AMD "Ryzen is Smoother" Misconception Benchmark & Explanation (Ryzen vs. Intel for Gaming)

[humbug]

I edited my answer - Intel doesn't believe in reviewers doing benchmarks,so what Steve did is now irrelevant to Intel in 2020.

That's what gets me about it, he's taken something that was, 'to some extent' true in 2017, and no longer is in 2020 and then applied it as if it was a falsehood being perpetuated for 2020, all so he could make a clickbait video putting his big brain angry hat on and put the internet to rights.

The backlash he got for this was deserved.

 

[Deleted User 456458]

@MartinPrince if your 3900x is IF and mem tuned, mind running Geekbench 4. Trying to see SC gap currently when both systems are tuned.

 

[KungFuSpaghetti]

That's what gets me about it, he's taken something that was, 'to some extent' true in 2017, and no longer is in 2020 and then applied it as if it was a falsehood being perpetuated for 2020, all so he could make a clickbait video putting his big brain angry hat on and put the internet to rights.

The backlash he got for this was deserved.

The backlash should solely be aimed a how verbose and monotonous his videos are... seriously, the guy needs to skip to the end. I like tech, but surely no one needs to be subjected to this.

I honestly only tune into his vids to see if he's bothered to get a haircut...

 

[humbug]

The backlash should solely be aimed a how verbose and monotonous his videos are... seriously, the guy needs to skip to the end. I like tech, but surely no one needs to be subjected to this.

I honestly only tune into his vids to see if he's bothered to get a haircut...

I don't know what his game is but he does have a massive chip on his shoulder, there are a lot of very happy Ryzen users out there, given the massive swing from Intel to Ryzen many of these people are former Intel owners, that includes me, its basically a huge army, the sort of army that Intel used to enjoy.

When he took a 2017 truth and twisted it into a 2020 myth so that he could make this video he insulted the intelligence of an army, he then rubbed salt into that wound by trying to discredit his now critics by calling them an AMD fanboy problem, again many, probably most of them are Ex-Intel.

Its an age old problem of someone who is quite clever, but nothing like as clever as he thinks he is.

 

[MartinPrince]

@MartinPrince if your 3900x is IF and mem tuned, mind running Geekbench 4. Trying to see SC gap currently when both systems are tuned.

Yes, the IF and MEM are as tuned as far as I can take them at 1866/3733Mhz and also C14 and 1T etc on the memory. 3800Mhz is just out of reach and not quite fully stable. I'll post up a Geekbench 4 later, anything else?

 

[KungFuSpaghetti]

I don't know what his game is but he does have a massive chip on his shoulder, there are a lot of very happy Ryzen users out there, given the massive swing from Intel to Ryzen many of these people are former Intel owners, that includes me, its basically a huge army, the sort of army that Intel used to enjoy.

When he took a 2017 truth and twisted it into a 2020 myth so that he could make this video he insulted the intelligence of an army, he then rubbed salt into that wound by trying to discredit his now critics by calling them an AMD fanboy problem, again many, probably most of them are Ex-Intel.

Its an age old problem of someone who is quite clever, but nothing like as clever as he thinks he is.

Yeah I find it all odd that anyone would be so brand biased, but then I've only been back into PCs in the last few years before the internet's was really a thing.
Ive always been brand agnostic and generally just went for whatever I could afford that suited my needs. Prior to 2700x and 3900x I've had mainly all Intel, bar a phenom CPU that I used for a media server as it was cheap.
I've never had a motherboard cater for 4 generations of CPU though, which was the reason for ryzen this time out. Played on a friend's 1800x when they released and was blown away by how good it was for the price so that was me sold.

Tech jeebus might be academically smart in his given field, but generally seems a bit of a plonker, but I'm probably old and out of touch and see people like him in my line of work all the time. Usually they are career limited, as it takes a smart man to admit he's stupid. Fortunately for him he's found a niche market for his dull presentations.

 

[ik9000]

what sound and music software you running? Certainly for ripping and batch converting I find it's thread dependent in my software and can use as many threads as you throw at it, so my 3900 wins. For sound recording and post processing though, not done enough tests.

 

[Deleted User 456458]

Yes, the IF and MEM are as tuned as far as I can take them at 1866/3733Mhz and also C14 and 1T etc on the memory. 3800Mhz is just out of reach and not quite fully stable. I'll post up a Geekbench 4 later, anything else?

GB4 is plenty. Mainly trying to see if both systems are tuned how much of a gap there is for SC. I focus on that as my vr is 144hz for sims and needs a lot of sc grunt.

maybe a smt off run to see if it helps any?

True 1T is hard to do on amd so well done!

 

[MartinPrince]

GB4 is plenty. Mainly trying to see if both systems are tuned how much of a gap there is for SC. I focus on that as my vr is 144hz for sims and needs a lot of sc grunt.

maybe a smt off run to see if it helps any?

True 1T is hard to do on amd so well done!

Yep, my 3900X is pretty decently tuned. I can do an all core clock of 4325Mhz but at default I get all core of ~4275Mhz due to this great board and bios, so I just leave it at that. My default CBr20 is ~7500 which is higher than most stock scores I've seen. Anyway, your request.

With SMT

SMT OFF

How does that compare to your 9900K?

 

[Deleted User 456458]

Yep, my 3900X is pretty decently tuned. I can do an all core clock of 4325Mhz but at default I get all core of ~4275Mhz due to this great board and bios, so I just leave it at that. My default CBr20 is ~7500 which is higher than most stock scores I've seen. Anyway, your request.

With SMT

SMT OFF

How does that compare to your 9900K?

That's a really big help. Thanks a ton man. It's so hard to find well tuned systems to compare with.

This is with HT off: https://browser.geekbench.com/v4/cpu/15713878 7333 SC
This is with HT ON: https://browser.geekbench.com/v4/cpu/15714317 7300 SC

Core 52x uncore 48x Mem = 2x8 16-16-16/4200 1T secondaries and tertiary timings tightened up. Daily setup.

HT On is about ~30-35 points lower for SC, similar to yours. It seems like there is still a decent way to go for SC performance to catch up once both are well tuned.

 

[MartinPrince]

what sound and music software you running? Certainly for ripping and batch converting I find it's thread dependent in my software and can use as many threads as you throw at it, so my 3900 wins. For sound recording and post processing though, not done enough tests.

I'm using DAW's such as Cubase and Ableton Live. I don't do audio batch conversions but for anything that involves batch conversion, where you can load all cores then definitely I'll use the 3900X but for audio work where latency comes into play then the 9700K is better. The memory latency on Intel is already lower and then the ability to run memory at higher frequencies than on Ryzen can help quite a bit in some scenarios.

 

[Rroff]

Latency tests.

 

[CAT-THE-FIFTH]

Interesting that the Core i5 10600K and Ryzen 7 3700X is as what GN would say are "functionally equivalent" but the Core i3 is worse than the Ryzen 3 3300X. If Zen3 is moving to a unified L3 cache,things are going to get very interesting!!

 

[humbug]

Interesting that the Core i5 10600K and Ryzen 7 3700X is as what GN would say are "functionally equivalent" but the Core i3 is worse than the Ryzen 3 3300X. If Zen3 is moving to a unified L3 cache,things are going to get very interesting!!

I'm glad he's put that to bed, people keep conflating Intercore Latency with Input Lag, i keep telling people those things do not equate but you know how it is... Internet experts.

 

[Swinbag]

Latency tests.

This is the one I'd been waiting for personally; my use-case is 90% gaming, I've always been quite sensitive (bless) to input latency and was worried about making the switch from my SB system to Zen2.

Great that they've put this to bed - I'll likely press on and grab an X570 board & 3600 for now, with the view to picking up an 8c Zen3 part next year.

 

[Rroff]

This is the one I'd been waiting for personally; my use-case is 90% gaming, I've always been quite sensitive (bless) to input latency and was worried about making the switch from my SB system to Zen2.

Great that they've put this to bed - I'll likely press on and grab an X570 board & 3600 for now, with the view to picking up an 8c Zen3 part next year.

To be a little contrarian I don't think it entirely tells the whole tale - the actual distribution of latency can have an impact just as microstutter with framerate where it can feel different despite the same average framerate. You would have to be severely sensitive to it to notice any difference between the current crop of Intel and AMD CPUs however. (This is where the X79 platform can potentially win out for complicated reasons - not that I'd recommend people go for that platform over something newer heh).

 

[humbug]

To be a little contrarian I don't think it entirely tells the whole tale - the actual distribution of latency can have an impact just as microstutter with framerate where it can feel different despite the same average framerate. You would have to be severely sensitive to it to notice any difference between the current crop of Intel and AMD CPUs however. (This is where the X79 platform can potentially win out for complicated reasons - not that I'd recommend people go for that platform over something newer heh).

Intercore Latency, i assume this is what you are talking about as there is no "input latency difference" is the difference between an electron travelling 2mm along the die traces vs 8mm at the speed of light, i don't know what that is in time but its probably a few dozen zeros .2 of a second.

What it is is the cause of a clock cycle stall, when a core needs to communicate with another outside of its own CCX the time it takes to do that reduces the speed of that cycle by about 10%, so if you can feel something so fast it delays a clock cycle by 10% then your Ivy Bridge E must feel like sliding down a cheese grantor bare back because your IPC is about 30% lower, so even with the clock cycle stall that Cycle on Zen 2 is still 20% faster than all the cycles in your CPU.

This argument is ridiculous.

 

[CAT-THE-FIFTH]

I'm glad he's put that to bed, people keep conflating Intercore Latency with Input Lag, i keep telling people those things do not equate but you know how it is... Internet experts.

Well the fact is in all of the games tested they were getting over 120FPS anyway,which is pretty good going even for the cheaper CPUs IMHO!

 

[Rroff]

Found an older video that includes X79 albeit the testing methodology leaves a bit to be desired.

https://www.youtube.com/watch?v=dsbVSknUK7I

What is really needed in histogram break downs of these results as with sync off frames aren't arriving at equal times so you can get a huge spread of button to pixel results that aren't limited to a floor of the average frame interval. If your average latency comes from bursts of low and high latency that is going to feel, even at these small ms numbers, different to an average latency that comes from a more consistent interval closer to your average frame latency - assuming your average frame latency isn't suffering from micro-stutter like conditions.

 

[Dave2150]

Interesting that the Core i5 10600K and Ryzen 7 3700X is as what GN would say are "functionally equivalent" but the Core i3 is worse than the Ryzen 3 3300X. If Zen3 is moving to a unified L3 cache,things are going to get very interesting!!

Are you aware that Ryzen is practically brand new architecture wise, and is on a cutting edge 7nm process?

Intel on the other hand, are on a 5 year old architecture (Skylake) and a 5.5 year old process (14nm). The fact that Intel still wins in some areas (gaming) by a small margin, mean it will absolutely demolish Ryzen once Intel releases first Rocket Lake (new architecture backported to 14nm) and then Alder Lake on LGA1700 shortly after (new architecture & new process).

That's what I call interesting, not current Ryzen struggling to beat Intel in gaming, or Zen's 3 perhaps matching Intel in gaming.