oops... bought a 3600! -benches incoming

NattyKathy

NattyKathy

[H]ard|Gawd
Joined
Jan 20, 2019
Messages
1,483
one thing led to another and here I am taking the train back from Microcenter with a new CPU in tow. Also took advantage of the bundle deal to pick up an ASRock B450M Pro4 for effectively $25... not that the board will be much of an upgrade from my B450M HDV but more memory slots, an extra PCIe slot, more fan headers, more VRM heatsinking, and 10Gbit Type-C seemed worth it.

I'll post some benchmarks vs my 1600 when I get it running :)
 
Yeah, I fried my 1800x Computer trying to install a AIO and walked into Microcenter the next day to buy a motherboard and saw 30 3900x instock so i bout the combo just like you. I am fucking blown away. 7 hour car render just turned into a 90 minute render for the same quality. I am shocked I'm not joking.
 
Epyon said:
Yeah, I fried my 1800x Computer trying to install a AIO and walked into Microcenter the next day to buy a motherboard and saw 30 3900x instock so i bout the combo just like you. I am fucking blown away. 7 hour car render just turned into a 90 minute render for the same quality. I am shocked I'm not joking.
Click to expand...

What kind of render are you running?
 
+1 for benchmarks. I was dissapointed at the initial numbers posted of R3 by AMD, but am very interested in apples to apples 1600 to 3600 numbers. Apparently they actually did get IPC up, and not just in cinebench. Plus the actual higher clocks.
 
Thecold,

I am running Keyshot 8.2 with my Ferrari render as a test of the 2 different times. This thing is fucking fast man.
 
travm said:
+1 for benchmarks. I was dissapointed at the initial numbers posted of R3 by AMD, but am very interested in apples to apples 1600 to 3600 numbers. Apparently they actually did get IPC up, and not just in cinebench. Plus the actual higher clocks.
Click to expand...

I've got a handful of gaming and content creation tests based directly on my real-world use, including 1600 @ both stock and 4Ghz all-core. running now :)
 
KATEKATEKATE said:
I've got a handful of gaming and content creation tests based directly on my real-world use, including 1600 @ both stock and 4Ghz all-core. running now :)
Click to expand...
Nice.
commercial reviews tend to provide little data and mountains of fluff. Almost written as if they are selling shit (This site was actually not like that, when it did reviews). Often the "other" processors they are comparing the new processor dont run as "good" as I know they do cause I own the bloody things and have years of experience using them. Upgrade itch is real, but i've hardly even used my 1600, it spent the first year of its life in the mines. Now that its my main gaming rig, and I bought a silly 144hz monitor...... (I like it, but i only ever see 100-120fps, first world problems right). GPU's cost too much presently, so a nice little drop in CPU might scratch the itch just hard enough that I dont blow another $1500 on new gear.
 
I kind of thought that there would be no way to see the Big picture as AMD has allowed the owner of the new product to test and review going back to cheap mATX 350b boards and have no drop off in performance and who is to say which chipset Gen 3 Zen likes best as 370x /470x/570x = we get to do the R&D now .

Remember how much freedom they give us with 2200g program and not a bad way to get the Black OPs R & D done in the field for free while they tune the drivers to your liking ..

And then Navi with Gen 3 and what Jays 2 cents stated about Fire Strike and why Navi excels in it .. the combined score test to see it start in the 40 fps off the bat with a 3600/5700 pairing (auto boost on cpu) on 370x tells me these two worked together in that RDNA some how when paired showed very strong http://www.3dmark.com/fs/20139458 .. That's waxed some Intel ass right there from just the 6 core model = 44.32 fps
 
Last edited:
crazycrave said:
I kind of thought that there would be no way to see the Big picture as AMD has allowed the owner of the new product to test and review going back to cheap mATX 350b boards and have no drop off in performance and who is to say which chipset Gen 3 Zen likes best as 370x /470x/570x = we get to do the R&D now .

Remember how much freedom they give us with 2200g program and not a bad way to get the Black OPs R & D done in the field for free while they tune the drivers to your liking ..

And then Navi with Gen 3 and what Jays 2 cents stated about Fire Strike and why Navi excels in it .. the combined score test to see it start in the 40 fps off the bat with a 3600/5700 pairing (auto boost on cpu) on 370x tells me these two worked together in that RDNA some how when paired showed very strong http://www.3dmark.com/fs/20139458 .. That's waxed some Intel ass right there from just the 6 core model = 44.32 fps
Click to expand...

sorry, not sure I understand?
I'm just testing how my new 3600 will compare to my old 1600 for my own personal workloads when paired with my Vega 64 on a B450 board. Nothing to do with Navi or B350/X370 platform tests or the APUs or any direct comparisons to Intel...
 
Last edited:
finished the stock/OC 1600 benches last night--going to install the new board + 3600 today, wish me luck!
 
spent all afternoon getting the mobo swapped. everything seemingly working fine on the first try, temps seem a bit high but so far so good. Running stock except for 2666MT memory set to 2933.

initial impressions-
-CPU is boosting to 4Ghz regardless of # of cores loaded @ ~1.35v
-3600 @ stock all-core reporting ~70W
-1600 @ stock all-core reported ~80W
-1600 @ 4Ghz all-core reported ~140W
-3600 @ stock hitting ~65C full load
-1600 @ stock hit ~45C
-1600 @ 4Ghz hit ~65C

I couldn't help but to test the best-case gaming scenarios first:

Far Cry 5, Ultra, 1080P, HD textures
+13% vs 1600 @ same clocks
+27% vs 1600 @ same power

Minecraft, 1080P, SEUS shaders
+ ~13% vs 1600 @ same clocks
+ ~28% vs 1600 @ same power

more benches working + detailed power consumption & overclocking
 
Randall Stephens said:
This on stock cooler?
Click to expand...
noooo that's with a Corsair H115i Pro RGB lol. I'm going to try remounting the pumphead but I don't have super high hopes.... power density of one Zen 2 CCX chiplet at stock is similar to highly-overclocked Zen 1 die. And the stock 3600 temps are about the same as I got on the 1600 @4Ghz & ~140W accordingly

I'm curious though, how well the stock cooler would do? H115i liquid temps stay roughly at ambient with the 3600 @4Ghz vs a noticeable rise with the 1600 @4Ghz. The cooling limitation is clearly getting that much heat out of such a small space
 
Last edited:
KATEKATEKATE said:
noooo that's with a Corsair H115i Pro RGB lol. I'm going to try remounting the pumphead but I don't have super high hopes.... power density of one Zen 2 CCX chiplet at stock is similar to highly-overclocked Zen 1 die. And the stock 3600 temps are about the same as I got on the 1600 @4Ghz & ~140W accordingly

I'm curious though, how well the stock cooler would do? H115i liquid temps stay roughly at ambient with the 3600 @4Ghz vs a noticeable rise with the 1600 @4Ghz. The cooling limitation is clearly getting that much heat out of such a small space
Click to expand...

How did remounting the pumphead go? :p:p:p
 
Randall Stephens said:
How did remounting the pumphead go? :p:p:p
Click to expand...
haven't done it yet too busy benchmarking & messing with ryzen master henever I sit down at the machine lol

I will though- I want to swap the 1600 back in to get more power readings anyways. I'll also try different pumphead alignment to see if repositioning the microfins helps at all... I think rotating it 90* from the "usual" might give better chiplet coverage. tbd.
 
Last edited:
if you need any other numbers from the 1600 let me know i can get some from mine although for some reason my 1600 runs at 1600x base but 1600 boost clocks when everything is set at default on my taichi so they might be slightly different.

i'll probably swap it out wed for my 3600/x570/ddr4 3600 i bought and haven't had the time to install.
 
KATEKATEKATE said:
noooo that's with a Corsair H115i Pro RGB lol. I'm going to try remounting the pumphead but I don't have super high hopes.... power density of one Zen 2 CCX chiplet at stock is similar to highly-overclocked Zen 1 die. And the stock 3600 temps are about the same as I got on the 1600 @4Ghz & ~140W accordingly

I'm curious though, how well the stock cooler would do? H115i liquid temps stay roughly at ambient with the 3600 @4Ghz vs a noticeable rise with the 1600 @4Ghz. The cooling limitation is clearly getting that much heat out of such a small space
Click to expand...


The stock cooler is decent but it is barely enough to tame a 3700x at stock...With ambient temps of 70f I was hitting 75c during benchmark runs.

It's amazing how much of an improvement the 3600 is vs the 1600, especially at the price point AMD offers it.


Will your ram do at least 2933? Zen2 really needs high memory and IF speed to shine. If you can get 3000c15 or better then you get 90%+ of the performance with higher memory speed.
 
sirmonkey1985 said:
if you need any other numbers from the 1600 let me know i can get some from mine although for some reason my 1600 runs at 1600x base but 1600 boost clocks when everything is set at default on my taichi so they might be slightly different.

i'll probably swap it out wed for my 3600/x570/ddr4 3600 i bought and haven't had the time to install.
Click to expand...


So you are replacing your x370 board with an x570? Any particular reason why? From what I have read the taichi does really well with the Zen2 SKUs.
 
ccityinstaller said:
So you are replacing your x370 board with an x570? Any particular reason why? From what I have read the taichi does really well with the Zen2 SKUs.
Click to expand...


hate dealing with asrocks bios and how you have to jump through hoops to update them so i'm going to give this board plus my 2600x to a family friends kid. the gigabyte x570 elite cost less than this taichi when i bought it new and everything on that board is either a side grade or an upgrade over the taichi.
 
ccityinstaller said:
The stock cooler is decent but it is barely enough to tame a 3700x at stock...With ambient temps of 70f I was hitting 75c during benchmark runs.

It's amazing how much of an improvement the 3600 is vs the 1600, especially at the price point AMD offers it.


Will your ram do at least 2933? Zen2 really needs high memory and IF speed to shine. If you can get 3000c15 or better then you get 90%+ of the performance with higher memory speed.
Click to expand...

agreed abt the improvement vs 1600. perf/MHz is as high as they come now and perf/W under heavy load is really good. And the architecture is different enough that there's opportunity for huge increases in certain non-gaming scenarios--for example, a common DNG -> TIFF export I do with batches of 1000's of medium-reso images now takes ~60% less time, clock-for-clock :eek: (the 32MB of L3 cache is doing its magic here, no doubt)

I've been running my memory at 2933 cl18, same as I had it on the 1600. I'll try and tighten up the timings once I'm done tinkering with CPU clocks. When I have the $$ to spare I plan on looking for a low-latency 3200MTs 32GB kit.

Those air temps are helpful thanks... I guess my AIO is doing some good, it just feels odd to see 60-65C at <80W with a 280mm rad. Dense silicon is dense :confused:

Good stuff these. Wish clocks were a bit higher but can't have everything for $200.
 
sirmonkey1985 said:
if you need any other numbers from the 1600 let me know i can get some from mine although for some reason my 1600 runs at 1600x base but 1600 boost clocks when everything is set at default on my taichi so they might be slightly different.

i'll probably swap it out wed for my 3600/x570/ddr4 3600 i bought and haven't had the time to install.
Click to expand...

I'd love to hear what your SVI2 readings are like with various numbers of cores loaded, if you have hwinfo64 and are so inclined :) building a picture of Zen1 perf/W at various levels for broader comparison to Zen2
 
KATEKATEKATE said:
I'd love to hear what your SVI2 readings are like with various numbers of cores loaded, if you have hwinfo64 and are so inclined :) building a picture of Zen1 perf/W at various levels for broader comparison to Zen2
Click to expand...

what kind of loads are you looking for? the ones i can give easily are that running BOINC seti@home full load my SVI2 readings are
CPU voltage 1.169v
SOC 1.1v
cpu current ~77A
SOC current ~12.5A
cpu package power 84w
cores stay between 8.5-9.1w each (core 0 is always my lowest power usage core, core 3 tends to always be higher than the rest)
cpu core power ~90w
SOC power 13.8w
total package 105w.

cpu temps are at 50.8C @ 3.4Ghz

at 60% load even though windows scheduler will still make it use all 12 threads
voltages unchanged
cpu current 63A
soc current 11.5A
cpu package power 70w
core power between 5.8-7.9w
cpu core power 74w
soc power 12.7w
total package 87w

cpu temps 48C @ 3.4Ghz (mind you the AIO is completely equalized since it's been sitting at full load under seti for about 3 weeks now 24/7)

idle(~5%) using ryzen power plan 90/100 set
voltages unchanged
cpu current 15.7A
soc current 10A
cpu package power 26w
cores between .37-.9w
cpu core power 18w
soc power 11w
total 29w

28C @ 3.4Ghz (windows claims it's running at 3.1Ghz but it's not)
 
so i decided to switch over to the 3600.. first boot, nothing changed in bios, no XMP settings nadda exactly 1000 points higher than my 1600.. what i find even more hilarious is that with my AIO my cpu just sits pegged all core at the 4.2Ghz max boost @ 67C. think i may just set the overclock offset to +200 and hope for the best but honestly with that much of a performance gain over the 1600 i'll probably just leave it at stock settings.

1600 @ 3.4Ghz 2494 ddr4 3000CL16
3600 @ 4.2Ghz 3496 ddr4 2400CL16

at full load i'm using between 20-30w less power at my UPS and at idle about 10w less.
 
Last edited:
dang, you're actually getting 4.2Ghz all-core @ stock? I think I lost the silicon lottery big time with mine... still haven't seen anything over 4Ghz sustained. PBO & auto OC does nearly nothing, maybe a few dozen extra Mhz. Not even touching advertised clocks at stock lol. Manual OC gets to 4250-4275 and hits a wall no matter how much voltage I throw at it. Latest ASRock UEFI with AGESA 1.0.0.3ABB (two Bs)

Thanks for taking the time to post your power readings :) That's exactly what I was looking for
 
Last edited:
KATEKATEKATE said:
dang, you're actually getting 4.2Ghz all-core @ stock? I think I lost the silicon lottery big time with mine... still haven't seen anything over 4Ghz sustained. PBO & auto OC does nearly nothing, maybe a few dozen extra Mhz. Not even touching advertised clocks at stock lol. Manual OC gets to 4250-4275 and hits a wall no matter how much voltage I throw at it. Latest ASRock UEFI with AGESA 1.0.0.3ABB (two Bs)

Thanks for taking the time to post your power readings :) That's exactly what I was looking for
Click to expand...

yeah i have BOINC running right now at 90%(11 of 12 tasks) sitting at 4.166Ghz all core @ 67C.. that seems to be the cut off, if i keep the temps at 66C or lower it'll sit at 4.190-4.2Ghz but if it goes over 66C it'll drop to 4.166. boards still on the F3 bios it came with, but i need to buy another flash drive to update it. also the default voltages if this helps you any are 1.35v LLC set to auto. idle voltage is .944v. big thing i need to do right now is play with my fan curve settings for my AIO but i want to see what kind of frame rates i can hit in some of the games i know i was cpu limited in before i go and mess with that some more.

as i've been typing this though my clock rates are dropping slightly as my AIO becomes heat soaked.. now at 71C i'm running 4.116Ghz compared to the 4.19@66C, 4.16@67C and 4.144@69C so that kind of shows how temp sensitive it is when it comes to clock rates..

also PBO's useless, you can set the auto OC independent of the PBO settings and should still see some effect if your temps are low enough.
 
KATEKATEKATE said:
dang, you're actually getting 4.2Ghz all-core @ stock? I think I lost the silicon lottery big time with mine... still haven't seen anything over 4Ghz sustained. PBO & auto OC does nearly nothing, maybe a few dozen extra Mhz. Not even touching advertised clocks at stock lol. Manual OC gets to 4250-4275 and hits a wall no matter how much voltage I throw at it. Latest ASRock UEFI with AGESA 1.0.0.3ABB (two Bs)

Thanks for taking the time to post your power readings :) That's exactly what I was looking for
Click to expand...

Running prime 95 I hit 4200 sustained on all cores at stock pretty much 100% of the time, temps are around 56-57c at load. Prior to water it was MUCH more variable. Temps are a huge deal.
 
Last edited:
I'm seeing 62~65C on the core under max load with a H115i, so temps shouldn't be too much of a problem. maybe I just got a dud chip, or my UEFI is doing weird stuff with my turbos (I'm on the latest "Destiny 2 compatible" firmware, maybe all the stuff they fixed broke other things?)
 
remounted the AIO the other day, quick test showed temps about identical to before. coldplate and liquid flow must be the limitation as others have posited elsewhere. I was unable to try rotating the pumphead orientation because of interference with the AM4 bracket and hoses so it will remain a mystery if positioning the microfins differently would change anything.

i have been a bit preoccupied but I'll get to posting benchmarks etc this week
 
You must log in or register to reply here.
Back
Top