Official AMD Ryzen 5000 Series & 7000 Series Thread

[Mangler]

Yah that looks great!

Trying to post a photo here:


What is the issue? They don't seem to be displaying. :/

Direct link works but embedding doesn't display.
https://imgur.com/a/p0IXIa7

I can see even after moving the 360mm radiator out of the system a 4090 might not fit in there even if I wanted one lol
These new GPUs are starting to become a real problem for more common atx cases.

https://i.imgur.com/ewiGR0Y.jpeg

[Mangler]

You know you have a flawed IHS design when companies selling thermalpaste are releasing products like these:

https://noctua.at/en/na-stpg1

The NA-TPG1 is a simple yet highly effective means to preclude thermal paste build-up in the cut-outs of the heatspreader of AMD AM5 CPUs. When the mounting pressure of the cooling solution is applied, excess thermal paste will be squeezed outwards. With AM5 CPUs, this excess paste tends to accumulate in the cut-outs at the sides of the heatspreader and may become difficult to remove. Simple and risk-free to apply, the NA-TPG1 prevents this undesired phenomenon. Combined with the ten included NA-CW1 cleaning wipes, the NA-STPG1 set gives you everything you need to keep your AM5 processor clean.

I think thermalgrizzly is also thinking about releasing something similar.

[Riptide]

Mangler, what did you do to fix the issue with that photo?

I’ll probably order one of those templates before I reapply paste. I’m g2g for now. The issue is a bigger problem for people that frequently reapply the paste/cooler.

[Mangler]

Mangler, what did you do to fix the issue with that photo?

I’ll probably order one of those templates before I reapply paste. I’m g2g for now. The issue is a bigger problem for people that frequently reapply the paste/cooler.

I just right clicked the image in that "direct" link/imgur docking page in order to get the actual address.

[pax]

Any of you guys gonna try the liquid metal maybe apply varnish on the resistors to see if it helps?

[bill dennison]

Any of you guys gonna try the liquid metal maybe apply varnish on the resistors to see if it helps?

not me


and this thing won't do squat
maybe if you sealed it to the IHS with silicone





just don't put your TIM on like your plastering a wall

[pax]

Looks like the ihs isnt that bad asus coming with a bios for setting tjmax and they get better perf at 90c than 95c in CB R23:

https://twitter.com/overclex/status/1578401868694306817

Almost 500 points better at 90c over 95c...

[bill dennison]

Looks like the ihs isnt that bad asus coming with a bios for setting tjmax and they get better perf at 90c than 95c in CB R23:

https://twitter.com/overclex/status/1578401868694306817

Almost 500 points better at 90c over 95c...

here it is without the twitter in non english

Control the temps of your AMD Ryzen 7000-series CPU with ASUS-exclusive PBO Enhancement

https://edgeup.asus.com/2022/control...o-enhancement/


..

MSI had this from the start i was hopeing Asus would add it

[pax]

They show multicore scores but wonder if the 95c one is more about single core perf... 90c would seem to be a good compromise.

[the_sextein]

My 7950 can handle a negative voltage offset of 20. At default I'm getting 37500 in Cinebench R23. With overclocking I hit 38500. I get 95C temps just like the reviews showed. If I set the thermal limit to 75C, it stays at 75C and my score is about 37050.

My mobo only showed support for two sticks of RAM for the kit that I bought but I've had no problems with stability running 4 sticks. I even used the "try it" settings in the bios and managed to drop my timings from C36 down to C30 which provided a little boost according to AIDA 64.

Everything's stable but the boot times are terrible and not just for the first boot either. These boards are slow as can be when it comes to system start up.

[Riptide]

Boot cycles on AM4 were awful at release too. My MSI is slow to POST. Approximately 2 minutes to get through it. No complaints otherwise though.

With same temp target set I am getting nearly identical performance in randomx with a 240 aio vs a 360.

[the_sextein]

Yeah I think AMD just set all of their chips to run at 95C and the only way to lose performance would be the little bit of extra voltage you can get with lower temps. If you lower the temp to 75C the performance difference is tiny so AIO's don't really matter with these.

I'm getting the same 5850Mhz single core and 5250-5300Mhz all core as everyone else. I'm pretty happy with the noise levels on my AIO this time though. I've owned 360mm AIO's for a very long time and this one is much more quiet than any of the others I've used. Even if you max it out it doesn't sound bad. Things just keep getting better.

[Nunz]

I suspect the thermal transfer limitation is between the CPU Die and IHS, which is why the method of cooling (like dropping from 360 to 240mm AIO) doesn't change the temps much despite losing cooling capacity.

If you put the chip direct-die, it would likely be much more noticeable.

Looks like Derbauer/Thermal Grizzly will have their direct-die mount ready in the next ~month. Highly recommend y'all grab it.

[pax]

Yeah I think AMD just set all of their chips to run at 95C and the only way to lose performance would be the little bit of extra voltage you can get with lower temps. If you lower the temp to 75C the performance difference is tiny so AIO's don't really matter with these.

I'm getting the same 5850Mhz single core and 5250-5300Mhz all core as everyone else. I'm pretty happy with the noise levels on my AIO this time though. I've owned 360mm AIO's for a very long time and this one is much more quiet than any of the others I've used. Even if you max it out it doesn't sound bad. Things just keep getting better.

What aio did you get? You mentioned the 420 earlier is it the arctic?

[bill dennison]

Asus finally has the 705 Bios up and they are not batas

https://rog.asus.com/motherboards/ro...helpdesk_bios/

[the_sextein]

What aio did you get? You mentioned the 420 earlier is it the arctic?

I bought the Corsair H150i Elite. Mainly because I liked the pump head design with the customizable LCD screen. If I set it to balanced mode, I can't hear it. Even on extreme mode it sounds like a GPU with it's fans at 50%. Corsair's new fans are very similar to the expensive Noctua fans that I bought on my last setup. I think the Fan controller that came with the H150i Elite is what really makes the difference. The way it handles speed variations and keeps RPM's from spiking during temp variations is key to keeping things quiet.

I pushed my 7950X as far as I could today. I managed to stabilize it with a -25 voltage offset and a 125Mhz clock boost. It hit 5.9Ghz single core and 5.3-5.4Ghz all core. R23 scores were over 39,000. It sits at the same 95C as usual when doing this and the fans are on balanced mode and inaudible. Originally I bought the 170i Elite from Corsair and it's larger 140mm fans probably would have been even more quiet. Unfortunately, the EATX mobo I bought is too large for a case that is designed to be compact. The Define 7 is a little less tall than most other mid tower cases and it didn't work out. The Define 7 is smaller than your average mid tower and very tidy. I loved the design of it and it made wiring my rig with a clean look, very easy. This was the most simple build I've ever put together from both a hardware and software side of things. This will be an excellent platform for the next four or five years.

[pax]

Thanks thats a good review and one bud here did buy a 240 mm corsair and its super quiet as well. Have had a few noisy aio's tho usually its the pump noise that bugs me but there are some with cheap fans. The race to cooler aios is usually met with ever faster pumps but theres a noisy price to pay for that. I def think quiet is worth 5 degrees extra.

[the_sextein]

Updated the Bios and chipset drivers.

Managed -30 voltage offset with 150Mhz boost. Didn't effect the scores much because the voltage limitation from the 95C temps. Better cooling could allow you to squeeze more out of this than I am. A couple cores hit 5.9Ghz. CPU power hit 250 watts. The best score I have been able to achieve is 39,429 so far.



I havn't done any manual tweaking of the RAM yet. I used MSI's pre configured RAM speed of 6000Mhz 30-36-36-76 but otherwise I havn't done anything else. Looks like my latency could use some work and I'm going to look into it. Also, Aida 64 that was provided by MSI doesn't seem to be optimized for the platform yet so it could be inaccurate.

[the_sextein]

Pushed my RAM from it's XMP profile of 6000mhz C36 to 6200Mhz C30. Broke 80,000 mb/s read speeds. Not bad for a 4 stick configuration on a new AMD platform. I tried 6400Mhz with C32 and it failed to boot so I think thats as far as it will go. I'm probably gonna have to look up a guide for manual minor timings to squeeze the latency down. I pushed 175Mhz boost with a -30 voltage offset on the CPU and it completed Cinebench with a score of 39506. Core 1 hit 5925Mhz. If I pushed it up 200Mhz I could get 5950Mhz but that is a far as it will go unless you tweak the second CCD to run at lower values in hopes of boosting the good CCD to 6Ghz. Don't think I'm going to mess with that. So now I have to make the decision of either running the chip at 95C with 39500 multi core performance or run it at 75C at 36500 multi core performance. Either way, it's been fun but I think I'm done pushing the limits of this thing.

[the_sextein]

One last thing, I set all of my fans to quiet mode with my highest OC yet and it stopped at the same 95C and scored 39100 points in Cinebench R23. So if you choose to run it this way you can overclock to the max with fan speeds bottomed out and still get within 500 points of what you had with fan speeds maxed out and at the same temps. Pretty cool and interesting way of handling the Voltages and heat on a CPU and it leaves alot of options for the users to customize to their liking. There is basically no point in running your fans above their lowest setting because it makes no difference in temps and the performance difference is so small it would go unnoticed in any real world application.



Fans speeds on quiet mode never went beyond 800RPM during the R23 multicore test and were completely silent. There was no difference in sound from idle at the desktop to full load during the R23 test and CPU temps stopped at the same 95C as usual. Cores hit as high as 5850Mhz which is basically AMD's designed top out and the performance difference was less than 500 points. Still managing a very respectable 39000 points.

[demo]

Are all [most?] AM4 water blocks compatible with AM5?

[the_sextein]

From what I have heard. AIO's that worked on last gen will work on this gen but specialty coolers for custom loops like EK water blocks will need a new backplate to work on the 7000 series.

[Riptide]

As a rule of thumb, if it worked with the stock AM4 backplate installed on your previous motherboard then it should work with AM5. If it needed a replacement backplate then you need hardware for it to work.

Based on the little testing I've done, with just thermal setpoints in the BIOS, performance suffers at most 4-5% going from Auto (no setpoint) down to 65c.

I'd like to do more in-depth testing to cut power consumption down and maintain near 100% out of box performance but I don't know what I'm doing when it comes to manually tweaking scalars, voltage, etc..

[bill dennison]

EK-Quantum Velocity² D-RGB - AM4 needed a backplate

EK-Quantum Velocity² D-RGB - AM5 they changed the bolts to work with the stock backplate and a T7 TORX to tighten the bolts feom the back side
https://www.ekwb.com/shop/EK-IM/EK-IM-3831109861684.pdf

[demo]

Thanks Bill, that is the exact block I was looking at buying. Can't seem to find the AM5 version for sale here.

[acroig]

Are all [most?] AM4 water blocks compatible with AM5?

The hook ones are, if cooler needs to replace backplate then no.

[bill dennison]

Thanks Bill, that is the exact block I was looking at buying. Can't seem to find the AM5 version for sale here.

it's pre order from EK only i think for now

i ordered a Nickel + Satin Titanium but it won't ship till the 21th

[Mangler]

I bought one of those corsair blocks on sale a month or two ago when I was unsure if I was going raptor lake or ryzen 7xxx a couple of months ago.

They are probably not the best blocks on the market but since I guess I lucked out as it does not really matter for am5.

[Nagorak]

If you're setting a curve offset with Curve Optimizer, especially if going to -30 on all cores, then be sure to test stability very thoroughly. CO can be risky because your system can seem stable in terms of not crashing, but cores may be suffering silent errors in various workloads/boost conditions.

To be confident about stability you really need to test cores individually using something like Core Cycler, but then also test all core since that results in greater power draw/voltage droop and can result in different failures. The individual core testing also takes forever on 12 and 16 core processors. It's honestly questionable whether it's worth doing, in my opinion.

[Nascar24]

I just recently used CO and just had ryzen master give me the number. Was -27 all core. Crashed the next day so I backed it to -25. Hasn’t crashed since. If it crashes or causes any other issue I will just back it off a couple more.

[andino]

I wonder what effect a reverse die water block would have on that IHS. Like a AIO or water block that could encase that funky IHS. It would add cooling inside of those funky crevasses on the sides.

[Nunz]

Just get rid of the terrible IHS that AMD designed and buy a block that is direct-die itself.

Waiting to see what supercoolcomputers comes up with. They had one for AM4

[pax]

People might want to wait a bit to jump in as sales arent great and we could get some great deals by Black Friday. 7950x and 7700x are selling pretty well but not the 7600x and 7900x oddly.

As for cooling Id rather lap the thing for hours to shave a couple mm off the top than delid. Be nice to see some tool from Derbauer just for that so you keep the lapping even.

[the_sextein]

I think I'm going to settle with a thermal limit of 85C. At this setting I'm getting 5170Mhz on CCD #1 and 5050Mhz on CCD # 2 under 100% load of all cores. Single core performance is generally 5.5Ghz with small spikes up to 5.7Ghz at times. I can leave my fans on balanced or even quiet mode all the time, doesn't really matter. It's technically an underclock and unquestionably stable. Scores about 38000 points in R23 at these settings and most games run at 5.5Ghz on all utilized cores.

My I9 9900K at 5Ghz all core would hit 86C in R23 unless I maxed out the Custom EK water loop that I had and it not only sounded like a jet engine but temps were still around 80C. Prime 95 would run at 95C with the loop maxed out and I didn't have the balls to try running it at a lower setting

Running the 7950X at an 85C thermal limit gives me nice and quiet operation without any worry of going over 85C under any load including AVX512 like DAZ studio or prime 95. It's averaging above 5Ghz with double the cores, quadruple the cache and a nearly 30% IPC increase while running more quiet and slightly lower power and temps with a more simplistic cooler. Can't complain about that.

I did experiment a little today though. In the PBO menu next to the thermal limits there are four performance modes. Each one bumps the single core boost max by 50Mhz. So performance mode #4 pushes single core performance to 5950Mhz. Pretty much 6Ghz and all core load is around 5.3Ghz across all cores. I didn't have any problems with stability after hours of prime 95 and a few hours of DAZ work but I just don't like the constant 95C temps this things is pushing. I'm sure it's fine but I don't like it.

One thing I do consider a negative about this chip is it idles at 50C and plays games at 70C. Considerably more hot than my 9900K but I'd rather have the extra heat at normal operating levels than at the breaking point.

I also lowered my RAM back down to 6000Mhz. I could push it to 6400Mhz with XMP settings or 6200Mhz with much tighter timings but AMD suggests that going over the 6000Mhz mark can cause instability so I've backed off for now and just pushed my timings as tight as I can. I'm doing hours of content creation that requires stability above raw performance and most of these overclocks make no difference in gaming what so ever so I think the 85C thermal limit is a good compromise for me. It's an a$$ hair faster than stock while running 10C below stock settings. Seems like a no brainer. 75C was tempting but CCD#2 drops down to 4.9Ghz under full load and I don't average 5Ghz all core or reach the multi core performance that was advertised. I'll admit it's pretty close but I think I'm going to pass on the 75C thermal limits.

[the_sextein]

When it comes to games like Cyberpunk. The chip runs all 16 cores at 5.5Ghz for a few seconds and then it will drop to 4.5Ghz on half the cores while maintaining 5.5Ghz on the other 8 for a few seconds before returning to all core 5.5Ghz.

What it's doing is it's cycling the 8 primary threads that the game is using so that they are always at 5.5Ghz while letting the other 8 cores cool down and then swapping them back and forth to keep temps in check. Pretty cool, temps are about 73C in Cyberpunk. Total CPU usage is around 20% of the 7950X at any given time. Worst case scenario would be if all 16 cores were hit with similar loads while half of them were down clocked then the chip would average 5Ghz all core but most of the time it seems to be around 5.5Ghz in the games that I have observed.

[Mangler]

Cool, the 20% rebate on my X670E strix board resulted in it being cheaper than a B650E strix, that is kind of ****ed up, asus has gone overboard with their PREMIUM pricing this gen.

It really looks like those who want good value for their money needs to go Intel this time, b650 and x670 are just too expensive for budget minded people.

[Megaman]

We have too many new tech out at the same time. We are being raped with premiums.

[Riptide]

Are you guys running your DDR5 modules at command rate 1T? It seems that the XMP profile at least defaults it to 2T for me.

Is that typical?

Kingston CL32 PC6400 M/T
Hynix-M

[Nunz]

Yes 2T is standard for that kit. There's no kits running 1T off XMP. It's do-able, but don't know how AMD is doing. Really depends on motherboard quality, at least for Intel.

You could try it and see what happens. I doubt it'll POST.

[the_sextein]

I'm using T1 on my kit. I need 64GB of RAM for the work that I do. My current projects use 50GB regularly. I couldn't find any 32GB sticks at 6000Mhz so I bought 4x 16GB sticks to do the job. Using 4 sticks instead of two can increase read and write speeds but it makes it much harder to push clock frequencies up higher and it also hurts latency considerably. My DDR 4 RAM could reach 4000Mhz but when I added another two sticks I could no longer push beyond 3600Mhz.

I posted results on the last page that shows 82000MB/s read and a latency of 72ns. In T2 mode my latency is 82ns.

Like Nunz said, it has alot to do with your motherboards power delivery and also the chip lottery on your RAM. I'm was running mine at 1.4V when I was pushing 6200Mhz 30-36-36-76 and my motherboard is way over engineered when it comes to the power delivery which may have contributed to it. I've seen my latency drop down to 68ns during some of my testing but it's never went lower than that and I've seen 5600Mhz 2 stick kits hit 62ns at stock which kind of sucks for me. My read and write speeds are through the roof though. My RAM is Corsair Dominator Platinum (Hynix).