Overclock.net banner

141 - 160 of 359 Posts

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #141
x264 / Y-Cruncher HWiNFO logs

3.8GHz_OC_x264_Y-Crun.zip 1822k .zip file


IIRC ~30min I let system idle after Y-Cruncher, then I fired up 3DM FSE set to combined test loop only, ~1.25hrs. The attached 3DM save file IIRC should open on another's system for viewing.





3DM_FSE_CT_Log.zip 423k .zip file


Next I ran SWBF, 1440P "Ultra" preset, with Crimson driver set to FRTC: 89FPS, FreeSync: On, Power Efficiency: On. Due to these settings the GPU will not stick to max clock in game, I don't find performance lacking or see any stutter/issues. Once stability testing of CPU OC finish will revert to my Fury X OC ROM 1145/545. HML file attached with all monitoring data that MSI AB support.

SWBF.zip 81k .zip file


Rig now running [email protected], been ~45min, will update thread with result once finish "the run!
 

Attachments

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #143
No worries
smile.gif
, thanks for support and viewing
wink.gif
, keeps me going when get "beat up"
thumb.gif
.

~2.5hrs in now, no new WHEA errors on profile so far.



The reason why I go nuts on stability with OC is not just that I'd hate a game bombing, data issues, etc, but for when I want my rig to do stuff like below:-

~174hrs straight run on [email protected] via my Q6600 rig (CPU/mobo had since 2007
biggrin.gif
, still rocking it in 2017).



HWiNFO log for run
 

·
Hardware Abuser
Joined
·
3,451 Posts
Quote:
Originally Posted by gupsterg View Post

No worries
smile.gif
, thanks for support and viewing
wink.gif
, keeps me going when get "beat up"
thumb.gif
.

~2.5hrs in now, no new WHEA errors on profile so far.



The reason why I go nuts on stability with OC is not just that I'd hate a game bombing, data issues, etc, but for when I want my rig to do stuff like below:-

~174hrs straight run on [email protected] via my Q6600 rig (CPU/mobo had since 2007
biggrin.gif
, still rocking it in 2017).



HWiNFO log for run
I understand completely, I have run [email protected] in the past and it takes ~ 100% stability to run consistently and error free.
 

·
Registered
Joined
·
3,921 Posts
So no Win 8.1 updates on ryzen? Also does 6 cores overclock higher than 8, anyone tried that yet?
 

·
Registered
Joined
·
1,367 Posts
@gupsterg

Im going back to 0902...and see difference between 1001 and 0902.

Update1: I get better temperatures with 0902! 10c lower on Load...
Funny Cant say which one is real!!
 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #147
@diggiddi

Info that I have read seems to point to no huge gains.

@majestynl

To me 0902 with Sense MI: [Disabled] seems right for temps. Only thing I'm noticing is sometimes CPU sensor from Asus EC is "whack", I reckon it's when I use lower polling rate in HWiNFO than default 2000ms. Will be testing theory sometime soon.

Total rig uptime since last reboot: ~35hrs, ~5.25hrs x264 > ~1hr idle > ~5hrs Y-Cruncher > 3DM FSE CT loop ~1.25hrs > SWBF 0.25hrs > [email protected] 20.5hrs.



Will be checking log once finish [email protected] on when tCTL reached >~76°C how many times it happened. As viewing screenshots at it's first occurrence (~9hrs in [email protected]), the fans on front intake/CPU did not ramp up above ~750 RPM and they should have. Viewing x264 tCTL over 48 loops MAX: ~70°C AVE: ~61°C and Y-Cruncher over 6hrs MAX: ~78°C AVE: ~70°C, fan ramping did occur to ~1200 RPM.
 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #148
Quote:
Originally Posted by The Stilt View Post

No need to use silly high VDDCR_SoC voltages
wink.gif

0.9125V set from bios, measures 0.915V under load.



Hynix AFR.
Quote:
Originally Posted by The Stilt View Post

I think we've yet to see the actual limitations of the memory controller, so probably all chips will do 3200MHz+. At least once the software & firmware (AGESA) side has been sorted.

The 1700 specimen I'm using has the lowest leakage of all Ryzen chips I've seen so far. So technically it should require higher voltage in every aspect, than the chips with higher leakage
wink.gif


I would suggest:

- Set the VDDCR_SoC voltage to 0.95 - 0.975V from the bios.
- Set 2933MHz MEMCLK & save changes, reboot.
- Re-enter the bios and change the MEMCLK to 3200MHz, save changes and reboot.

DRAM training should pass now, as long as your memory can do the frequency with the settings you are using.

Despite the single rank (sided) modules are currently 2933-3200MHz capable, running dual rank (sided) modules will require using lower DRAM ratios and increasing the BCLK (which is not recommended).

Corsair 3000C15 kit (Ver. 5.30) uses Hynix AFR ICs.
3200C16 could be the same, or Hynix MFR or Samsung x-Die (check the version number from the label).
5.xx = Hynix, 4.xx = Samsung.
+rep, thank you for this info
thumb.gif
.
 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #149
Members viewing my x264 post 136 and Y-Cruncher data post 138 will see mobo / CPU socket temp sensor is not stuck at 31°C, but is in [email protected] post 147.

I do not know why this occurred, I believe it is a board UEFI / Asus EC FW issue. As fans are not ramping when this issue occurs. I would assume board uses CPU socket sensor for fan ramping and not tCTL as well.

So I ended [email protected] ~23.5hrs.



I then updated from v5.47-3109 to latest version released today, v5.47-3115. I loaded HWiNFO and then RB, I got same stuck mobo / CPU socket sensors at 31°C.



So I reboot and restart testing.





No stuck mobo / CPU socket sensor and fans are ramping.
 

·
Registered
Joined
·
3,921 Posts
Quote:
Originally Posted by gupsterg View Post

@diggiddi

Info that I have read seems to point to no huge gains.
So a V8 is more desirable than a V6 ...again, just like the FX, but with M$oft shennanigans
mad.gif

I might just have to defect and get an I7 6700K
Repped UP BTW
 

·
A fish with teeth!
Joined
·
1,076 Posts
Yeah I think it is just limited by the characteristics of the 14nm LPP node. The 6- and 4-core chips are supposed to use the same die but with some cores disabled (though that means there could be potential for unlocked cores!).

@gupsterg

I should have my chip tomorrow but unfortunately my board probably isn't going to ship until next week
mad.gif
so I won't be able to provide any details on max stable clocks, but I can provide the info off the heatspreader. Do I just post it here or is there a link to a form I should fill out?
 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #152
@diggiddi

I was gonna wait til R5 released, in a Bit Tech review I noticed 1600X is ~$259 and say 1700 is ~$329 to me it seemed for the extra 27% cost I got extra 25% cores/threads. So as the plan is to go longer term use than my i5 I thought I'd stump up for it.

Dunno if the 1600X like the R7 X CPU will not come with a cooler, if it doesn't and seeing how the Wraith Spire on ebay uk has been going for ~£40, I believe if I sell the cooler on a FVF promo+cheap shipping I could net ~£30 back.

3x R7 1700 @ CPU: 3.8GHz Cache: 3.8GHz DF: 1200MHz RAM: 2400MHz C14

3x i5 4690K @ CPU: 4.9GHz Cache:4.4GHz RAM: 2400MHz C11

3DM FSE the gap on combined test closes compared with FS, GS/GT1/GT2 remains within run to run variance and in both case similar gap
redface.gif
.

FSE R7 1700 vs i5 4690K

FS R7 1700 vs i5 4690K

I know above is only one case of SW, but the R7 has double the cores and 12 more threads than i5 and is slightly behind in GS/GT1/GT2/Combined test.

It will be interesting to see the results when I do disable SMT and/or cores on R7.

[GUIDE] 3DMark Score Calculation - how to calculated your 3DMark Score

@bardacuda

No form yet
redface.gif
, will try to setup, post up details when you like and will add to DB
smile.gif
.
 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #153
So yesterday due to
post-flame-small.gif
The Stilt's
post-flame-small.gif
info on SOC and feeling confident on my 3.8GHz OC stability, the new round of testing began last night!
biggrin.gif
.

i) measure voltages for SB, 1.8V PLL, VDDP, DRAM, NB SOC on DMM when CPU at stock, loaded with x264.

ii) fix all measured voltages in UEFI manually to gain same as testing in step i

iii) alter PState 0 for 3800MHz, set CPU voltage offset to +137.5mV so final VCORE measured on DMM was ~1.350V when under load. I also set LLC LVL1 as wanted to make sure if left on "Auto" UEFI "Auto Rules" were not adjusting it to higher levels.

iv) Global C-State Control [Enabled] in AMD CBS menu as wanted to make sure if left on "Auto" UEFI "Auto Rules" were not changing it.

Updated is just enable D.O.C.P 3 so mobo setup XMP profile of 2400MHz C14.

1st up x264, increased loops from 48 to 64
biggrin.gif
.



Currently on Y-Cruncher, ~4hrs in so far
smile.gif
.



When viewing above data bare in mind been having some issues on CPU socket/mobo temp sensors. To me CPU socket sensor is used for fan profile in UEFI, so if stuck low fans keep low and if stuck high, even if tCTL lower fans will stay at x level.

To me the new tests are a) confirming my OC is finally stable b) no need for extra SOC or VCORE for 2400MHz C14 vs 2133MHz C15.
  • SB: ~ 1.069V
  • 1.8V PLL: ~1.815V
  • VDDP: ~0.915V
  • DRAM: ~1.227V
  • NB SOC: ~0.843V
  • VCORE: ~1.355V
 

·
A fish with teeth!
Joined
·
1,076 Posts
Model: 1700
Product SKU: YD1700BBAEBOX
CPU SKU: YD1700BBM88AE
Batch: UA 1708SUT
Country: China
CPU Clock: TBD
Voltage: TBD
Cooling: Hyper 212 Evo
Max Load Temp: TBD
App Loading CPU: TBD
Mobo: ASUS Prime X370-Pro
Firmware: TBD
RAM: F4-3200C16D-16GVKB
Amount: 2x8GB
Clock: TBD
Timings: TBD
Validation Link: TBD

 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #155
@bardacuda

+rep
thumb.gif
, thank you
thumb.gif
.

@subscribers

So earlier post I had extended test on x264 for 2400MHz C14 vs 2133MHz C15, 64 loops (6.75hrs) vs 48 loops (~5.25hrs).

Well I have success in Y-Cruncher as well
biggrin.gif
, 2400MHz C14 (~7.5hrs) vs 2133MHz C15 (~6hrs)



To me, on my CPU, SOC needs no adjustment (~0.843V) for 2400MHz C14 vs 2133MHz C15.

Gonna game for an hour, then [email protected] time for rig
wink.gif
.
 

·
Always Learning
Joined
·
1,194 Posts
Quote:
Originally Posted by elmor View Post

Using DMM to measure voltage will be accurate only in idle. During load you will read higher than what the CPU is actually getting because of power plane droop being accounted for when the VRM outputs voltage.

The VRM uses on-die sense outputs to read accurate voltage at the "destination". It will thus output a higher voltage during load because there will be significant voltage drop across the CPU and ground power plane. To get a more accurate reading you need to measure at the MLCCs at the back of the CPU socket, and be sure to also get your ground point from there.

Posting this here, because many people were worried about DMM reporting higher Vcore readings.
 
  • Rep+
Reactions: gupsterg

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #157
Cheers
wink.gif
.

Already added to OP earlier today, also there is post from member comparing "Pro belt" VCORE with socket, ~19mV difference.

I have no access to rear of socket on my case, may mod one in
wink.gif
. But then again the small difference I reckon it's better to go with the slightly higher VCORE read from Pro belt. But valid info for all of us to know
thumb.gif
, so thanks.
 

·
Registered
Joined
·
367 Posts
My Gigabyte Aorus Gaming 5 board defaults at 1.1V SOC should I manually reduce that? I was thinking of putting in 0.95v Running at 2400mhz across 4 dimms.
 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #159
The Stilt has shown we don't need excessive SOC even for 3200MHz RAM, post 148.

Then in my small bump in RAM speed from 2133MHz C15 to 2400MHz C14 I'm not requiring more than stock SOC. I'm using 2x 4GB single sided RAM. I'm waiting on the return from RMA the Trident Z 3200MHz C14 2x 16GB back to do further RAM increase.

My mobo on UEFI defaults 2133MHz selects ~0.843V, then just selecting 2400MHz it shoots to ~1.050V when SOC on [Auto] and no other changes to setup. This to me seems excessive, I think this by looking at what the current official support RAM info is. I would assume from that data if we are staying within official RAM parameters we should be able to use stock SOC.

Try 0.95V and share your experience
smile.gif
. I added a Memory Stability testing section in OP today, grab HCI Memtest and run several instants required to max RAM usage, AFAIK as this creates some load on the CPU, cache is being tested as well.
 

·
Meddling user
Joined
·
7,423 Posts
Discussion Starter #160
Quote:
Originally Posted by The Stilt View Post

Just keep the voltages at sane levels (< 1.45V for VDDCR_CPU, < 1.10V for VDDCR_SoC) and don't tamper with the load-line settings, unless you actually MEASURE significant amounts of droop, under load (which is not likely on C6H). Voltage overshoot hurts just as bad as undershoot, when it comes to stability. If you need to increase the load-line setting (i.e. introduce overshoot) to maintain stability, then your voltages are not set correctly to begin with.

The load-line options in bios translate to:

Auto = ±0% (1.425mOhm)
Level 1 = -40% (0.855mOhm)
Level 2 = -50% (0.7125mOhm)
Level 3 = -75% (0.35625mOhm)
Level 4 = -85% (0.21375mOhm)
Level 5 = -100% (0.0000mOhm)

I personally recommend to keep the load-line settings at "Auto" at all times, unless you are doing LN2 runs.

The main difference between the different Ryzen 7-series SKUs (aside of the clocks) is the leakage. The 1700 SKUs have low leakage characteristics, while both 1700X & 1800X are high(er) leaking silicon. Because of that 1700 requires even less load-line biasing than the other two (due the currents being lower).
So based on bold text, the info you supplied for say AMD GPU (quoted below) apply for CPU as well? thanks in advance, just wish to be clear
smile.gif
.
Quote:
Higher leakage means the GPU will require less voltage to operate, however it´s maximum safe voltage level is lower at the same time. Lower leakage parts require higher voltages, however their break down voltage is slightly greater too.
Quote:
High ASIC "Quality" (Leakage) = Lower operating voltage, larger current draw, hotter, less energy efficient (due higher losses)

Low ASIC "Quality" = Higher operating voltage, lower current draw, cooler, more energy efficient

Unless you are using LN2 you definitely want the leakage to be as low as possible. Even under LN2 the high leakage characteristics are only desired because the difference in voltage scaling. All ASICs despite the leakage have some sort of design specific absolute voltage limit. The low leakage ASIC might run into this limit prior reaching the maximum clocks.
 
141 - 160 of 359 Posts
Top