.

well in game tests are certainly not wrong or flawed
like I said,720p built-in testing is very consistent but max cpu performance is not the limiting factor,the min is.
like that i5 result,sb sees that and thinks his i5 is ready to handle odyssey at 90 fps,buys a 120hz panel and then sees the cpu limiting performance to 60 fps in every city.

 

[Vayra86]

stock vs. stock the K skus look meh compared to ryzen on price terms,non-k will be cheaper for the cpu and cheaper for the mobo while the performance hit is not that big.
it's just another option,and tbh it's not silly at all.

well in game tests are certainly not wrong or flawed
like I said,720p built-in testing is very consistent but max cpu performance is not the limiting factor,the min is.
like that i5 result,sb sees that and thinks his i5 is ready to handle odyssey at 90 fps,buys a 120hz panel and then sees the cpu limiting performance to 60 fps in every city.

Of course, but the blame should be placed not on the CPU but the game. All CPUs drop heavily in these scenarios, and the gist of that is just that you want the fastest CPU available, there are barely any exceptions to that. And that conclusion also echoes from the 720p tests.

 

[cucker tarlson]

Of course, but the blame should be placed not on the CPU but the game. All CPUs drop heavily in these scenarios, and the gist of that is just that you want the fastest CPU available, there are barely any exceptions to that. And that conclusion also echoes from the 720p tests.

well there are big differences in how much those cpus drop,the whole point of testing cpus for gaming.balanced performance is not high avg. fps,it's when the basement is as close to the ceiling as possible.

 

[Vayra86]

well there are big differences in how much those cpus drop,the whole point of testing cpus for gaming.

....drop in that specific scenario. A different game may put other CPUs on top in its own heaviest scenarios. Those tests have little value beyond a specific type of game, type of engine and/or even only just that specific game. They say little about the 'overall' relative CPU performance. 720p does give you that.

Go figure, the two examples you keep giving are two very similar scenarios. Different engines, but 3rd person open world and both in a population hub. That is precisely the scope of the results you are looking at.

 

[cucker tarlson]

....drop in that specific scenario. A different game may put other CPUs on top in its own heaviest scenarios. Those tests have little value beyond a specific type of game, type of engine and/or even only just that specific game. They say little about the 'overall' relative CPU performance. 720p does give you that.

that's why you test across games,engines and apis.
look at gamersnexus,they do just a bunch of games,but each is different engine,type,api.
if a cpu does well in 5 wildly different scenarios,chances are it's gonna do well in every other one.
that's why I always say that the best cpu is not the one with highest avg. fps,but the one that always makes it to the top 3 of any chart.if you get an 8700k you can be sure you're never gonna see your performance fall off a cliff in any game.

 

[Vayra86]

that's why you test across games,engines and apis.
look at gamersnexus,they do just a bunch of games,but each is different engine,type,api.
if a cpu does well in 5 wildly different scenarios,chances are it's gonna do well in every other one.
that's why I always say that the best cpu is not the one with highest avg. fps,but the one that always makes it to the top 3 of any chart.

I think the relative performance charts on TPU still cover that quite well. Or can you put two reviews side by side and point out a different top 3 based on that between GN and TPU?

A large test selection kinda mitigates the outliers and arrives at similar results as to the 'overall best'.

 

[cucker tarlson]

I think the relative performance charts on TPU still cover that quite well.

in terms of cpu rank list - yes
in terms of measuring actual cpu bottleneck performance deltas - not quite

fact is I don't care much for 720p testing,but that's cause I have more personal experience with several dozens of games ran on a high refresh rate panels,that's why I look for cpu heavy location testing.
both are indicative of something that someone might be looking for.

 

[Vayra86]

in terms of measuring actual cpu bottleneck performance deltas - not quite

Alright I get ya, but that really is equivalent to a deep dive, one you would expect to be looking for AFTER buying your CPU. A lot of these game specific quirks can be fixed either through software tweaks, or hardware OC. And in other examples, a simple game patch makes all problems go away... so again... I highly question the value of going that deep into it, when it comes to a purchase decision. For that, overall relative performance is where its at, unless you are very keen on serving just a single type of workload.

 

[cucker tarlson]

Alright I get ya, but that really is equivalent to a deep dive, one you would expect to be looking for AFTER buying your CPU. A lot of these game specific quirks can be fixed either through software tweaks, or hardware OC.

ryzen 3000 would not like what you just said

 

[Vayra86]

ryzen 3000 would not like what you just said

Amen to that, its the reason Intel had dominance for so long, right? The CPUs were versatile. They excelled both at high FPS and at high load scenarios. Now, the tables are turning. Good thing Ryzen 3000 kills stuff right out of the box.

 

[cucker tarlson]

Amen to that, its the reason Intel had dominance for so long, right? The CPUs were versatile. They excelled both at high FPS and at high load scenarios. Now, the tables are turning. Good thing Ryzen 3000 kills stuff right out of the box.

well,another thing is min fps is just part of the story.frametime consistency is just as important.I'd like to see how that humongous l3 cache on ryzen 3000 handles that.

 

[ratirt]

Amen to that, its the reason Intel had dominance for so long, right? The CPUs were versatile. They excelled both at high FPS and at high load scenarios. Now, the tables are turning. Good thing Ryzen 3000 kills stuff right out of the box.

Intel has slight advantage now because its long dominance which shrinks every time AMD releases a new CPU.

well,another thing is min fps is just part of the story.frametime consistency is just as important.I'd like to see how that humongous l3 cache on ryzen 3000 handles that.

I agree. Min FPS is the key and benching 720p is ridiculous from one stand point. Who cares about 120 max when lows go below 40 for example. The narrow the gap between min and max then better. I see reviews when benching at 720p the FPS isn't much different from 1080p or even 2k and reviewers tend to say "CPU bottleneck". It is kinda horse crap due to the fact the CPU is being utilized with 5-10%. How can CPU be a bottleneck when it still has such a headroom. The game is the bottleneck, poor scheduler (maybe), execution of code and process cause it can't utilize the resources. The game software gives/feeds commands to CPU and tells him how to make it happen not all the way around. For me a CPU bottleneck is when CPU is at 100% struggles to keep up and the GPU has to wait for it to finish whatever is there to finish hanging on at 60% for instance.
That's a definite CPU bottleneck. I know this will change and it has already started but it is happening slower than CPU arch and core number for some reason.