Why do my two Ryzens perform so differently?

[Arigs]

Hey guys, I have built two machines that I use at home for mostly just running stuff that requires a lot of processing power. I am having a very weird issue where I am running the exact same software on both and its performing completely differently. Ryzen 1 seems to be very stable and typically floats around 30-60% CPU usage at all times with 50 threads active in my software (its something I coded myself not any commercial product). Ryzen 2 with the 1700 (not 1700X) CPU seems to be very odd...whenever threads are launched/closed it will spike up much more than the Ryzen 1 machine. For example if its at 30% CPU usage and I re-launch 1 out of 50 threads it will spike up to like 60% CPU then back down whereas the Ryzen 1 (1700X) machine will go from 30% to 35% as expected. On top of this, when I look at the CPU usage in Task Manager, about half the time I see that the CPU usage is maxed out at 100% and the line is just literally up at the top of the graph for a solid minute. The tasks being performed are the exact same. The ram is the same. The SSDs are pretty much the same...the GPU as well as the same (very little GPU is used anyway). I kept the bios settings the same initially but I have tried fiddling with SMT, Core C6 state, Core Performance Boost, all the different options, tried disabling and enabling them in different combinations and currently what I have is SMT enabled, C6 disabled, Core Performance Disabled. Both machines are running a clean version of the same Windows 10 OS as well. Both are on High Performance Power Plans. 

 

Here are the two builds:

 

Ryzen 1 

- Tomahawk B350 Motherboard

- Ryzen 1700X (OC'd to 3.8GHZ @ 1.35v)

- Corsair Vengeance LPX 32GB (4x8GB) DDR4 DRAM 3200MHz (PC4-25600) (CMK16GX4M2B3200C16) (running at 2133 speed)

- Noctua NH-L9x65 SE-AM4 CPU Cooler

- R9 380 2GB GPU

- ADATA SP920 SSD

- Corsair 650W PSU

 

Ryzen 2

- Tomahawk B350 Motherboard

- Ryzen 1700 (OC'd to 3.7ghz @ 1.35v)

- Ballistix Sport LT 32GB (2x16GB) Single DDR4 2400 MT/s BLS16G4D240FSE (running at 2133 speed)

- Cryorig H7 CPU Cooler

- R7 370 4GB GPU

- ADATA SU800 SSD

- EVGA 500W PSU

 

I have no idea why Ryzen 2 is so slow and maxes out the CPU so hard...like Ryzen 1 almost never go above 50% even (in the screenshot I used it was during a very active part of the code but usually its 20-40% CPU usage and performs the same amount of work as Ryzen 2 because I can see the results being the same). One odd thing I do notice in task manager's CPU graph is the fact that on Ryzen 1 (Stable Machine) when I show the CPU graph PER processor it shows all processors are DIFFERENT usages, whereas Ryzen 2 (unstable / laggy machine) has all of the CPUs seemingly at the exact same usage (pictures attached).

 

Ryzen 1: 

 

 

Ryzen 2: 

 

 

Not sure if that is any indication of what may be going on...but yeah.

 

I've tried updating Chipsets Drivers, AMD drivers, flashing the B350 Tomahawk to latest bios and all that, so I can't really figure out what it is exactly. Anyone have any ideas of what it might be, or things I can try to fix this? 

 

[The Sloth]

RAM, ryzen loves ram speed, with  3600mhz ram a ryzen system can almost match a 7700k in 1080p gaming( can even get better results in some games). 

[AlwaysFSX]

The infinity fabric runs at half of whatever your ram speed is, lower creates more delay between when cores can talk to each other and when you have workloads that have a lot of cross-core talking going on you're going to have a bad time.

[Orangeator]

7 minutes ago, nerdslayer1 said:

RAM, ryzen loves ram speed, with  3600mhz ram a ryzen system can almost match a 7700k in 1080p gaming( can even get better results in some games). 

 

5 minutes ago, AlwaysFSX said:

The infinity fabric runs at half of whatever your ram speed is, lower creates more delay between when cores can talk to each other and when you have workloads that have a lot of cross-core talking going on you're going to have a bad time.

Read his post, they are both running at 2133mhz...

[typographie]

7 minutes ago, nerdslayer1 said:

RAM, ryzen loves ram speed, with  3600mhz ram a ryzen system can almost match a 7700k in 1080p gaming( can even get better results in some games). 

If I read it correctly, the RAM in both systems is running at 2133 MHz, despite being 2400 MHz and 3200 MHz kits.

[SaintPosum]

1 minute ago, typographie said:

If I read it correctly, the RAM in both systems is running at 2133 MHz, despite being 2400 MHz and 3200 MHz kits.

more channels due to one being a 4 stick kit as opposed to a two set????? maybe?

[The Sloth]

5 minutes ago, Orangeator said:

 

Read his post, they are both running at 2133mhz...

I meant to say channels. 

[Arigs]

15 minutes ago, AlwaysFSX said:

The infinity fabric runs at half of whatever your ram speed is, lower creates more delay between when cores can talk to each other and when you have workloads that have a lot of cross-core talking going on you're going to have a bad time.

 

16 minutes ago, nerdslayer1 said:

RAM, ryzen loves ram speed, with  3600mhz ram a ryzen system can almost match a 7700k in 1080p gaming( can even get better results in some games). 

 

Ah yes, that does make sense but both systems are running at 2133Mhz as mentioned by the above 2 posters. That being said, one ram is clearly superior to the other (able to run up to 3200Mhz if I wanted to try that but that ram is already working fine at 2133 so I never bothered to mess with it), and it is running as 4 sticks instead of 2, but could that really lead to such a significant difference (20-40% average CPU usage versus constantly jumping to 100% and staying there for some time?). 

 

I mean I could try and put these sticks in a separate machine performing less CPU intensive tasks and buy another 4x8GB sticks but I prefer not to spend another $400 on ram if I don't have to  So could that 4 channels versus 2 channels and better ram really be the cause of this huge difference?

 

 

[AlwaysFSX]

15 minutes ago, Orangeator said:

Read his post, they are both running at 2133mhz...

Didn't see that part, my bad.

7 minutes ago, Arigs said:

Ah yes, that does make sense but both systems are running at 2133Mhz as mentioned by the above 2 posters. That being said, one ram is clearly superior to the other (able to run up to 3200Mhz if I wanted to try that but that ram is already working fine at 2133 so I never bothered to mess with it), and it is running as 4 sticks instead of 2, but could that really lead to such a significant difference (20-40% average CPU usage versus constantly jumping to 100% and staying there for some time?)

Ryzen is a bit weird running 4 sticks vs 2 sticks (AMD recommends running 2 sticks of RAM for high speeds) but at 2133 it shouldn't be that affected. Swap the RAM around and see what happens?

[Arigs]

Just now, AlwaysFSX said:

Didn't see that part, my bad.

Ryzen is a bit weird running 4 sticks vs 2 sticks (AMD recommends running 2 sticks of RAM for high speeds) but at 2133 it shouldn't be that affected. Swap the RAM around and see what happens?

Going to try them in different slots now and also just push it up to 2400Mhz as well since it supports that. Will report back results in 30 minutes.

[Ordinarily_Greater]

Just now, Arigs said:

Going to try them in different slots now and also just push it up to 2400Mhz as well since it supports that. Will report back results in 30 minutes.

Ok, I start PM you if I need something

[Arigs]

Okay, looks like my very first test resolved the problem...Moved each ram stick over 1 slot, used the default XMP profile to set it to 16-16-16-39 2400Mhz, and suddenly my CPU graph on Ryzen 2 is exactly the same as Ryzen 1 now. No more spikes when threads are launched, hovering around 20-60%, never hitting 100%.

 

It did not occur to me that the slots themselves could cause bad performance...such a weird thing...I know you can have 'bad slots' where memory may not register, or the system won't boot, but slots making the sticks perform at slower speeds??? That one is new for me, but a good experience here nonetheless.

 

Thanks for the suggestion, can't believe how easily this was resolved!

[AlwaysFSX]

2 minutes ago, Arigs said:

Okay, looks like my very first test resolved the problem...Moved each ram stick over 1 slot, used the default XMP profile to set it to 16-16-16-39 2400Mhz, and suddenly my CPU graph on Ryzen 2 is exactly the same as Ryzen 1 now. No more spikes when threads are launched, hovering around 20-60%, never hitting 100%.

 

It did not occur to me that the slots themselves could cause bad performance...such a weird thing...I know you can have 'bad slots' where memory may not register, or the system won't boot, but slots making the sticks perform at slower speeds??? That one is new for me, but a good experience here nonetheless.

 

Thanks for the suggestion, can't believe how easily this was resolved!

Oh yeah, motherboards (specifically for Ryzen) are really specific on what slots you're going to populate if you're only using two sticks. Strange to think about if you've been using any computer ever lately where that hasn't mattered much, if at all.

[typographie]

3 hours ago, SaintPosum said:

more channels due to one being a 4 stick kit as opposed to a two set????? maybe?

Ryzen doesn't support quad-channel. Four DIMMs is still dual-channel on this platform, provided it's two matching pairs.