After a long needed hibernation

[muzz]

Thats a loooooong wait dude , but yea, thubans were great, i had fx later but it was a wash ,not really worth the upgrade.Now im on some old intel, but im getting a ryzen 3000 when the dust settles .
Im just not that excited about the whole x570 thing , my pc at idle/low load is QUIET. And i want it to stay that way.The way mainboards are reviewed these days is not my cup of tea.I need data, not information on oled screens and led connectors.

True Story, good things come to those who wait!
I thought about jumping back in since the 1090, but they just weren't competing, I wasn't using my desktop very much, didn't really wanna spend the $ on subpar performance, and I don't do intel unless it's my work laptop.
I don't even really NEED to build anything, but I'm finally excited enough for AMD to most likely do it, and gaming is something I miss doing.
Unfortunately this will require CPU/Mobo/Mem/VCard AND OS as well, so not sure how I wanna go about it money wise, I may just get a 470...still TBD.

No LEDS please, geez I see some of the reviews with the flashing leds and I SMH. To each their own, but nah, I'm good!

[bignazpwns]

True Story, good things come to those who wait!
I thought about jumping back in since the 1090, but they just weren't competing, I wasn't using my desktop very much, didn't really wanna spend the $ on subpar performance, and I don't do intel unless it's my work laptop.
I don't even really NEED to build anything, but I'm finally excited enough for AMD to most likely do it, and gaming is something I miss doing.
Unfortunately this will require CPU/Mobo/Mem/VCard AND OS as well, so not sure how I wanna go about it money wise, I may just get a 470...still TBD.

No LEDS please, geez I see some of the reviews with the flashing leds and I SMH. To each their own, but nah, I'm good!

I got my x470 gaming 7 just for the excessive amount of RGB LED's lol. And it was pretty much spot on max oc as my crosshair vii but pbo for some reason hit higher on it and I could always push memory about 100.hz higher.

[bluestang]

I got a X470 Taichi with a 2700X in it now which will eventually be swapped for a 3900X.

Also, have an empty X470 CH7 that is waiting for the 3950X whenever that may be.

Can't wait to see your tips and tweaks!

[muzz]

Having Fun yet chew*?

[chew*]

Not yet. have a Mercedes sprinter at shop that needs to get finished before I can play. its being stubborn.

[vario]

I have this general question to everyone.
How is windows 7/8 compatibility ? x470 chipset and x570 chipset.
I KNOW that there are functioning x570 /ryzen 3000 drivers for these systems, cinebench records before E3 were made on windows 8.1.

[muzz]

Not yet. have a Mercedes sprinter at shop that needs to get finished before I can play. its being stubborn.

Sounds like an ASUS K7M, that thing was a PITA, and the buggiest board I've ever owned. Good Luck!
I have a few items in 2 carts, but haven't committed...YET!

[FlanK3r]

finally, first one is home, tomorrow 3800X

[chew*]

quick summary so far.

taichi overvolting cpu go manual voltage. from what I understand its not just them.

2 chip runs hot. I had to ditch stock cooler and install a noctua n12 due to thermal shutdown but its still not enough heatsink.

3 this 3700x is a dud but the 3900x is not if I can tame its temps

4 the asrock other than the vcore issue is doing exactly what I tell it to do so no quirks there so far.

[charged3800z24]

My 3700x is a dud as well. Looking forward to the 3900x results Chew*

[IFMU]

Here is the comparison system. I guess its not a bad chip but not a great chip.

For all intended purposes I will be running it stock with some tweaks since while it Ocs decent for a 9900K its a very very hot piece of glass @ 1.4v @5.1 it can exceed 100c on a dark rock pro 4 under full loads.

I run it stock due to this and the fact that in gaming the oc nets little to no gains vs the temps that get out of hand so how I run it is how I test it.

asrock phantom gaming 9
9900K
2 fury x
4x8g 3600 cl15

nice

[chew*]

so investigating this chip a little more.

as you can see I'm borderline temp/overheating fixed @ 4.1 @1.25v fyi cb brings chip to about 70c and prime brings the noise @ 96c

2.jpg


then I checked what volts amd is using 1.3 for 4124

3.jpg


then I checked single core volts and its a whopping 1.475 for 4349 single core

4.jpg


the only real win here is I can do 4.1 @ less volts and score better in cb somehow with a fixed lower clock.

[chew*]

So this is full on chew* no F's given no lift shifting and not realistic@1.45v....

5.jpg


This is most likely where reality lies.

6.jpg

[muzz]

Well, you just convinced me not to bother with that Little Wraith Cooler, I knew that thing was a joke.
Thanks chew, keep'm coming.

[chew*]

Here's a 2700x at same speeds as the 3700x minus nb/memory. Only reason I have 1 is testing there compatibility on newer boards, seems to be a tad quirky on this taichi.

No clue what these chips are supposed to do as I skipped ryzen refresh due to very little changes but clock/voltage scaling wise its almost identical to the 3700x which is interesting due to 3700x die shrink.

IPC wise its not even a contest however.

7.jpg

[chew*]

So anyway I ran some tests on the 3900 on air

scaling vs voltage required.

1.3v for 4300 air
10.jpg


1.4v for 4350 air
11.jpg


1.45v on water, was a temp issue on air.
12.jpg


Then as always the more things change the more they stay the same. The only magic in getting good scores at a given clock speed is less throttling.
swapping to water and running at 4.3 @ 1.3v on water netted better results than forcing chip to do 4350 on air and 4400 on water.
Simply put find the chips happy place and cool it better.

13.jpg

[chew*]

So this is what I tossed together although with the 3900x it is pretty fruitless and most overclockers are out of a job lol.

Its a swiftech pump/block combo and one of the few that's inlet actually would hit the new core layout first paired with a swiftech aio but its not enough radiator it seems, it seemed like a good idea on paper at least lol.


The chips smart and can do what we can not do which is overclock and underclock on the fly as well as adaptive voltages.

Basically same situation as my 9900k. pointless to OC it so focus on total system performance tweaks and ship it.

under a severe work load


under a less demanding workload

[chew*]

PI is slow........
R1 @ 4.0 for LCC


R3 @4300 honestly probably not worth my time to further investigate. a 300 MHz advantage even win 7 tuned cant make up the difference imo.
I guess look at the bright side, its a fairly fast run for r3 and something I just threw together quick.

[chew*]

timespy and firestrike on the 9900k with the 5700xt.
I will repeat for the tests on the amd chips but first I'm actually going to get seat time in on apex just to see how it compares to the fury x I've been using and get some real world time in on the card.





3700x 5700xt timespy.jpg
3700x 5700xt firestrike.jpg

3900x




3700x





For all those runs try to not to focus on total score but more so on the game tests as I noticed a trend with the 5700 as well.

Oddly enough the 9900k with the 4.7 all core clock and lower res of these benches could not seem to differentiate itself in the game tests and pull ahead.

I'm sure that's not the case always but if and when it can it will most likely be minimal at best. considering the 3700x is a lot cheaper.......enough said.

[chew*]

Now as I said I went and gamed with the 5700xt on the 9900k system as benchmarking doesn't always catch things.

In apex I was hitting a solid 144 in fact everything went smooth except yep always and exception.....I was using overlay to monitor if it was throttling, it did not however 3 hours in the overlay artifacted its overlay window away.

The game itself however continued to play fine. I'm sure that's a driver bug of some sort.

Other observations were well it totally kicked my fury x's azz and maybe placebo effect but I was able to hit targets a lot easier.

3900x 5700




3700x 5700




I'll give the 9900k a run with the 5700 tomorrow after work and game on that card for a bit as well.

So far the trends showing the 3700x edging out the 3900x in the frames ever so slightly like the extra cores are hanging it up just a tad.

9900k 5700


Now we start to see the 9900K slightly edge out in firestrike game tests where I would have expected it to. possibly due to less graphics power = cpu bottleneck.

[bluestang]

What RAM is that you're running and at what volts?

[chew*]

Just some old gskill c15 3600 sku bdie i used on r1 reviews. 1.45v

more importantly its not just something tossed together and bench only stable.

My job is not to show unrealistic benchmark results around this time, I'm solely focusing on real world avg joe end user.

I've actually already deviated from that by slapping water cooling on that's not quite aio but considering the situation unless the temps are being reported wrong.....I cant see anyone taming these cpus without an extremely expensive water setup and anyone who is seeing 70c in CB IS NOT STABLE IN PRIME as that's worth 20C + difference and they will thermal shutdown. crash is a crash..

Only way not to thermal shutdown is let AMD do what they seem to do exceedingly well constant variable voltage and overclocking. The cooler you can run the chip the higher it will oc itself anyway.

I was supposed to be doing stock cooler and maybe an aio.....that ships already sailed. I am in a heatwave right now but not everyone has A/C and everyone has summer so realism is in full effect.

[Charles Wirth]

Good to see you back Brian, my 3700X is a dud and my 3900X is really good.

I am benching on the Hero with the 3900X on air, 4.2Ghz 1.4v 24/7 on stock heatsink. I will update my thread with modified air results.

[chew*]

It will bench fine Charles. It wont prime fine. Running a custom prime test will either crash because not enough volts or enough volts but excessive heat.

Good measure I have seen so far is if you hit 70c in some back to back CB runs your not going to survive prime stability.

IMO no ones buying the 12 core 3900x to game, they are buying it to put it to work thus workload testing it gets...

prime passes stock so if it can't manually oced it is what it is.

If you sat there and maintained say a 40c temp with some ln2 and an evap the cpu would more than likely exceed your ( 1.4v 4.2 OC ) as from what I can see amd's variable auto OC is solely temp based.

The best results your going to get is a really high end water setup with an updated block that hits the off center core.

fwiw I thought dealing with TR full load temps was a chore but this one is a bit more challenging.

Also fwiw in some of the easier tests in prime the 3900x is hitting 60c and running over 4.2 on its own making a fixed 4.2 OC rather pointless. drop them temps to 50c and it starts running 4.4+

[FlanK3r]

issue of Pi is WIndows10. Win10 is very lazzy compared win7 on Ryzens Superpi results. U can check it with older Ryzen on new OS Win10 .
Seesm, in windows 7 are Pinnacle and Matisse clock to clock very similar in SUperpi, maybe new Ryzen is slightly better (but not much). FOr Win7+Matisse I know only use X470 board. I saw also few tips to get x570 for Win7. But...