Page 3 of 4 FirstFirst 1234 LastLast
Results 31 to 45 of 54

Thread: 4790K - 3950X speed comparison

  1. #31
    Registered User
    Join Date
    Jan 2004
    Location
    Stavanger, Norway
    Posts
    166
    Quote Originally Posted by Marander View Post
    @Amerelium

    Thanks! I really appreciate your post about your Noctua and Ryzen 3950X experience, otherwise I might have chosen the 3900X.

    About the second cooler: Very important is a good overall airflow. I added a slightly smaller Antec fan instead of the second Noctua (which I used in the case instead). The high RAM modules have enough space that way.

    Large tower with 1x front case cooler with air in, 2x CPU coolers and 2x rear coolers for air out. I may add another front cooler eventually.

    I don't think the PCI-E lanes should have any impact on render times / CPU.

    m.2 yes I think that takes away a bit from PCI-E but with 2x m.2 SSD Samsung EVO 1TB and 2TB plus 1x 2 TB SATA SSD and a RTX 2070 Super installed I don't see any issues.

    I do GPU rendering with on another machine and dual RTX and a third workstation with dual GTX so I leave this one with single GPU for now.

    My 3950X system runs quietly at 54-59 degrees under full load and scores between 8950 and 9360 points in Cinebench R20 depending on the Bios / TPU settings, so I'm very satisfied.
    You wanna get the hot air out of the tower as efficient as possible, then the rest takes care of itself. I installed a Noctua outake fan at the top of my cabinet, and that lowered the GPU (!) temperature at load with about 15C.

    M2s will definately slow down your CPU, IF you run out of lanes. Did so with the 4790K, where's there's just 16 in total.

    Just did a Cinebench 20 test by the way: 9457, no overclocking.


    Sensei: Yea the language thing, saw another thread about that. Another one of Win10s charming features, it will sudden ly change language settings, even WHILST one is typing. I discovered that when the same thing suddenly happned on my current Win10 Pro setup.

    I've spent the last few weeks basically hacking the Win10 installation, to get it into a bearable working condition. I have managed to reduce running processes from 145+ down to 114 - which still makes it a travesty of an operatiing system. Hacking that is mostly impossible to do in Home Edition by the way, so stay way clear of that one.
    - Ignorance is bliss...

  2. #32
    Registered User
    Join Date
    Jan 2004
    Location
    Stavanger, Norway
    Posts
    166
    Quote Originally Posted by Amerelium View Post

    Just did a Cinebench 20 test by the way: 9457, no overclocking.
    BY THE WAY: That score is after freshly rebooting the system. Score before rebooting (latest reboot yesterday) was 9274 - so there's Win10 rot for yah...
    - Ignorance is bliss...

  3. #33
    Registered User
    Join Date
    Jun 2014
    Location
    Right here
    Posts
    1,698
    Quote Originally Posted by Amerelium View Post
    BY THE WAY: That score is after freshly rebooting the system. Score before rebooting (latest reboot yesterday) was 9274 - so there's Win10 rot for yah...
    These are great Cinebench scores, above 9450!

    You mentioned no overclocking, but did you use TPU I or II automatic setting? What are your RAM specs and settings? At what frequencies does the CPU run during the benchmark?

    I don't want to push the system and CB around 9100-9300 (I think my score of 9360 with TPU I) is fine for me but if there's an easy setting without getting much higher temps then of course I don't mind.

    You mentioned you reduced the amount of processes. Do you mean during runtime only? I usually disable all Windows Services I don't need, remove tasks from the Windows Task Scheduler and configure the startup applications. Also note that the msconfig tool still exists in Win10. That helps reducing the background activities and I get rather consistent rendering / benchmark results.
    Last edited by Marander; 01-12-2020 at 12:02 PM.

  4. #34
    Registered User
    Join Date
    Jan 2004
    Location
    Stavanger, Norway
    Posts
    166
    Quote Originally Posted by Marander View Post
    These are great Cinebench scores, above 9450!

    You mentioned no overclocking, but did you use TPU I or II automatic setting? What are your RAM specs and settings? At what frequencies does the CPU run during the benchmark?

    I don't want to push the system and CB around 9100-9300 (I think my score of 9360 with TPU I) is fine for me but if there's an easy setting without getting much higher temps then of course I don't mind.

    You mentioned you reduced the amount of processes. Do you mean during runtime only? I usually disable all Windows Services I don't need, remove tasks from the Windows Task Scheduler and configure the startup applications. Also note that the msconfig tool still exists in Win10. That helps reducing the background activities and I get rather consistent rendering / benchmark results.
    Have not messed with any TPU settings that I am aware of. This is what I have changed:

    AI Overclock Tuner to D.O.C.P auto - sets the RAM to it's correct 3200 MHz (and the timers from auto to 14-14-14-34)
    Performance Enhancer to Level 3 (OC)

    By the way, I mounted my slightly smaller Noctua CPU fan from the 4790K onto the new cooler (and managed to fasten the new spare large one to it as replacement): The temperature went down a few degrees, but also, interestingly, I lost a bit of performance; about 50 points on the cinebech test, 10-20 seconds added to that LW frame I've used for testing here. Cores clock themselves to about 4,1 GHz constantly during the LW render, down from 4,15 GHz from before adding the fan. It's like it's stealing some voltage or something.

    BUT, since cooling became better, I decided to mess about a bit with the CPU multiplier. Tested 42 at first, with voltage at auto, managed to get it up to 43,5 with manual voltage offset, running stable.

    The attached screengrab will tell you how potent this CPU is.

    Note: This also shaved another minute off the rendertime of my LW testframe, albeit that is somewhat compromized by the fact that the time spent adding corona became linger - LW only uses two cores when doing that, and with ratio overclocking there's no turbo boost beyond set value. Max registred temperature 83C

    Click image for larger version. 

Name:	Untitled-1.png 
Views:	81 
Size:	21.7 KB 
ID:	146643
    Last edited by Amerelium; 01-15-2020 at 12:03 PM.
    - Ignorance is bliss...

  5. #35

    Wow, darn good, way higher OC than i thought was possible.  
    LW vidz   DPont donate   LightWiki   RHiggit   IKBooster   My vidz

  6. #36
    Registered User
    Join Date
    Jan 2004
    Location
    Stavanger, Norway
    Posts
    166
    Quote Originally Posted by erikals View Post
    Wow, darn good, way higher OC than i thought was possible.  
    ...and with air cooling, not water
    - Ignorance is bliss...

  7. #37
    Registered User
    Join Date
    Jun 2014
    Location
    Right here
    Posts
    1,698
    Wow great performance Amerelium! I will revisit the Bios settings to get some more juice out of the system.

  8. #38
    Registered User
    Join Date
    Jan 2004
    Location
    Stavanger, Norway
    Posts
    166
    Quote Originally Posted by Marander View Post
    Wow great performance Amerelium! I will revisit the Bios settings to get some more juice out of the system.

    I will probably will not push it that hard when I do my large render projects - buzz around the net is that you degrade the CPU fast if you run the voltage above 1,35 constantly.


    If I can get LW to run non-stop with 43 multiplier and voltage no higher than 1,3ish, I'll do that - if not just let it turbo itself as good as it can.
    - Ignorance is bliss...

  9. #39
    Registered User
    Join Date
    Jun 2014
    Location
    Right here
    Posts
    1,698
    Tested my 3950X again with some overclocking... Indeed great performance with a score of 10010 points and temps of max 75 degrees!

    Click image for larger version. 

Name:	Mar_CB20_R9_3950X.png 
Views:	47 
Size:	1.61 MB 
ID:	146843

    Now back to normal settings with variable clock speed... don't want to constantly push it.

  10. #40
    Super Member Qexit's Avatar
    Join Date
    Feb 2003
    Location
    Warrington, UK
    Posts
    1,118
    Sorry, just couldn't resist finding out how far behind state of the art systems my current LW PC is in terms of benchmark speeds. My dinky little Xeon E5-1620 v3 scraped in with 1380 pts multi-threaded at stock speed

    Click image for larger version. 

Name:	Cinebench20 LWPC MultiCore 01.jpg 
Views:	25 
Size:	424.4 KB 
ID:	146847
    Kevin F Stubbs

    Remember...one size does NOT fit all

  11. #41
    Super Member Chris S. (Fez)'s Avatar
    Join Date
    Feb 2003
    Location
    Virginia
    Posts
    3,000
    That 3970 is killing it in those marks. Woah.

  12. #42
    Registered User
    Join Date
    Jan 2004
    Location
    Stavanger, Norway
    Posts
    166
    Quote Originally Posted by Marander View Post
    Tested my 3950X again with some overclocking... Indeed great performance with a score of 10010 points and temps of max 75 degrees!

    Click image for larger version. 

Name:	Mar_CB20_R9_3950X.png 
Views:	47 
Size:	1.61 MB 
ID:	146843

    Now back to normal settings with variable clock speed... don't want to constantly push it.

    4,18 GHz - that's an odd clock rating - what is you OC settings?

    I notice you scored better than when I set my cores to 4,20 GHz, albeit I managed to push mine to 4,35 GHz and score 10113
    - Ignorance is bliss...

  13. #43
    Registered User
    Join Date
    Jun 2014
    Location
    Right here
    Posts
    1,698
    Quote Originally Posted by Amerelium View Post
    4,18 GHz - that's an odd clock rating - what is you OC settings?

    I notice you scored better than when I set my cores to 4,20 GHz, albeit I managed to push mine to 4,35 GHz and score 10113
    I think it was 4.2 GHz but my board is slightly slower clocking everything (like 3.79 GHz instead of 3.80). Used TPU II and stopped at 4.2. RAM is set manually at 3200 but I changed it to D. O. C. P (XMP) which improved speed alot on the 3900X so I will do this on the 3950X too.

    I prefer having optimized standard bios settings where the CPU multiplicator adjusts automatically (and goes down to 2.2 GHZ in idle or up to 4.7 GHz for individual cores). Slower then then with fixed high clock speed but still great performance.

    What do you think about variable vs fixed speed? Temp wise, both seems ok.

  14. #44
    Goes bump in the night RebelHill's Avatar
    Join Date
    Nov 2003
    Location
    jersey
    Posts
    5,783
    Coming in just shy of 7300 pts on the 3900X here... all stock, standard turbo, etc.
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	Capture.PNG 
Views:	31 
Size:	31.2 KB 
ID:	146857  
    LSR Surface and Rendering Tuts.
    RHiggit Rigging and Animation Tools
    RHA Animation Tutorials
    RHR Rigging Tutorials
    RHN Nodal Tutorials
    YT Vids Tuts for all

  15. #45
    Registered User
    Join Date
    Jun 2014
    Location
    Right here
    Posts
    1,698
    Quote Originally Posted by RebelHill View Post
    Coming in just shy of 7300 pts on the 3900X here... all stock, standard turbo, etc.
    That's a good score for a 3900X system with standard settings!

    On the 3900X I get 7136 with standard settings and 7587 with OC (but I leave it with standard settings).

    I chose the 3900X as main workstation (and occasionally games), also for single core stuff due to the faster base clock.

    For rendering I use the 3950X and it turned out to outperform even in single core tasks and runs cooler under full load. For interactive design work, there is no noticable difference between the two.

    They're both very good processors allowing to render so much quicker then the Intel systems I used before. Working with the Ryzen 9 feels very snappy and that's maybe also due to the excellent PCI-E / m.2 I/O performance.

Page 3 of 4 FirstFirst 1234 LastLast

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •