Who's ready for the new incoming AMD CPU/GPU hardware?

[chew*]

Hi

Ryzen 5000 FTW! Amazing IPC performance and best price but hard to mem OC
I have heard AMD RX6000 has lower image quality in games! is that so?!

vs 5xxx? I see no visible difference, maybe it would require screenshot comparisons on same drivers. I like my eye candy though and have a high attention to detail and have not noticed anything "off".

PSU be here monday. It's only an ax850 but that fits the criteria for real world as most won't need more than that anyway and it should already be more than enough power even with ambient "total platform tuning" as I prefer to call it vs overclocking now.

Arctic liquid freezer II 280 showed up ( my dime ) also picked up an open box ITX ASUS ROG X570-I open box to play with ( my dime ). Wall Mount came in for monitor ( my dime ) A retail 5800x ( my dime ) is on the way to bin as well although i think the chip that really need replacing may be the 5900x. It would be nice to find a 5.0+ capable 5800x though.

Something is arriving from EK on tuesday so lets add them to the special thanks list. I honestly have no clue what it is ( third party via AMD )

May change up the hard drive config but undecided yet as I don't want to run 2 m2's and starve bandwidth on one of the x16 slots.

Things left to get are the phanteks p500a case ( black non rgb or rgb or white rgb doesn't matter it's just for airflow/realworld rgb looks cool in moderation but does not cool ) and a 5 pack of p14 artic fans. I may add one set of memory to the list but i'm not quite sure it's needed except to simplify side by side as this set of 32gb 16-16-16 3600 can pull the 3600 14-15-15 @ 1.45 spec just fine as well as up to 2000 16-16-16 @ 1.45

[BeepBeep2]

I've come to terms with the fact that CH6 Hero is probably not the ideal motherboard for overclocking memory being a T-topology board and the VTTDDR voltage is adjusted in 0.0066v steps while VDIMM is adjusted in 0.005v steps. Not sure who thought that was a bright idea. Basically only 1.325, 1.35v, 1.380v, 1.4v, 1.425v are functional at all at 3600 MHz+, and more VDIMM above 1.35v is just progressively worse no matter what I try.

I can pass 75% coverage HCI memtest at DDR4-3800 16-16-16-32-48 1T tRFC 352 1.35v VDIMM/.6732 VTTDDR, could do 70% even up to 1919/3838 (101.0 BCLK) but can't actually get it stable whatsoever because any choice of higher VDIMM and VTTDDR is less stable / fail between 2-35% coverage.

CLDO_VDDP is happy at 0.80, 0.90, 1.0...all perform exactly the same and slightly less happy at 0.85, 0.95, everything else is a memory hole @ 1900, nothing changes from 0.80 to 1.0.

CLDO_VDDG IOD needs 0.90/1.00/1.05v for 1900, can run CLDO_VDDG CCD at 0.80, 0.90, 1.0, or 1.05. Anything else +- 15mv creates a memory hole for all of these voltages and help really from going 0.90-1.05v except linking the two at 1.05v is marginally better. 1.1v with 1.15v SOC is worse.

SoC voltage likes 1.10v with LLC 1-2-3 but don't really see any difference between them with vdroop anyways. Boots 1.15v but worse, anything above is no POST. Anything except 1/1.05/1.1/1.15v creates memory hole so I have four separate voltages to balance that all create memory holes if set 0.0066v too high or low

I can drop down to 3600/3666/3733 dividers and have the same issue.

Can't do GDM off at all, can boot 3800 CL14 but unstable even up to 1.55v, can't even get 3600 CL14 stable. Tried screwing with setup times and drive strengths, basically changed nothing.

I've given up on trying to get 3800 CL16 working at this point and guess I will try to drop down to 3733 16-16-16-32 1.35v since the VTTDDR/VDIMM mismatch is killing everything.

I'm pretty sure if I plopped this CPU in a new board it would do 1900 all day.

[chew*]

not necessarily. It's very board and chip dependent. my master v1.0 did 3800 2x8 , ch8 3733 same chip memory. taichi 3800 till later agesa's then 3733.

Bios and agesa matter to. pre 3003 bios and on hero i was doing 1900 NP with my 5900x and now i'm not.

[BeepBeep2]

not necessarily. It's very board and chip dependent. my master v1.0 did 3800 2x8 , ch8 3733 same chip memory. taichi 3800 till later agesa's then 3733.

Bios and agesa matter to. pre 3003 bios and on hero i was doing 1900 NP with my 5900x and now i'm not.

Interesting.

My board, RAM, CPU and I had a talk and decided to do the opposite of the rest of the internet. Apparently my IOD is special. Failed at 95% coverage this time, so making progress. I couldn't get 3666/3733 stable either so I started working on 3800 again.

No matter what I set the board won't complete DRAM training at POST above 3800 even at 1:2 FCLK so I assume the T-Topology has something to do with that vs. every other board on the market being daisy chain idk. Tried every VDIMM boot voltage between 1.3-1.5v

SoC 1.000v LLC 1 (0.981 w/ vdroop)
CLDO_VDDP 0.800v
CLDO_VDDG 0.815v CCD/IOD
VDIMM 1.380v

The difference between this and not stable at all is razor thin. Gonna keep going lower on VDDG and VDDP in 0.005v steps, if I can eventually pass 400% I'm going to call it good enough. SoC power dropped from 25w to 17w under load which should help PBO boost in the end as well.

[chew*]

07 post code = fclk wall which the vast majority of chips hit @ 3800 or 3866 nothing to due with T topology it's the typical fabric limitation. drop the 1/1 ratio and you can do 5000+ memory.

PBO is usually almost always limited by EDC which if i run prime always bangs 100% of 200 A while TDC and PPT are usually in the 50-60% range, at least that's the case when i run the 16 core chips.

I've got the liquid freezer II 280 installed now, trying to close the clock gap on min max avg and shift it higher on 5950x now.

testmem5 works with anta extreme it's harsher than hci and faster. Found errors with it that even 2000% coverage in hci did not.

Case is ordered ( ended up with black P500a non rgb as it's what became available ) , just need to order a 5 pack of p14 fans probably later on today.



This is just a baseline run with PBO +100 and manual set current/watts no curve optimizer yet

[BeepBeep2]

07 post code = fclk wall which the vast majority of chips hit @ 3800 or 3866 nothing to due with T topology it's the typical fabric limitation. drop the 1/1 ratio and you can do 5000+ memory.

PBO is usually almost always limited by EDC which if i run prime always bangs 100% of 200 A while TDC and PPT are usually in the 50-60% range, at least that's the case when i run the 16 core chips.

I've got the liquid freezer II 280 installed now, trying to close the clock gap on min max avg and shift it higher on 5950x now.

testmem5 works with anta extreme it's harsher than hci and faster. Found errors with it that even 2000% coverage in hci did not.

Case is ordered ( ended up with black P500a non rgb as it's what became available ) , just need to order a 5 pack of p14 fans probably later on today.



This is just a baseline run with PBO +100 and manual set current/watts no curve optimizer yet

...

Tried 1/2 FCLK and still couldn't train DRAM between 3866-4800 CL18-20.

This post by elmor kind of confirms what I'm seeing. I believe it is a board limitation, he was only able to hit 3666 stable 2x8GB w/ B-die: https://www.overclock.net/threads/ry...-x570.1728878/

I'll look into other memory tests. Passed 3600 CL16 1.35v 1800% HCI as a sanity check last night, 3666+ fail all around the same point 70-100% HCI coverage. Now working on tuning 3600 CL14 since I can't get the higher dividers stable.

[demonkevy666]

Chew* do you meant it can vary just by board by board, Like same exact board, just some work better than others?
or is it a combination cpu's and boards liking each other ?

[chew*]

Chew* do you meant it can vary just by board by board, Like same exact board, just some work better than others?
or is it a combination cpu's and boards liking each other ?

I think it boils down to board tuning trace and bios wise. The board I was referring to that pulled 3800 was a aorus master v1.0 with 2x8gb it pulled 3800 on every chip i tested 100% stable.

Swap to DR and it was not stable at all but could boot. then they revised board like 2 more times. makes you go hmmm..

Asrock on older agesas also pulled 3800 then later 3733. the chviii pulled 3733 from launch bios to current with the same 3000 series cpu. I tested every version.....just to be sure.

Both the Asrock and the ASUS were stable with 3733 DR setups as well though.

So i guess it's a combination of alot of things. If you sacrifice something I guess you can get certain things stable or you can tune the board for a majority of configs while maybe losing something.

I gave the gigabyte away for the overall compatibility with SR and DR on CH8 as it had more options available that I could manually set if I needed even though it hit lower speeds.

Right now I can't test vendor to vendor variance as far as board tuning goes as I have 3 ASUS boards all fairly similar well except the 570 I but that should run similar to impact and I wouldn't exceed over an 8 core part in it when testing say prime.

I might pick up an Msi b550 unify X for my vendor to vendor cpu's behave differently testing but i'm not 100% sure I will as I've been waiting to see what other boards might be built electrically for 5000 series.

[BeepBeep2]

I've boiled down my issues to VTTDDR mismatch with VDIMM. This board does not allow the two to match up in a way that most VDIMM settings actually work. Was tipped off by this post at overclock.net https://www.overclock.net/threads/ro.../post-26547301

3600 CL14-15-14-xx boots at 1.4-1.425v but unstable. The RAM itself needs higher than this to be stable. 1.45v is close to stable but again fails between 60-150% in HCI even at 3600 CL16-16-16 with high tRFC when 1.35v passed. Temps on each stick are below 50c when failing but I think heat is working against me as well when on the edge. Everything in between 1.425-1.45v is less stable and everything from 1.455-1.525v is less stable yet, but gets marginally better or worse depending on VTTDDR selection.

1.35v is my best bet, so back to CL16 and working out how high I can go or if I need to stay at 3600 with loose secondaries. Another thing for me to look at is BCLK, I can do exactly 101 before I lose boost but I have an audio latency issue at that freq.

I might end up dumping this board and picking up an X570 board in the next few weeks but I'm not sure if it's worth forking over the cash.

I'm going to try CL18 at 3800 with loose timings and low VDIMM to see if I can get that stable for kicks. Might be able to swing 3800 16-17-16-xx at 1.35v if that is the case. This kit has a little more headroom with tRCD = tCL+1.

[Mehdi_FXX]

vs 5xxx? I see no visible difference, maybe it would require screenshot comparisons on same drivers. I like my eye candy though and have a high attention to detail and have not noticed anything "off".

PSU be here monday. It's only an ax850 but that fits the criteria for real world as most won't need more than that anyway and it should already be more than enough power even with ambient "total platform tuning" as I prefer to call it vs overclocking now.

Arctic liquid freezer II 280 showed up ( my dime ) also picked up an open box ITX ASUS ROG X570-I open box to play with ( my dime ). Wall Mount came in for monitor ( my dime ) A retail 5800x ( my dime ) is on the way to bin as well although i think the chip that really need replacing may be the 5900x. It would be nice to find a 5.0+ capable 5800x though.

Something is arriving from EK on tuesday so lets add them to the special thanks list. I honestly have no clue what it is ( third party via AMD )

May change up the hard drive config but undecided yet as I don't want to run 2 m2's and starve bandwidth on one of the x16 slots.

Things left to get are the phanteks p500a case ( black non rgb or rgb or white rgb doesn't matter it's just for airflow/realworld rgb looks cool in moderation but does not cool ) and a 5 pack of p14 artic fans. I may add one set of memory to the list but i'm not quite sure it's needed except to simplify side by side as this set of 32gb 16-16-16 3600 can pull the 3600 14-15-15 @ 1.45 spec just fine as well as up to 2000 16-16-16 @ 1.45

Not vs 5000, vs RTX
actually I dont know the fact. I tested two 3080s weeks ago and took some shots and i'm waiting for RX 6000 to make comparison

[chew*]

Not vs 5000, vs RTX
actually I dont know the fact. I tested two 3080s weeks ago and took some shots and i'm waiting for RX 6000 to make comparison

Ahh, 3080 is a mixed bag due to dlss and about 3-4 different settings within like say cyberpunk. I'd probably shut all that crap off and compare native to native.

[BeepBeep2]

Passed 1000% HCI last night at 1900 FCLK. Looks like the IOD / mem controller on this CPU are actually excellent, these voltages are the lowest I've seen. I guess somewhat luckily, stability for SOC/VDDP/VDDG influenced results even with bad VTTDDR so I settled on the same voltages for those as before.

DDR4-3800 16-17-16-32-48 tRFC 342 (180ns) GDM on @ 1.360v manual VTT set 0.6798
1.000v SoC LLC 1 (no LLC)
0.800 CLDO_VDDP
0.815 CLDO_VDDG CCD and IOD

Triple confirmed at DDR4-3600 16-16-16-32-48 that 1.35v and 1.36v VDIMM are the only voltages I can actually use because of the VTTDDR steps.

Kind of infuriating but then again this board was designed around CPUs that could only do 3200-3466 anyway I guess. My 1600X couldn't do more than DDR4-2933 stable at any SoC voltage on this board and now I'm thinking the VTTDDR issue might have been why.

Hand tuned subs. Might try to tighten these up further but not much left to gain. Will see what affect going lower on tRCDWR, tRDRDSCL, tWRWRSCL, and tCWL has on performance and maybe try to bring tRFC down a little more.

Moving to TestMem5 with anta777 extreme preset.

[BeepBeep2]

Passed 3 cycles of TM5. Might change config to 10-20 and test overnight. Gonna see if I can get latency down further, unfortunately a bit limited at max 1.36v

[chew*]

try pulling in your trfc your really loose for 2x8, I'm running 304 with 2x16gb you should be able to do 298 or less easy.

[BeepBeep2]

try pulling in your trfc your really loose for 2x8, I'm running 304 with 2x16gb you should be able to do 298 or less easy.

Yep had it loose while I worked down the rest just to be safe. Brought it down to 304 last night. Tried 285 for 150ns, won't POST but I can try something in between.

Dropped tRCDWR to 8 and tCWL to 14 as well which bumped writes and copy bandwidth significantly, and latency is now under 64ns. tCWL 10/12 was no POST and tRDRDSCL/tWRWRSCL couldn't get below 4 - these both run 2 easy with more voltage but I'm limited to 1.36v

Was seeing weird behavior in Cinebench R20 I assume was clock stretching(???) at 1.00v SoC and a little at 1.05v so bumped it to 1.10v. Score went from 7050-7250 back to expected 7450-7500 with PBO same clocks in HWInfo.

My copy bandwidth is still super low for dual CCD CPU but must be a product of board/AGESA... going to try earlier BIOS version, low copy bandwidth happens at every divider and timings. Scores look good for CL16 otherwise

[chew*]

Turbocool Leaf blower 1200 finally removed from the test rig. Special thx to corsair and my contact there. You know who you are.



Got some EK stuff to test with the new AMD hardware. Maybe we can get rid of that cpu overtemp issue that I got rid of on everything but the 5950x with the liquid freezer 280. 5950x is just to much for that AIO it seems.

My ears are totally thanking corsair right now. It was so loud in here with the dinosaur power supply I couldn't think straight. The loudest thing in this room now is the rather dated version of the AX1200 in my gaming PC and thats barely audible.

https://streamable.com/zpxt5p

[demonkevy666]

Turbocool Leaf blower 1200 finally removed from the test rig. Special thx to corsair and my contact there. You know who you are.

Got some EK stuff to test with the new AMD hardware. Maybe we can get rid of that cpu overtemp issue that I got rid of on everything but the 5950x with the liquid freezer 280. 5950x is just to much for that AIO it seems.

My ears are totally thanking corsair right now. It was so loud in here with the dinosaur power supply I couldn't think straight. The loudest thing in this room now is the rather dated version of the AX1200 in my gaming PC and thats barely audible.

Are those blocks Copper on the inside with no nickel Plating ? I'm so tired of nickel plated everything.

chew*(Brian) your voice has gotten deeper over the years O.o!

[zanzabar]

Are those blocks Copper on the inside with no nickel Plating ? I'm so tired of nickel plated everything.

Those come in copper only

https://www.ekwb.com/shop/ek-quantum...d-copper-plexi

Most things come in copper only as an option, but I would get acetal (delrin) and nickle on anything new I buy. Cleaning copper is such a pain in the ass and the cost is basically the same.

[demonkevy666]

Those come in copper only

https://www.ekwb.com/shop/ek-quantum...d-copper-plexi

Most things come in copper only as an option, but I would get acetal (delrin) and nickle on anything new I buy. Cleaning copper is such a pain in the ass and the cost is basically the same.

I like full copper my self. yeah it is a pain to clean. Something rough to take off oxidization helps, fine sand paper for base plates. I forgot what my father told me to clean the internals with a while a back I thought it was a write brush wheel maybe or something. :-/

[chew*]

vinegar. anything acidic will make it shine when pumped through it.

Test rig is mocked up and in a case now. Still need to "clean" things up a bit (braided cables/accent lighting) and get the proper fans in it when it all arrives.

[Shocker003]

vinegar. anything acidic will make it shine when pumped through it.

Test rig is mocked up and in a case now. Still need to "clean" things up a bit (braided cables/accent lighting) and get the proper fans in it when it all arrives.

clean setup

I am still using my old Coolermaster comos S.
I will be happy to upgrade to a case with at least one optical bay and option for a 360 rad at the top of the case.
i can?t wait anymore for 5900x so placed an order for Ryzen 7 5800x as upgrade for my aging 1700x.
I will be glad if someone can point me towards a decent pc case Without RGB and thanks in advance.

[chew*]

clean setup

I am still using my old Coolermaster comos S.
I will be happy to upgrade to a case with at least one optical bay and option for a 360 rad at the top of the case.
i can?t wait anymore for 5900x so placed an order for Ryzen 7 5800x as upgrade for my aging 1700x.
I will be glad if someone can point me towards a decent pc case Without RGB and thanks in advance.

This fits a 360 front and top but does not fit the optical drive criteria. Most modern cases won't fit that criteria.

Took me a while to finally ween myself off my blue ray player and ditch optical bays. You can can always go the external route as that's what most people who still need one do.

Cooling wise I think you will like the 5800x. Easy to tame temps and maximize clocks. Price wise ehh I've already explained my personal opinion on that and its solely based on the rest of the product stack and those products pricing.

My cooling requirement breakdown is this

5600x 240mm aio
5800x 280/360mm aio
5900x custom loop 360mm min
5950x custom loop 360+

This is of course if you want to get the max potential out of the cpu's and just my personal opinion having used them.

Case does look like it has a tad to much room but room to grow is always a criteria. I'll give you idea where this is going.

Keep in mind it's just a test rig so It needs to fit and accommodate a large range of configurations.

https://streamable.com/pdqi1h

[Shocker003]

This fits a 360 front and top but does not fit the optical drive criteria. Most modern cases won't fit that criteria.

Took me a while to finally ween myself off my blue ray player and ditch optical bays. You can can always go the external route as that's what most people who still need one do.

Cooling wise I think you will like the 5800x. Easy to tame temps and maximize clocks. Price wise ehh I've already explained my personal opinion on that and its solely based on the rest of the product stack and those products pricing.

My cooling requirement breakdown is this

5600x 240mm aio
5800x 280/360mm aio
5900x custom loop 360mm min
5950x custom loop 360+

This is of course if you want to get the max potential out of the cpu's and just my personal opinion having used them.

i already have a custom 360 rad loop just for cpu cooling, running in my sig system.
it is hard to let go or discard my bluray writer/burner. I had to settle for the 5800x after reading about thermal problems on the 5950x and 5900x. I will be playing cold war and apex legends on the rig. i woun?t be streaming and no content creation.
Thanks once more for you advice. I will still be on the look out for a case.

[zanzabar]

I like full copper my self. yeah it is a pain to clean. Something rough to take off oxidization helps, fine sand paper for base plates. I forgot what my father told me to clean the internals with a while a back I thought it was a write brush wheel maybe or something. :-/

I specifically could not clean the skived portions of my EK evo supremacy so it is not useless as the fins are clogged, but anything I do now will bend or destroy them. The nickle ones you can use some vinegar then a pressure washer or sonic cleaning bath to clean. I am going to try a sonic cleaner on the copper block but dont have high hopes.

[ZenEffect]

electrolysis is another non-abrasive option