Overclock.net banner

901 - 920 of 952 Posts

·
Registered
Joined
·
951 Posts
Broke out the trusty old cinebench tonight. let the 3950X loose at close full blast to check for any signs of degradation. I'm happy to report I have noticed nothing. Voltages to hit these frequencies are exactly the same as they were back in January when I made my highest runs. These arent my highest ever scores but I still feel happy about them!

11015 in R20 4625/4625/4550/4550 @1.375V
4832 In R15 4700/4700/4625/4625 @1.375V
If I may ask, what is your temps while running these at that voltage. I am just asking as I cannot get close to 1.3vcore before my cpu goes into the 90C and above.
 

·
Premium Member
Joined
·
5,026 Posts
Not hating 🙂

You are still benchmarking cache. Easy to see with the 4k random read @ qd=1. If it's faster than one ssd, and WAY faster it is cache speed 🙂
Youre ssd is about 60-62 MB/s 4k random read @ qd=1. If you are getting 170MB/s it's cache.
But hey, high numbers is fun to watch sometimes.

Try 10-15 GB test size 🙂
It's two Gen 4 M.2's in RAID 0, and if I'm using 1GB test size, it's not testing the cache.

By the way, if you go to the bottom of the overclock.net webpage on mobile and choose Full Site, not the Mobile version, your posts won't look like

Not hating 🙂

You are still benchmarking cache. Easy to see with the 4k random read @ qd=1. If it's faster than one ssd, and WAY faster it is cache speed 🙂
 

·
Registered
Joined
·
2,330 Posts
Not hating 🙂

You are still benchmarking cache. Easy to see with the 4k random read @ qd=1. If it's faster than one ssd, and WAY faster it is cache speed 🙂
Youre ssd is about 60-62 MB/s 4k random read @ qd=1. If you are getting 170MB/s it's cache.
But hey, high numbers is fun to watch sometimes.

Try 10-15 GB test size 🙂
It's two Gen 4 M.2's in RAID 0, and if I'm using 1GB test size, it's not testing the cache.

By the way, if you go to the bottom of the overclock.net webpage on mobile and choose Full Site, not the Mobile version, your posts won't look like

Not hating 🙂

You are still benchmarking cache. Easy to see with the 4k random read @ qd=1. If it's faster than one ssd, and WAY faster it is cache speed 🙂
4k rr @ qd=1 don't scale in raid-0. If it's scaling it's negative, due to a bit higher latency in raid-0.

4k rr @qd =1 is the latency of the ssd. You can't scale latency lower...

Optane is about 300MB/s 4k rr qd=1.

Have fun with youre cache result.
 
  • Rep+
Reactions: ThrashZone

·
Premium Member
Joined
·
5,026 Posts
4k rr @ qd=1 don't scale in raid-0. If it's scaling it's negative, due to a bit higher latency in raid-0.

4k rr @qd =1 is the latency of the ssd. You can't scale latency lower...

Optane is about 300MB/s 4k rr qd=1.

Have fun with youre cache result.
You don't seem to understand I'm using two 5000MB/sec read, 4500MB/sec write Gen 4 M.2's in RAID 0 and using 0Fill with multiple queues and threads.

This is 8GB, top, 0Fill, and the bottom, Default Random, with queues at 16-16-32-32 and threads at 32.

I showed the queues and threads in the previous benchmarks, even one 0Fill with the default queues and threads.



 

·
Registered
Joined
·
73 Posts
If I may ask, what is your temps while running these at that voltage. I am just asking as I cannot get close to 1.3vcore before my cpu goes into the 90C and above.
83C max temps for 1.35V. I ran this on chilled water to about 8C. Temps maxed out at 61C
 

·
Registered
Joined
·
88 Posts
I decided to try no SMT after coming across an article claiming increase gaming speed. Ryzen master gaming mode didn't to much for me but manually OC was fun to do.
Maybe 3% increase in single core vs PBO. Multicore wasn't as bad of a drop as expected. Score increased slightly in time spy. I had a worse score with time spy extreme without smt. I gain 1 fps in rdr2. The memory transfer rate and latency had a nice improvement as did the fabric clock latency. Not worth it.

https://www.3dmark.com/spy/11414221
 

Attachments

·
Registered
Joined
·
521 Posts
I know this is not the right CPU, but given the similarities and the question, i think it fits the thread.

System:

Define S, 3 x nf a14 front intake, 1 stock fan rear exhaust. Front panel and filter removed.
Cooler nh d15
Mobo crosshair 6 hero (bios 7803)
3900x

In prime95 at 4,3ghz, 1,275v, 160w (core+soc power) hits 93-94ºC at max fan speed and 25ºC ambient temperature, which would make about 70ºC delta. Is that normal? Isnt that too much? Ik ryzen 3000 thermal density is high, but im seeing GN temps with 198w on 3950x (same exact internal layout like 3900x) sitting at 55ºC delta on a test bench, which is probably worst compared to the "wind tunnel" i have here.
 

·
Robotic Chemist
Joined
·
2,928 Posts
In prime95 at 4,3ghz, 1,275v, 160w (core+soc power) hits 93-94ºC at max fan speed and 25ºC ambient temperature, which would make about 70ºC delta. Is that normal? Isnt that too much? Ik ryzen 3000 thermal density is high, but im seeing GN temps with 198w on 3950x (same exact internal layout like 3900x) sitting at 55ºC delta on a test bench, which is probably worst compared to the "wind tunnel" i have here.
Have you tried remounting? That does seem hot but I think GN has a very good CPU, or very good contact with the heat spreader, or something. The thickness of the solder can vary CPU to CPU.

GN had better deltas v.s. air than I did too, until I used liquid metal between the heat spreader and my water block. The EK Velocity isn't a great water block for a 3950X but with >1 GPM and 3x560mm keeping the water at ~2°C above ambient I also did not expect to be beaten by GN's test bench. They do say to not compare these numbers to your own, so I suppose this is an example why.
 

·
Overclocker
Joined
·
11,388 Posts
It also matters if you can dump all that heat out of the cooler into case and then out of the case. It is no joke to be dumping 300+W GPU and a 250W CPU into a case haha and then use say dust filters and only 2 intake fans. You make a hotbox.
The question is is your cooler getting cold while heatpipes are hot or are even heatpipes cold, or the whole thing is hot and the case is hot and you made a hotbox?
No cooler will save you if it has no air to breathe.
 

·
Registered
Joined
·
1,130 Posts
I know this is not the right CPU, but given the similarities and the question, i think it fits the thread.

System:

Define S, 3 x nf a14 front intake, 1 stock fan rear exhaust. Front panel and filter removed.
Cooler nh d15
Mobo crosshair 6 hero (bios 7803)
3900x

In prime95 at 4,3ghz, 1,275v, 160w (core+soc power) hits 93-94ºC at max fan speed and 25ºC ambient temperature, which would make about 70ºC delta. Is that normal? Isnt that too much? Ik ryzen 3000 thermal density is high, but im seeing GN temps with 198w on 3950x (same exact internal layout like 3900x) sitting at 55ºC delta on a test bench, which is probably worst compared to the "wind tunnel" i have here.
I think it is normal.
If you let amd auto regulate through pbo OC , prime smallfft will push the vcore droop to around 1.23-1.25v at 4ghz. At your clocks and vcore, i m not surprised the heat density overwhelming your cooling.

You can try loosening llc to let it vdroop or lower vcore
 

·
Registered
Joined
·
521 Posts
Cooler was badly mounted and software was buggy as well. It dropped to 79-80°C after changes.

Sent from my Redmi K20 Pro using Tapatalk
 

·
Registered
Joined
·
1 Posts
I just tried 4.4 GHz all cores, and VCore stays at 1.248V (set on auto) during AIDA64 CPU Stress Test. Max temperature is 64°C (with a Noctua NH-D15 SE-AM4), otherwise in IDLE 35°C (room temp = 25°C) . Motherboard used is ASUS Strix X570-E Gaming with bios v2204. Ram are also overclocked from 3200MHz to 3600MHz (Corsair Vengeance LPX 2x16GB). I measured the data with CPUID HWMonitor, AIDA64 and HWiNFO64.
 

·
Registered
Joined
·
57 Posts
I just tried 4.4 GHz all cores, and VCore stays at 1.248V (set on auto) during AIDA64 CPU Stress Test. Max temperature is 64°C (with a Noctua NH-D15 SE-AM4), otherwise in IDLE 35°C (room temp = 25°C) . Motherboard used is ASUS Strix X570-E Gaming with bios v2204. Ram are also overclocked from 3200MHz to 3600MHz (Corsair Vengeance LPX 2x16GB). I measured the data with CPUID HWMonitor, AIDA64 and HWiNFO64.
That's very low temperature, did you do a fpu test in AIDA? You could try a CCX OC since first chiplet sould be better, LL3 and 1.275 as I remember should give you a 1.25V and less under high load. You probably can push ram to 3733MHz, the newest batches can do a 1900MHz on fclk, but you have to try it yourself, 1866MHz should work on almost all.
 

·
Registered
Joined
·
521 Posts
I just tried 4.4 GHz all cores, and VCore stays at 1.248V (set on auto) during AIDA64 CPU Stress Test. Max temperature is 64°C (with a Noctua NH-D15 SE-AM4), otherwise in IDLE 35°C (room temp = 25°C) . Motherboard used is ASUS Strix X570-E Gaming with bios v2204. Ram are also overclocked from 3200MHz to 3600MHz (Corsair Vengeance LPX 2x16GB). I measured the data with CPUID HWMonitor, AIDA64 and HWiNFO64.
When you give us temps, give us power consumption too, it's way easier to contrast data.

Sent from my Redmi K20 Pro using Tapatalk
 

·
Registered
Joined
·
2,330 Posts

·
Registered
Joined
·
73 Posts
Congrats on 29. place on Hall of fame for single card :)

Was it with chiller on cpu or even colder?
Thanks yeah I know im never gonna get one of the very top positions as I dont run Intel or LN2. But im happy with this considering its Ryzen and Its just chilled water. This winter ill try and push further as condensation wont be an issue and water temps should be colder.
 

·
Registered
Joined
·
22 Posts
Good evening, guys. How are you all doing?
I have a 3950X and I'm looking to make my FCLK overclock stable. I can boot and run Cinebench just fine with 1900 FLCK, but fail at Prime95 Small FFTs test after less than a minute. Temperature never goes above 85 C (360mm AIO).
With 1800 FLCK, I can go as low as 1.13V and still do well on Prime95 with a fixed 4.0Ghz overclock (with beautiful temps). However, with 1900 FLCK it'll fail even with 1.20V.
My SOC settings are:
SOC Voltage: 1.09375V
VDDG CCD: 0.965V
VDDG IOD: 0.965V
VDDP: 0.950V
Board is a Asus TUF X570-Plus.
Seems like increasing any of these make it less stable. Should I try further lowering then? Any ideas are welcomed. If there's any information missing, let me know. Thanks in advance.
 

·
Robotic Chemist
Joined
·
2,928 Posts
I gave up on 1900 flck on my 3950X, it just won't do it no matter what. :(
 

·
Premium Member
Joined
·
5,026 Posts
901 - 920 of 952 Posts
Top