Latest Benchmarks - Intel vs AMD

Cockpit res “every frame“ is a huge FPS Killer in certain modules. You should take EF off for testing, for sure with a 1060.

Otherwise, great work.

Now we need a x080ti in your rig to see the real deal.

Gesendet von iPhone mit Tapatalk

I squeezed 5-10 more fps from 0.1% lows if I leave VCore voltage at Auto Which is insane voltage that goes up to 1.5v, hence I lowered it, sorry for not mentioning it, I didn't think it would have any effect. Lol is 1.5v even safe :D? The cpu/fan behaviour is annoying on default voltage/windows power plan settings. CPU load spikes either randomly or when you launch apps, and fan annoyingly spools up and down.

It's safe. AMD stated that Ryzen (that was in case of 2xxx) series is designed to handle such voltages and it's just different from Intel. I think a lot of people make simple mistake of applying their approach from Intel cpu to AMD one when the two have very different arch.

My 2700X is well cooled with AIO and despite voltages reaching 1.52 V it stays below 75 - 80 deg at all times while boosting on most cores beyond 4.352 Ghz and some reaching 4.397 to 4.402 Ghz. And that's mostly with default settings.

With 3xxx series I've read that it's best to let cpu do its own thing and not mess around. I'm waiting for ASUS to deliver Bios with new AGESA and then will pick 3900X.

Iirc from what I have seen from YT vids this is no safe and will degrade your 3000 series in a few months.

Iirc it was Actual Hardcore Overclocking Talking about X570 boards.

For 7nm 1.45 or even 1.5v sounds a lot

Gesendet von iPhone mit Tapatalk

From AMD on Reddit:

Please note that it is totally normal for your Ryzen to use voltages in a range of 0.200V - 1.500V -- this is the factory operating range of the CPU. It is also totally normal for the temperature to cycle through 10°C swings as boost comes on and off. You will always see these characteristics, as they're intended, so do not be surprised to see such values. :)

Wow, hell yes. You are right.

When you measure the cores under a full core load of prime95 with AVX, Small FFT's, what does it read ? and how many watts ( aka Ampere ) are fed ?

My 95w TDP CPU has peaks at 205w ...waahaaa, thats like 157 Amps at 1.30v under load

Looks like undervolting doesn’t help much at all.

I was also kind of shocked to see stock voltages in the 1.45-1.5 range when I plopped my 3700x into my x470 board. Thought asrock might have gotten things wrong for the bios but apparently it’s totally normal. Even asked on another big forum I frequent ar15.com and folks were saying similar things.

Gamers nexus just put out a video about undervolting and it looks like it actually hurts performance...

Old setup:

CPU: i5 3570k OC (boosts up to 4.2GHz)
RAM: 24GB DDR3-1866 (CL 10-11-10)
GPU (same): Gigabyte GTX 1060 6GB

New setup:

CPU: Ryzen 3700x Stock (boosts to 4.3GHz+)
RAM: 32GB DDR4-3600 (CL 17-19-19)
GPU (same): Gigabyte GTX 1060 6GB

Game settings:

 

 

Man thanks for this, it's a nice benchmark and I can tell you put some time into it. I just upgraded from an i5-4670k w/ 2080Ti to a 3800x w/ 2080Ti. While I didn't benchmark before the upgrade (should have!), here are my numbers on the new system.

3800x at stock (boosts to 4.5ghz)
32GB DDR4-3200 (CL14-14-14)
2080Ti Founders Edition at stock (factory OC)
Game settings: I made them identical to yours except I don't use reshade.

I didn't make a pretty graph, but here are the numbers:

27-07-2019, 14:01:09 DCS.exe benchmark completed, 33220 frames rendered in 195.813 s

Average framerate : 169.6 FPS

1% low framerate : 99.4 FPS

0.1% low framerate : 52.3 FPS

It does look like you're GPU limited but the 1060 does keep the 0.1% lows near mine, so I wonder how much of that is affected by CPU. That looks to be around the difference most reviewers saw between otherwise-identical 3700x/3800x systems (just a few percentage points).

The results are on 2k resolution?

The results are on 2k resolution?

Yes

thx

Man thanks for this, it's a nice benchmark and I can tell you put some time into it. I just upgraded from an i5-4670k w/ 2080Ti to a 3800x w/ 2080Ti. While I didn't benchmark before the upgrade (should have!), here are my numbers on the new system.

 

3800x at stock (boosts to 4.5ghz)
32GB DDR4-3200 (CL14-14-14)
2080Ti Founders Edition at stock (factory OC)
Game settings: I made them identical to yours except I don't use reshade.

 

I didn't make a pretty graph, but here are the numbers:

27-07-2019, 14:01:09 DCS.exe benchmark completed, 33220 frames rendered in 195.813 s

Average framerate : 169.6 FPS

1% low framerate : 99.4 FPS

0.1% low framerate : 52.3 FPS

 

It does look like you're GPU limited but the 1060 does keep the 0.1% lows near mine, so I wonder how much of that is affected by CPU. That looks to be around the difference most reviewers saw between otherwise-identical 3700x/3800x systems (just a few percentage points).

I'm hopefully gonna get 2080 ti by end of this week or next week, so i can test that too. Anyways i found out that my memory wasn't running most efficient settings. My infinity fabric seems to suck big time in silicone lottery, i can't even run 1800 on it, so running memory at 1800(3600) doesn't seem to be most efficient thing. Need bios updates to see if it fixes anything.

2x 16GB or 4x 8GB ?

In either case, the first has dense modules and the latter has 4 modules. Each one has it's pain.

If 4, try with only 2 and see if that works, just as a test.

2x 16GB or 4x 8GB ?

 

In either case, the first has dense modules and the latter has 4 modules. Each one has it's pain.

 

If 4, try with only 2 and see if that works, just as a test.

2x16

Ya, that is the drawback of 32GB. In either case, it's tougher than 2x 8GB.

Are they in the QVL withthat rated speed ?

I doubt the IMC is uncapable for 1800MHz silicon wise. According to AMD, theyshould all do DDR-3600, just maybe not with 2x 16GB or 4x 8GB. That is the question to find the answer for.

Old setup:

CPU: i5 3570k OC (boosts up to 4.2GHz)
RAM: 24GB DDR3-1866 (CL 10-11-10)
GPU (same): Gigabyte GTX 1060 6GB

New setup:

CPU: Ryzen 3700x Stock (boosts to 4.3GHz+)
RAM: 32GB DDR4-3600 (CL 17-19-19)
GPU (same): Gigabyte GTX 1060 6GB

Game settings:

Resolution: 2560x1440
Everything on minimum or off if possible, except the following:
Textures – High, Terrain Textures – High, Visib. range - high,
Res. Of cockpit Displays: 1024EF, Preload raidus: 60500,
Anisotropic filtering: 16x,[x] Full Screen, [x] Scale GUI

Reshade: Default Lumasharpen and HDR effects on.

 

As you can see I’m probably (and hopefully) GPU bottlenecked, and difference between 6 y.o. setup is not big.

The way I measured it is I recorded a track(Caucasus map), where your location gets shot by MLRS after some time (which causes fps drop). Camera motions and plane motions in the track are the same (or close to being so I hope). All I did after is played track 3 times and averaged the values (I start playing the track by pressing pause and about half second later I press the button to record benchmark data with MSI afterburner for 3m:16s). I will attach the track I used (you need su-25A or flaming cliffs 3 to play it).

In first day of installation 3700x performance was actually worse in 1% lows by 10 fps, but today Ryzen gods had mercy and made it slightly better.

Such an improvement to your benchmark, if it would be better to choose a module SU-25T which is free.

The query in the test are the mirrors off?

Such an improvement to your benchmark, if it would be better to choose a module SU-25T which is free.

The query in the test are the mirrors off?

Mirrors are off. I might do new one with something free and no reshade, after i receive new gpu hopefully soon.

Ya, that is the drawback of 32GB. In either case, it's tougher than 2x 8GB.

 

Are they in the QVL withthat rated speed ?

 

I doubt the IMC is uncapable for 1800MHz silicon wise. According to AMD, theyshould all do DDR-3600, just maybe not with 2x 16GB or 4x 8GB. That is the question to find the answer for.

Yes it's on QVL list of the motherboard. I can run memory at 3600, but then infinity fabric clock gets automatically set to 1600, if i try manualy set it to 1800: PC doesn't last long and freezes/doesn't boot.

Should be interesting to see how PCIe 4.0 plays into all this

https://www.google.com/amp/s/amp.tomshardware.com/news/what-we-know-about-pcie4,39063.html

Should be interesting to see how PCIe 4.0 plays into all this

https://www.google.com/amp/s/amp.tomshardware.com/news/what-we-know-about-pcie4,39063.html

Ya, read that back in May. PCIe v5 is close too, so v4 will have a short life.

Good so, I hate PCH bottlenecks and a myriad of USB but you cant use them. What a freaking joke that is.

Mirrors are off. I might do new one with something free and no reshade, after i receive new gpu hopefully soon.

Thank you for the information and I look forward to the next benchmark.

Apropos Voltage, PBO and Marketing:

I wasn't too wrong when I said those Voltages seem dangerously high, or at least questionable in the big picture.

From the comments:

That info comes from The Stilt, an ASUS community member who makes custom BIOSes etc. 1.325V is the figure stated for a safe voltage when all cores are being utilized, when running a benchmark or encoding video at 100% usage. Especially for hours or days at a time. To go above that risks degrading the chip (over time) and to go below that gives you some headroom and also lowers temps if you want a quieter case or have a subpar cooler (or it's summer). So if one was to manually OC all cores to the same frequency, that is the number that's been stated for maximum safe voltage. If you're just using stock boost, that number is sort of irrelevant. Unless you've got a dodgy BIOS that's using voltages a lot higher than that, which some do. In that case, some people have been using a negative voltage offset to bring the temps more inline with that figure. And also keep in mind, after more time and feedback.. the info out there could change. It's still very early and everyone is effectively a beta tester at this stage.

Yes, ASUS is consistently overvolting my 2700X. I wish I could see typical voltages seen on MSI or Gigabyte boards.

Apropos Voltage, PBO and Marketing:

 

 

I wasn't too wrong when I said those Voltages seem dangerously high, or at least questionable in the big picture.

 

 

From the comments:

 

That info comes from The Stilt, an ASUS community member who makes custom BIOSes etc. 1.325V is the figure stated for a safe voltage when all cores are being utilized, when running a benchmark or encoding video at 100% usage. Especially for hours or days at a time. To go above that risks degrading the chip (over time) and to go below that gives you some headroom and also lowers temps if you want a quieter case or have a subpar cooler (or it's summer). So if one was to manually OC all cores to the same frequency, that is the number that's been stated for maximum safe voltage. If you're just using stock boost, that number is sort of irrelevant. Unless you've got a dodgy BIOS that's using voltages a lot higher than that, which some do. In that case, some people have been using a negative voltage offset to bring the temps more inline with that figure. And also keep in mind, after more time and feedback.. the info out there could change. It's still very early and everyone is effectively a beta tester at this stage.

So you are saying my cpu will die in couple of month? :) coz my voltages reach almost 1.5 while doing anything rly, even randomly on idle.

So you are saying my cpu will die in couple of month? :) coz my voltages reach almost 1.5 while doing anything rly, even randomly on idle.

Watch Builzoid's YT video, it will tell you that.

The IDLE volts are that high cuz it CAN, that's the new algo at work.

Moving the mouse, the classic single thread light-load, causes highest allowed volts as it THEN is able to get to the Turbo Frequencies advertised.

The more you load it, the lower it should drop in Volts and MHz. Full core loads MUST have lower volts and thus MHz to not break the volt/temp limits.

Watch that video, regardless if you only understand 10%, it's worth the time.

Run prime95 v.29.5b6 with SMALL FFT's ( as in the popup, select the one on top ) THAT will put your CPU under 100% all core load. I bet it will drop FAR below 1.4 volts.

As that Beta Bios builder says, 1.325v and no higher under full load or it will degrade fast. I take that serious as someone who builds beta bios's for sure knows more about that CPU than we all together right here in this forum

edit*

link: couldnt find the version I have, this is newer and will also obey AVX: https://www.guru3d.com/files-details/prime95-download.html

screenshot shows how mine drop to 1.28 under load, tho that has a totaly different reason ( LLC ) but it shows you were to look in HWinfo. Start HWinfo BEFORE you start p95 :smilewink: