Official AMD Ryzen 5000 Series & 7000 Series Thread

[pax]

Ya this looks like DeepCool revised their lineup in 2020 at some point. New copper plate and new pump. It beats or matches the Arctic Cooling Freezer 2.

Some 360's beat it but I cant fit that in my old case easily. Can only do uptop and I prefer front position for cool air intake into the rad.

I hadnt looked at coolers in some time until this cpu came long. I think the whole 3d stacking thing is gonna force cpu/gpu cooler companies to upgrade their stuff in the next few years as well.

[bill dennison]

Ya this looks like DeepCool revised their lineup in 2020 at some point. New copper plate and new pump. It beats or matches the Arctic Cooling Freezer 2.

Some 360's beat it but I cant fit that in my old case easily. Can only do uptop and I prefer front position for cool air intake into the rad.

I hadnt looked at coolers in some time until this cpu came long. I think the whole 3d stacking thing is gonna force cpu/gpu cooler companies to upgrade their stuff in the next few years as well.

or force people to go full water

https://www.titanrig.com/watercool-h...102-00-xx.html

[pax]

Im not even sure that would cool well if there was more than one stack... I mean if they stack 3-4 cpus on tops of each other plus a L3 die. The interface between cpu and cooler will have to be rethought... cant just be a flat copper plate at some point.

I look at the AM5 IHS and wonder what all the side openings are as they look like vents.

[Nunz]

What they need to do is stop selling chips with the stupid IHS and just sell them bare-die meant for direct-die cooling. The IHS is a large part of the problem.. once you remove it and go straight to the die, thermal transfer increases dramatically.

It won't happen, but that's the only fix we have right now until there is a new development in how the substrates are designed. I'm not sure how it would work with AMD since it seems no one really de-lids on that side of the pond..

[pax]

I remember putting on the the copper shim for the old athlon in the day so we wouldnt crack the cpu edge while applying the cooler. Something along those lines in a factory installed shim would be nice...

[pax]

Welp the Deepcool is in and max temps dropped from 91c to 83c in geekbench. Pretty big diff. Gonna see about an undervolt. Most drop -0.05 stable.

Also gonna grab some 140 fans for push pull.

[bill dennison]

Im not even sure that would cool well if there was more than one stack... I mean if they stack 3-4 cpus on tops of each other plus a L3 die. The interface between cpu and cooler will have to be rethought... cant just be a flat copper plate at some point.

I look at the AM5 IHS and wonder what all the side openings are as they look like vents.

looks like a socket and very thick IHS

maybe a cooler plate with 8 fingers to hold it like a socket more surface area and to draw heat from the sides

[pax]

There was this one company that had a delid kit and replacement ihs... it claimed 5-8c drops in temp from the wider and heavier copper ihs.

Forget the name of it now tho.

[Nunz]

The thicker the IHS, the worse it will be. Thermal transfer gets worse the further away you are from the heat source.

[bill dennison]

The thicker the IHS, the worse it will be. Thermal transfer gets worse the further away you are from the heat source.

depends on what it's made of

if it's plated copper it won't matter much

and it may not be thicker just taller because of stacking

a water block is nothing but a thick copper plate with fins on the inside

stacking one copper plate on another would not matter much


problem is a lot of IHS are stainless steel i think or some where

[koralis]

Seems like what someone needs to do is have dedicated "pins" in the chip (more like posts) that channel heat away from the silicon to the MB, ultimately to route to the Heat Ssink. MAybe some wiring posts that you can screw from the MB to the CPU block or the case itself?

Gets rid of localized heat (the area immediately around the posts) and routes them to the cooler in a "Safer" way. If nothing else, should spread the heat load over a much larger area and improve the thermals at the IHS layer.

Just spitballing ideas.

[Mangler]

depends on what it's made of

if it's plated copper it won't matter much

and it may not be thicker just taller because of stacking

a water block is nothing but a thick copper plate with fins on the inside

stacking one copper plate on another would not matter much


problem is a lot of IHS are stainless steel i think or some where

AFAIK, both amd and intel use nickelplated copper for their heatspreaders.

[Nunz]

depends on what it's made of

if it's plated copper it won't matter much

and it may not be thicker just taller because of stacking

a water block is nothing but a thick copper plate with fins on the inside

stacking one copper plate on another would not matter much


problem is a lot of IHS are stainless steel i think or some where

Even if it's pure copper, the thicker it is, the further away from the die you are, and that makes the thermal transfer limited by what the IHS can move. The stock IHS from both AMD and Intel should be nickel-plated copper. The die is getting too small and the IHS just can't spread the heat out efficiently enough.

Direct-Die is the way. What used to be the norm in the past will be the norm in the future, I think

[bill dennison]

AMD Ryzen 7000 ‘Raphael’ Desktop CPU Leaks Out: 8 Zen 4 Cores, 5.2 GHz Max Clocks, RDNA 2 ‘GFX1036’ Integrated Graphics Clocked at 2.0 GH

https://wccftech.com/amd-ryzen-7000-...-igpu-2-0-ghz/

[bill dennison]

Even if it's pure copper, the thicker it is, the further away from the die you are, and that makes the thermal transfer limited by what the IHS can move. The stock IHS from both AMD and Intel should be nickel-plated copper. The die is getting too small and the IHS just can't spread the heat out efficiently enough.

Direct-Die is the way. What used to be the norm in the past will be the norm in the future, I think

we are not going to get it .

because of the idiot factor





too many idiots will kill too many dies .


.....

and it's not just one die anymore but 4
one or two CPU dies a controller die and a GPU now

[pax]

Welp the Deepcool is in and max temps dropped from 91c to 83c in geekbench. Pretty big diff. Gonna see about an undervolt. Most drop -0.05 stable.

Also gonna grab some 140 fans for push pull.

Down to 80c in GeekBench now I had the fans on the wrong fan header. They werent ramping up.

[demo]

Does it boost to full speed now? Or are you still waiting on BIOS?

[pax]

It boosts to full speed at 4.55 ghz since we got 1.2.0.7 but it runs a bit hot. Thing is ryzen master has disabled voltage control for this cpu. BIOS as well.

There is one little app that can do it but you have to reapply it everytime you boot.

https://www.overclock.net/threads/58...#post-28975484

PBO2 tuner

I mean temps arent that bad at 80c max with the DC AIO. But woulda been nice to cool it down a bit more.

[KAC]

It boosts to full speed at 4.55 ghz since we got 1.2.0.7 but it runs a bit hot. Thing is ryzen master has disabled voltage control for this cpu. BIOS as well.

There is one little app that can do it but you have to reapply it everytime you boot.

https://www.overclock.net/threads/58...#post-28975484

PBO2 tuner

I mean temps arent that bad at 80c max with the DC AIO. But woulda been nice to cool it down a bit more.

Don't worry about temps. Anything under 90 C is fine. Try running COD Vanguard and then post temps. Game that pushes my 5900X to 85 C routinely.

[pax]

I was used to the 1700x never going above 65c but that chip only boosted to 3.9 ghz with no PBO since one of the early bioses.

Im leaving things as is for now. Some say they wont let us undervolt because the 64 meg chip needs it...

Gaming, WOT and CP2077, seems to hit 70c max but Ill try some other games see how that goes.

[Riptide]

I have this 5950x undervolted with a negative offset of -0.125v core. 64MB L3.

Seems to run OK.

[the_sextein]

This is wccftech,https://wccftech.com/amd-ryzen-7000-...ryzen-5-7600x/

A site that takes MLID seriously because it brings in cash but still. Rumors swirling around of 16/32 core chip with 5.4Ghz single thread and 5Ghz all core with 20% IPC gain and a doubling of L2 cache while maintaining 170 watt package.

Apparently a chinese leaker is saying a 24/48 core variant with slower core clocks may be available as well but if that is true it would only appeal to people who favor heavy multithreading over moderate multithreading like gaming due to the lower core clocks.

I'm moving up from a 9900K so a chip with double the cores and threads with the same all core clock rate and a major IPC gain with the same power draw sounds amazing to me. AMD will be up to date on PCIE 5 and DDR5 so they appear to be aligned to put out a great chip this fall.

If Intel increases their clock rate to 5.8Ghz and improves their IPC by 15% it will probably be able to compete in gaming and adding another 8 small cores should keep it up to date in multithreading. Rumors are saying the adjustments to silicon will allow up to 100 watt power savings over the 12900KS. Thing is, it will be coming after the 7000 series and will probably still be less power efficient while pushing similar performance and features.

I think we might see a repeat of the 5000 series 5950X this time. I'm interested for sure. Both companies look like they are pushing competitive products but I'm happy to see AMD push out their chips first with confidence and no compromises.

[acroig]

AMD will have a 5.4GHz cpu this fall? Would love to see that.

[the_sextein]

The current AMD line up has 4.9Ghz single thread boosts and a 10% clock increase would allow another 500Mhz increase pushing it up to 5.4Ghz. All core clocks are usually 400Mhz slower so 5Ghz all core seems reasonable to me. Still, these are just rumors. I updated my post with a link to the article.

[acroig]

The current AMD line up has 4.9Ghz single thread boosts and a 10% clock increase would allow another 500Mhz increase pushing it up to 5.4Ghz. All core clocks are usually 400Mhz slower so 5Ghz all core seems reasonable to me. Still, these are just rumors. I updated my post with a link to the article.

Thanks for the link.

[pax]

It boosts to full speed at 4.55 ghz since we got 1.2.0.7 but it runs a bit hot. Thing is ryzen master has disabled voltage control for this cpu. BIOS as well.

There is one little app that can do it but you have to reapply it everytime you boot.

https://www.overclock.net/threads/58...#post-28975484

PBO2 tuner

I mean temps arent that bad at 80c max with the DC AIO. But woulda been nice to cool it down a bit more.

Push pull doesnt do squat for this AIO. Back in the day it helped the old CM ML 240 cooler but seems the 2 140 fans push enough air already. 93 cfm they are little tornadoes at load... The 2 new fans are RGB tho so might keep them and same 93 CFM.

3 runs of GB gave the same max temps of 80-82c in push pull as with just 2 fans. Save your money on this. 140 mm fans arent cheap.

[KAC]

Push pull hasn’t done squat for me in ages lol. I am back to push only and don’t give a shite anymore. AMD has taught me not to worry about temps and let it burn.

There is also a certain peace of mind when you know that all your equipment is fully paid up due to mining.

[pax]

https://videocardz.com/newz/asrock-r...yzen-7000-cpus

At ASRock’s YouTube channel one might find video showcasing a new AM5 motherboard as its thumbnail.

The video will premiere in 5 days, which is shorty after AMD’s Computex 2022 press conference where one should expect X670 motherboards. Those are AM5-based series for upcoming Ryzen 7000 desktop processors featuring Zen4 for consumer platform.

The description confirms that the board will use X670E chipset, which is supposedly the premium X670 variant featuring PCIe Gen5 support for both the GPU and storage.

Sounds like pcie5 will be only for the top end.

[Hapatingjaky]

https://videocardz.com/newz/asrock-r...yzen-7000-cpus




Sounds like pcie5 will be only for the top end.

The only reason that PCIE 5 is being pushed so fast is because of storage media, not GPU bandwidth. When PCIE-4 hit it was roughly 2% faster. No GPU has saturated or come close to saturating PCIE-4 bandwidth. So GPU's based on PCIE-5 won't really matter much when it comes to performance differences between PCIE Gens.

[pax]

Wonder how things will shake out in 4-5 years. Always good to future proof but I think 2nd gen mobos will be worth waiting for for am5. Not sure Id pay ~1k for a mobo for the pcie5 privilege tho.

[acroig]

Wonder how things will shake out in 4-5 years. Always good to future proof but I think 2nd gen mobos will be worth waiting for for am5. Not sure Id pay ~1k for a mobo for the pcie5 privilege tho.

I'm sure I will not!

[Mangler]

It would be nice to have those extra pci-e 4th gen nvme slots though.

[pax]

Sweet Mother Mary:


https://www.pcgamer.com/dr-lisa-su-i...ng-am5-boards/

26 stages... dayum.

ASRock has also kinda confirmed it will be showing off X670 boards, too, as Komachi Ensaka has uncovered a private YouTube video via Twitter. This details a host of upcoming Zen 4 motherboards, as well as… excuse me while I swallow back this little bit of bile in my throat… an NFT Edition Intel Riptide board "co-designed by thousands of people."

The video description reads: "ASRock reveals a series of motherboards based on the revolutionary AMD X670 chipset. The new motherboards are equipped with many exciting features and new technology such as PCIe 5.0 & DDR5 capability, the insane IO capability including the support of Thunderbolt 4.0 has pushed the X670 platform even further into high end desktop category.

"ASRock went all out on the VRM design too, with total of 26 phases SPS Dr.MOS design on the X670E Taichi, it is the most powerful AM5 product ASRock can offer

[Nagorak]

I finally got around to putting the 5800X3D in. The whole process went really smoothly. Just flashed the BIOS and swapped the CPU, then set memory to my 3733 CL14 timings and it was stable right from the start.

Only hitch was it wouldn't boot at 3800 or 3867 MHz (previously my 5900X would do 3867 MHz CL15, although at sort of high SOC voltage). I may have put the settings in wrong though. In any case, the difference in performance between those speeds is negligible, especially since the higher one is CL15.

My chip seems to do -30 on all cores with Curve Optimizer. I set it using the PBO2 tool and have tested it pretty extensively now. In terms of performance anything below about -15 seems to sustain max multi-core boost (4.45 GHz) at all times.

On the whole this has been one of the easiest CPUs I've had to "max out".

[Nagorak]

It boosts to full speed at 4.55 ghz since we got 1.2.0.7 but it runs a bit hot. Thing is ryzen master has disabled voltage control for this cpu. BIOS as well.

There is one little app that can do it but you have to reapply it everytime you boot.

https://www.overclock.net/threads/58...#post-28975484

PBO2 tuner

I mean temps arent that bad at 80c max with the DC AIO. But woulda been nice to cool it down a bit more.

https://drive.google.com/file/d/1DL6...gHhAk6AgT/view

There's a new version of this that allows it to be set via command line, so you can set it up to apply at Windows log on. The main benefit is that by lower voltage and temps you should be able to peg the multi-core boost at max frequency, which may improve performance by 5% or so. I think mine was running around 80C stock, and now more like 65-70C after CO adjustment.

Push pull doesnt do squat for this AIO. Back in the day it helped the old CM ML 240 cooler but seems the 2 140 fans push enough air already. 93 cfm they are little tornadoes at load... The 2 new fans are RGB tho so might keep them and same 93 CFM.

3 runs of GB gave the same max temps of 80-82c in push pull as with just 2 fans. Save your money on this. 140 mm fans arent cheap.

AIO radiators aren't really thick enough to benefit from push pull. There's just not that much resistance in them to start with. I did have my watercooled 1080 Ti running in push pull for a while and it did help to the extent that I could run the fans at lower RPM so it was quieter, but in terms of cooling performance there wasn't much difference.

The same thing went for "pressure optimized" versus "airflow optimized" fans in my testing. There just wasn't enough resistance in a thin AIO radiator for pressure optimize fan blade to make any difference in push orientation. Although they did work a lot better in pull orientation, since the airflow optimized fans made a terrible turbulent air noise that sounded like grinding. A lot of that fan stuff is kind of marketing. My conclusion was that in general you're better off with something with a more general blade design, like the Corsair ML series, or a straight-up airflow optimized design, unless you're facing extreme restriction on the output side, or pulling through restriction on the input side. For a thick 2 inch custom loop radiator those factors no doubt play a bigger part.

With your 5800X3D I expect you're facing a heat transfer bottleneck from the heat spreader to the water block anyway, so more airflow through the radiator can't do a lot to help that. That's where running a negative curve offset can be beneficial.

[Nagorak]

I will add that the CPU is fast and it feels really smooth in games. However, it can't work miracles. Poorly optimized games still end up being CPU limited, just at higher level.

[pax]

AMD is coming out with a memory ocing tool. Might improve things inthat dept but sounds like it may be for AM5... we'll see.

https://www.techspot.com/community/t.../#post-1966102

According to the patent, the new technology would alter frequency and specific RAM timings for the RAM kit and hardware combination being used, then test and adjust the settings to ensure stability. The automated stability tests include Error Correction Code (ECC) and other tests designed to detect bit errors. Once the frequency and timing-based tests are complete, the tool generates the new RAM profile using the subtimings and settings specific to that PC's hardware.

Gaming Im max 65c on load. I bought a 3 fan Inwin Saturn with same 93 cfm as oem fans that came with the cooler. So its a 1.5 push pull config lol. Fan curve only gets the fan to ramp up at 70c tho. So its almost always quiet. Temps barely budged doing that. I still get 38c to 45c desktop. I could get it down to 32c desktop but fans would ramp all the time. I left the fans at a minimum up to 70c. The deepcool cooler does a good job but the pump has started putting out a mosquito type buzz... not sure Ill live with that long.

[KAC]

I finally got around to putting the 5800X3D in. The whole process went really smoothly. Just flashed the BIOS and swapped the CPU, then set memory to my 3733 CL14 timings and it was stable right from the start.

Only hitch was it wouldn't boot at 3800 or 3867 MHz (previously my 5900X would do 3867 MHz CL15, although at sort of high SOC voltage). I may have put the settings in wrong though. In any case, the difference in performance between those speeds is negligible, especially since the higher one is CL15.

My chip seems to do -30 on all cores with Curve Optimizer. I set it using the PBO2 tool and have tested it pretty extensively now. In terms of performance anything below about -15 seems to sustain max multi-core boost (4.45 GHz) at all times.

On the whole this has been one of the easiest CPUs I've had to "max out".

Any noticeable gameplay improvements (without looking at FPS counters)?

[Nagorak]

They games I've played on it do feel really smooth, but honestly it could just be placebo.

[KAC]

They games I've played on it do feel really smooth, but honestly it could just be placebo.

Bench and let me know.