Bottleneck calculators / 'can you run it'

[wazzickle]

https://pc-builds.com/cyri/

My current setup is an i5 10500 paired with a 1070 (ghost s1, 16gb ram, h470m)

I'm thinking about upgrading to a 3070, should a reasonably-priced one become available.

The above site is giving the following results (1440p)

1070:



3070:



The first seems somewhat realistic; my ram isn't the fastest, but with resolution downscaled by about 20-30%, I'm getting frame rates above 100fps with mostly medium but some high settings.

It's also suggesting that the gpu is bottlenecking my current setup (not specified on a per game basis - I'm told that warzone is relatively cpu-based) by about 5%, and that if I got a 3070, the cpu now would be bottlenecking my system by less than 1% - so a good buy. But on a 3070, I'm still apparently gpu-limited when I go above medium settings.

Does anyone know how these tools work? Are there good alternatives out there?

[Output]

The one that immediately comes to mind, is the Can You Run It? from System Requirements Lab, which involves a download of the detection tool I believe (it has been around for years, but I haven't used it in a long time).

As for how it works, I would assume it just bases it on the minimum and recommended system requirements for each game, and simply compares those to your detected hardware specs.

I don't think it guarantees something will or won't run, just that it gives an indication of the likelihood based purely on the given requirements.

EDIT: Or at least, that's the way I believe the one I mentioned works. Yours could very well be calculating things very differently based on actual measurements involving various CPUs and GPUs for each game.

[Jonj1611]

Looks like they are just comparing things on specs as there is nothing to download, it doesn't detect anything, just asks you what you processor and gpu is and even then I would take the results with a pinch of salt.

It says on low settings and 4k I should be getting around 75fps in Cyberpunk. I really really wish that was the case!

Then goes on to say : AMD Ryzen 7 3800X will need at least 64GB of RAM to work well.

Seems a bit high?

[kalniel]

I can't speak for the accuracy of that particular site, but, for the particular benchmark they're running, it could well be right - most 3D games are bottlenecked by GPU at 1440p, which is why you see an increase in performance when you upgrade.

But the best alternative is to check GPU review sites for benchmarks.

[Scryder]

Yeah, that site seems to be just comparing ideal settings. And as @Output said, is similar to Can You Run It... a site I used to go to over 10 years ago, as my laptop wasn't up to spec at the time.

I've seen a site called https://www.userbenchmark.com/Software where you can compare your specs, I think I used it once when I first got this desktop setup completed. Seems a bit more advanced as to what they do.

They also have a build comparison section.

[Xuse]

I wish I had found a similar calculator before I went and bought a 3070 putting me in to a tricky situation financially as I now scramble to build a PC to run it...
Also don't know how accurate the site is, but looking at some test videos on youtube as much as 50% FPS loss...

[Xlucine]

Easy fix, just get a higher res monitor!

[kalniel]

Easy fix, just get a higher res monitor!

Or run DSR on your existing monitor

[ik9000]

Yeah, that site seems to be just comparing ideal settings. And as @Output said, is similar to Can You Run It... a site I used to go to over 10 years ago, as my laptop wasn't up to spec at the time.

I've seen a site called https://www.userbenchmark.com/Software where you can compare your specs, I think I used it once when I first got this desktop setup completed. Seems a bit more advanced as to what they do.

They also have a build comparison section.

IIRC take that site with a very generous pinch of salt. I think it's the place where people complain that they seem to ditch results that don't fit their agenda. So you could have a pimped AMD machine and they will lose its benchmarks to keep the averages where they want them. Or a lame intel machine and they'll lose those. And intel chips will seem to be recommended vs AMD ones, on spurious weightings etc etc. even where independent testing shows the AMD chip to be far better performing across the board.

[ik9000]

IIRC take that site with a very generous pinch of salt. I think it's the place where people complain that they seem to ditch results that don't fit their agenda. So you could have a pimped AMD machine and they will lose its benchmarks to keep the averages where they want them. Or a lame intel machine and they'll lose those. And intel chips will seem to be recommended vs AMD ones, on spurious weightings etc etc. even where independent testing shows the AMD chip to be far better performing across the board.

Here's an example. intel i7-11700k vs AMD ryzen 7 5800x. https://cpu.userbenchmark.com/Compar...00X/4107vs4085 Supposedly two similarly priced chips according to userbenchmark's priceguide.

Tom's hardware calls it "the chip of last resort" yet userbenchmark says it's the better chip. It only compares thermal performance in terms of TDP, and not actual testing where THw logged a whopping 80-140% extra - no typos - eighty to one hundred and forty percent - increase in power draw for the intel chip across tasks.

It's no secret that Intel has dialed up the power with Rocket Lake to compete with AMD's vastly more efficient chips, so you'll have to ignore the higher power consumption if you choose to go with an 11th-gen Intel chip. As such, there are no real surprises here — the Core i7-11700K draws more power in every measurement than the Ryzen 5000 lineup, and also more power than its 10th-gen predecessor, the Core i7-10700K.

check out this chart - power is wayyyy more

and this chart


AMD beasts it out the park in terms of thermals, heat and power usage. Here's the visual metric - AMD has approx the same performance for way lower power and energy usage. It is simply a better designed chip. Your machine will run real-world at the same speed, but cooler, quieter, and with the delight of knowing you didn't give greedy intel any of your hard earned cash for a substandard band-aid product that they never wanted to give you in the first place.



Ah but gaming you say, gaming userbenchmark says the intel is faster in gaming... 5%-8% faster.

not when you don't cherry pick your titles:

here's the aggregated data from all the titles THw test:

at 1080p, ryzen is 2.4% faster:


at 1440p ryzen is 2fps or 1.2% slower


what userbenchmark are reporting is more the 99th percentile fps values, where sure, you might get 6% over (THw for 1440p) but gameplay cares about min fps and average fps far more than peak fps value.

as THw put it:

Flipping over to 1440p reveals more slim deltas, with the Ryzen 5 5600X leading at stock by 2%, while the 11700K takes a ~1% lead after overclocking. The Core i7-11700K does have noticeably better 99th percentile measurements at 1440p, with a 3.4% lead at stock and 7% lead after overclocking. We didn't notice any outwardly visible signs of noticeably smoother gameplay during our tests with the Core i7-11700K, but those are appreciable 99th percentile deltas. That said, keep the measurements from the overclocked config in perspective — this could vary.

And here's the real rub - the cheaper 5600x does just as well in gaming in most cases. So what userbenchmark doesn't tell you, in any way, is get that one instead and save money. Stick those two into their benchmark and it tells you the 5600x is 20% worse, and in their ranking 2nd/1316 for intel vs 32nd for the 5600x AMD. Hence people suggest it might be akin to garbage if you consider the real world and proper testing.

Given the lack of rigour elsewhere on their site I wouldn't rely on their bottlenecking guides either. Is there a more reliable alternative anywhere people know of?

[kalniel]

Given the lack of rigour elsewhere on their site I wouldn't rely on their bottlenecking guides either. Is there a more reliable alternative anywhere people know of?

The reliable alternative is looking up GPU and CPU reviews so you can find the actual bottlenecking for your games of interest. If reviews haven't been done then if you have access to the hardware you can test yourself to see what the bottleneck in your system is, but it's not so useful if you're comparing theoretical future purchases.

[Xuse]

I used to have a 4k TV, but since it died within the warranty period I took my money back and haven't bothered since. Anyway, the point being I'm limited to 1080 right now anyway.