2400G @ 4175MHz opinions?
Printed From: ASRock.com
Category: OverClocking(OC) Zone
Forum Name: Achivement&Record
Forum Description: Share your HWBOT or Futuremark score here!
URL: https://forum.asrock.com/forum_posts.asp?TID=8553
Printed Date: 21 Nov 2024 at 11:55pm
Software Version: Web Wiz Forums 12.04 - http://www.webwizforums.com
Topic: 2400G @ 4175MHz opinions?
Posted By: cristy6100
Subject: 2400G @ 4175MHz opinions?
Date Posted: 29 Apr 2018 at 5:43pm
|
Just for fun I got my 2400G up to 4175MHz to see the Single/MultiThreading score https://valid.x86.fr/fwp2c8" rel="nofollow - https://valid.x86.fr/fwp2c8 With these scores, it even beats a 2700X overclocked to 4350MHz in Single threading So all in all I would call Raven Ridge the fastest CPU from AMD in regards to Single Threading Here's a 2700X @ 4350MHz for comparison, using 3200MHz memory also https://valid.x86.fr/s8ng35" rel="nofollow - https://valid.x86.fr/s8ng35 Opinions? ------------- Main Rig: AMD Ryzen 2400G | ASRock Fatality X370 ITX/ac UEFI 4.90 | Corsair LPX DDR4 2400@3200 1.35V | Corsair RM650i PSU | CM ML120L AiO Cooler |
Replies:
Posted By: Spaceminer
Date Posted: 30 Apr 2018 at 12:38am
|
Would you try a run @3.825ghz? My 1600 might actually top your single thread. I don't want to raise my voltage that high or I would try mine at 4.175ghz too. Here's mine @3.825ghz. http://valid.x86.fr/bench/yfl6r9/1" rel="nofollow - http://valid.x86.fr/bench/yfl6r9/1 I don't know how to submit a score and get the cpuz image to show up like yours did. Every time I try it won't let me stay anonymous. Edit: I got it to submit, but now the original link doesn't work. I can only find it on the bench page. I don't know what gives with cpuz. I updated the link. |
Posted By: PetrolHead
Date Posted: 30 Apr 2018 at 3:51am
There's something wrong with those 2700X numbers. Single threaded performance should be pretty much the same clock for clock between all Ryzen CPUs (of the same generation). If anything, the 2400G should be slower due to having only one fourth of the L3 cache the 2700X has. I suspect that the 2700X isn't really able to keep those clocks stable during the benchmark or that whoever ran the test had some other programs running in the background. For comparison, my 1500X scores between 435 and 438 in the single threaded benchmark @ 3.8 GHz (3790.95 MHz) / 3333 MHz (3325.4 MHz). This means your ~10% higher clocks net you a ~10% higher score. The 2700X has ~4% higher clocks than you do, and should thus also get ~4% higher scores. Even the multi-threaded score of the 2700X isn't as high as one would expect (it should be more than double your score). ------------- Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit |
cristy6100 wrote:Posted By: Spaceminer
Date Posted: 30 Apr 2018 at 9:11pm
Just wanted to give a thanks for posting your numbers. Your 1500x and my 1600 are looking the same clock for clock. I think we both have well binned CPUs. I usually dust other 1600's at the same clock and even some 1700x's. I'd image your chip is bit of an over performer too. |
Posted By: cristy6100
Date Posted: 01 May 2018 at 4:21am
|
Ok, so here is my 2400G @ 3818MHz with 1.152V and memory at 3200 with 1.35V and everything else default. I did a re-run with 1.4V CPU voltage and the scores jumped to about 438-440 at the same 3818MHz with 3200 Memory, I cant do a validation yet because it will take me to my previous one. I will post it once I can validate it It seems lowering the Vcore bellow its default voltage will affect performance on Raven Ridge, I lost about 15 points in single thread just by lowering the default voltage from 1.398V to 1.152V
https://valid.x86.fr/q4int0" rel="nofollow - https://valid.x86.fr/q4int0 EDIT: I think I need to wait or something between validation, if I submit it again it will take me to my old one, maybe its because I only changed the Vcore voltage. Here is a screen shot with everything default except RAM at 3200 and CPU downclocked to 3.825GHz PS: I wanted one of the CPU-Z windows to show the memory tab but forgot to switch to it before taking the screen shot  ------------- Main Rig: AMD Ryzen 2400G | ASRock Fatality X370 ITX/ac UEFI 4.90 | Corsair LPX DDR4 2400@3200 1.35V | Corsair RM650i PSU | CM ML120L AiO Cooler |
Posted By: PetrolHead
Date Posted: 01 May 2018 at 5:57am
Well, I'd expect pretty much any 1xxx CPU to give similar results at these clock speeds. After all, the cores are the same in all of the CPUs. If someone gets significantly different results, my first assumption would be that the cause is differences in the environment the CPU has to work in (installed Windows updates, programs running in the background, CPU-Z version etc.), not the CPU itself. Regarding binning, I don't think my CPU is particularly well binned. I wasn't able to overclock it beyond 3.7 GHz and 3200 MHz with reasonable voltages and temperatures using the stock Wraith Spire cooler. This may be of course be due to the fact that unlike many reviewers, I used Prime95 for stability testing, but I think a good chip could have been able to push to 3.8 GHz or 3.9 GHz with the same settings. Now that I'm using a liquid AIO cooler, I'm able to run the CPU at 3.8 GHz and 3333 MHz, but this required a notable jump in Vcore (from 1.25V to 1.325V in BIOS) and 3.9 GHz seems to still be out of the question with less than 1.4V (in BIOS, the real Vcore is higher on my motherboard). Not that I'm unhappy with the performance of the 1500X or the speeds I'm reaching, I just don't think I've won the silicon lottery with my CPU. ------------- Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit |
Posted By: PetrolHead
Date Posted: 01 May 2018 at 6:14am
Voltage should have zero effect on performance as long as the CPU is stable at the used voltage. More likely there was something happening in the background that threw the benchmark off. It wasn't too long ago some of my CPU-Z scores were lower than they should have been because Windows had decided it was a great time to start dowloading some updates. ------------- Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit |
Posted By: Spaceminer
Date Posted: 01 May 2018 at 6:43am
" rel="nofollow -
Have you tried using a voltage offset? I get the same thing at higher voltages. If I need to run up 1.35v, then I'll set the Vcore at 1.30v with a +50mv offset. This allows you to essentially keep a hard limit on it. LLC is trying to account for Vdroop and bumping you up. I drop ~0.0625v under load. If I'm using the stock clocks, I actually have to set a -50mv offset instead, otherwise the VDD hits 1.47v briefly on auto. Which is absolutely insane to me. I like to be very cautious. I can run 3.9ghz at 1.33725v. And 3.925ghz at 1.35625v. Anything less and I will crash just from running Dolphin emulator for 2-3min. I haven't tried beyond this. I know a decent voltage jump is needed somewhere beyond that, and I'm not willing to try it on air cooling. I wouldn't count yourself out just yet. Prime95 really is a torture test. I say, if you can do the things you want with the OC you have, be it gaming, video editing, compiling code etc., without it crashing, or hitting dangerous temps, then it's stable enough for every day use. I like my vcore on the low side, so I keep mine at 3.725ghz with 1.31v for everyday use. LLC will occasionally push me to 1.32v when I set it this low. I peak at about 54°C under load with ambient temps of 22-23°C. |
Posted By: cristy6100
Date Posted: 01 May 2018 at 7:02am
|
Well, here's a validation at full stock settings, just after the UEFI downgrade to 4.60 (AGESA 1.0.0.1a from the Beta 1.0.0.2). Nothing changed in UEFI, just rebooted after upgrade, so memory is at default XMP's of 2400 for these modules. https://valid.x86.fr/q65tcy" rel="nofollow - https://valid.x86.fr/q65tcy As you can see in the validation Voltage is at around 1.344V but that is with Vdrop, in UEFI its 1.398V and thats default, the Turbo Boost voltage is 1.45V but with Vdrop it stays at about 1.42V, CPU-Z cannot report the Boost voltage for Ryzen, only default voltage. Raven Ridge has very high Vcore from the get go, some users report even 1.5V when CPU is boosting. Summit Ridge has lower default voltage even when boosting but it cannot sustain the boost on more than 1-2 cores. In this validation you can see the CPU boosted all cores to 3.75GHz, and went to 3.8GHz for the single core benchmark. Default clock for 2400G is 3600MHz Regarding LLC, the ITX variants dont have LCC option in UEFI to compensate, but I would say the drop is not that high anyway. I have seen much higher Vdrop's on AM3+ FX CPU's :))
------------- Main Rig: AMD Ryzen 2400G | ASRock Fatality X370 ITX/ac UEFI 4.90 | Corsair LPX DDR4 2400@3200 1.35V | Corsair RM650i PSU | CM ML120L AiO Cooler |
Posted By: PetrolHead
Date Posted: 01 May 2018 at 1:49pm
Nope. It seems that my motherboard has some auto LLC going on by default. If I set 1.325V in the BIOS (as I have), the voltage (SVI2 TFN) is about 1.362V when idling or under light load and about 1.337V when doing Prime95 testing, so I'm always above what I set in the BIOS.
On my board the auto voltages go beyond 1.5V at times. It does seem odd, but it shouldn't actually be harmful to the CPU (after all, it's the stock setting). The voltage keeps jumping up and down so those high spikes are not kept for very long and this is mainly happening when the CPU is not under load, which means the current running through the CPU isn't that high.
You're probably right. I just want it to be a bit more stable than stable enough.  ------------- Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit |
Posted By: PetrolHead
Date Posted: 01 May 2018 at 2:04pm
The link doesn't seem to work. In any case those voltages seem quite high, since 1.35V is supposed to be the safe limit for 24/7 (or 1.4V according to some sources), which means under load you're already near the "limit" (it's not a hard one, of course). But then that voltage also depends on which sensor it's taken from, and the limits may not be exactly the same for Raven Ridge and Summit Ridge. ------------- Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit |
Posted By: Xaltar
Date Posted: 27 May 2018 at 6:27am
|
" rel="nofollow - The APUs don't use the infinity fabric to link 2 core clusters like other Ryzen CPUs do. This alone will result in better per core performance. Interesting results for sure. I know the R3 and R5 1k series both disable 2 cores from each CCX rather than utilize a single 4 core CCX. I wonder if the APUs even have the infinity fabric at all? ------------- 
|
Posted By: PetrolHead
Date Posted: 26 Jul 2018 at 6:29am
Based on Hardware Unboxed's (somewhat limited) tests, the performance difference between using one CPU complex and two CPU complexes is largely negligible outside synthetic (L3) cache benchmarks. ------------- Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit |