zen 4 AMD Ryzen 9 7950X breaks four HWBOT world records with AIO cooler

[BiG StroOnZ]

Summary

AMD revealed some impressive overclocking results of its own today. Using "standard liquid cooling", a 280mm AIO, the company's XOC team was able to push a Ryzen 9 7950X as high as 5.5GHz all-core and complete several benchmark runs. In total, four records now been broken by Sampson and blueleader who got a green light to share their Ryzen 9 7950X results.

 

 

 

 

 

Quotes

Quote

Using either AORUS X670E Xtreme, ASUS ROG Crosshair X670E Extreme, ASRock’s X670E Taichi or MSI X670E MEG ACE motherboards and Corsair H115i liquid cooler AIO cooler, both overclockers managed to break four records: Cinebench R15, R20, R23 and 7-zip.

 

The CPUs have been running at 5.4 to 5.5 GHz with all cores with temperatures between 87°C to 108 °C if Benchmate data is to be considered. The data also shows that Ryzen 9 7950X has been running with 226W to 244W max power, so within the maximum allowed range for AM5 socket.

 

Ryzen 9 7950X world records:

 

Cinebench R23 – 40,498 pts nT

• 5.40 GHz—ASROCK X670E Taichi—32GB Kingston DDR5

 

Cinebench R20 – 15,771 pts nT

• 5.35 GHz—GIGABYTE X670E AORUS Master—32GB G.SKILL DDR5

 

Cinebench R15 – 6,900 pts nT

• 5.50 GHz—ASUS ROG Crosshair X670E Hero—16GB G.SKILL DDR5

 

7-Zip – 228,992 MIPS

• 5.45 GHz—MSI MEG X670E ACE—32GB Corsair DDR5

 

My thoughts

This is quite impressive, as none of these records were broken using LN2 (Liquid Nitrogen). What's also impressive was the fact that there was a CPU-Z Validation at almost 5.8GHz all-core and only at 1.4v. It's a possibility that the 5.8GHz all-core overclock is a golden sample, but it could also mean Zen 4 overclocks well judging by these HWBOT results. As far as the records go, if this is what Zen 4 is capable of doing with a simple AIO CLC, I'm wondering what it's able to achieve with more exotic cooling.

 

Sources

https://videocardz.com/newz/amd-ryzen-9-7950x-breaks-four-hwbot-world-records-with-aio-cooler

https://hothardware.com/news/ryzen-7950x-zen-4-cpu-hits-impressive-overclock

[StDragon]

Damn!!! AMD came out swinging hard Too bad the MBs are expensive; no wonder the CPUs are priced lower than initially thought.

[Beerzerker]

I can promise you all chips used for these runs were golden samples - Esp if those guys are "In House" OC'ers and they are, being they are on the company's XOC team.

[porina]

From an overclocker perspective, are these even overclocked? The R20 result isn't even at PPT limit. Temps are already pretty high so these are looking like hot chips as is traditional for Ryzen. So on one hand, I think at least the R20 result might not be far off stock, and there is much more to get out of it. On the other hand, cooling for best performance is going to be challenging even at stock.

[starsmine]

These dont look overclocked at all?

[Shzzit]

108c ?  Haha not good.

[Doobeedoo]

Cool. Excited to see game tests eventually though. Really especially 3D cache ones when they're out. I just hope those can retail clocks and temps.

[WereCat]

6 hours ago, Shzzit said:

108c ?  Haha not good.

For long term probably not good but it may still last years. 

For short term it's a non issue really. 

[Fasterthannothing]

6 hours ago, Shzzit said:

108c ?  Haha not good.

Yeah exactly anyone who looks at that temperature and says that thing overclocks well is just being a disingenuous fanboy. Nothing about a 108c temp screams good overclock. 

[ZetZet]

10 hours ago, porina said:

From an overclocker perspective, are these even overclocked? The R20 result isn't even at PPT limit. Temps are already pretty high so these are looking like hot chips as is traditional for Ryzen. So on one hand, I think at least the R20 result might not be far off stock, and there is much more to get out of it. On the other hand, cooling for best performance is going to be challenging even at stock.

 

9 hours ago, starsmine said:

These dont look overclocked at all?

7950X all core boost is 5.1 GHz. 

[porina]

3 minutes ago, ZetZet said:

7950X all core boost is 5.1 GHz. 

Remind me, how does AMD specify all core boost, if at all? Is that the absolute maximum clock that will be hit when all cores are loaded? Or is it just a typical value?

 

Since Zen 2 (maybe even Zen+ but I can't remember that clearly), boost clocks have been opportunistic according to load, power, current and thermal limits. For most Zen 2/3, you can get around thermal limits with sufficient cooling, which leaves the power limit as the primary factor to how far it boosts. So I'm used to seeing the power limit being the one to look for. PPT for Zen 4 would fall around 230W, and some of these benches aren't reporting that. What I failed to consider is that it looks like thermal limit is now the bigger factor.

[leadeater]

13 minutes ago, porina said:

Remind me, how does AMD specify all core boost, if at all? Is that the absolute maximum clock that will be hit when all cores are loaded? Or is it just a typical value?

From memory typical value based on highest achieved under light computational type loads. Zen 3 had maximum all core spec but you didn't achieve it on more demanding workloads.

 

As you said it's power limit all the way and not much else really. Power and heat go hand in hand, so still the same thing really, other than bolting on epic cooling and maximizing thermal transfer, which has to transition to things like delidding etc quite quickly to really do anything.

[ZetZet]

16 minutes ago, porina said:

Remind me, how does AMD specify all core boost, if at all? Is that the absolute maximum clock that will be hit when all cores are loaded? Or is it just a typical value?

It's probably some average on some workload, still 5.35GHz and 5.45GHz are both over what the CPU does by itself. 

[starsmine]

Its not just the CPU
The ram speeds are both all over the place, and slow. 
 

52 minutes ago, ZetZet said:

 

7950X all core boost is 5.1 GHz. 

Sure, Which is expected behavior at stock
But I half expect professional over clockers to get close to single core turbos on most cores with quick and dirty OC.

I may be very much in the wrong, but this looks like they did the minimum OC necessary to break the records and AMD are just sandbagging again by letting this leak, why are all those ram kits all over the place otherwise? Like they were all instructed to just 1up the record, dont let the chip show off what it has.

Like that r23 score, they DOWNCLOCKED the ram from 5600 to 4800MT with very loose timings and downclocked the zen4 uclock to 2400mhz when its supposed to be 3000mhz?

Im wondering if that also explains the temps as they just put them on "fuck it" voltages to do a stable OC run in under 5 min

[ZetZet]

13 minutes ago, starsmine said:

I may be very much in the wrong, but this looks like they did the minimum OC necessary to break the records and AMD are just sandbagging again by letting this leak, why are all those ram kits all over the place otherwise? Like they were all instructed to just 1up the record, dont let the chip show off what it has.

NDA drops in 4 days. We will see then. 

[porina]

18 minutes ago, starsmine said:

The ram speeds are both all over the place, and slow. 

Cinebench family is minimally affected by ram performance. That's why it's been preferred by higher core count benchers to demonstrate their potential.

 

I'm not so familiar with 7zip benching but I'd expect that to depend on a combo of cache size and ram perf somewhat.

[Ultraforce]

Related to the overclocking Gamer's Nexus has an interview with AMD's overclocking team when they were doing some liquid nitrogen stuff.

 

[starsmine]

5 hours ago, Ultraforce said:

Related to the overclocking Gamer's Nexus has an interview with AMD's overclocking team when they were doing some liquid nitrogen stuff.

 

>48k r23 score. which, damn. 20% better then a record 12900k and its just the start of the generation

[porina]

Writer of Y-cruncher benchmark had early access to Zen 4 and just released his observations along with everyone else. Part of it explains what we saw in these benchmark results.

 

https://mersenneforum.org/showthread.php?p=614191#post614191

 

Quote

However, the chip as a whole is very difficult to cool due to the dies being so small. So the bottleneck is the contact area between the die and the IHS. For all practical purposes, under a suitable all-core load, the 7950X will be running at 95C Tj.Max all the time. If you throw a bigger cooler on it, it will boost to higher clock speeds to get right back to Tj.Max. Because of this, the performance of the chip is dependent on the cooling. And it is basically impossible to hit the 230W power limit at stock without direct die or sub-ambient.

If 95C sounds scary, wait to you see the voltages involved. AMD advertises 5.7 GHz. In reality, a slightly higher value of 5.75 GHz seems to be the norm - often across half the cores simultaneously. So it's not just a single core load. The Fmax is 5.85 GHz, but I have never personally seen it go above 5.75.

If CPUz's reading is to be trusted, the 5.75 GHz requires 1.5v vcore - at stock settings. Time will tell if there are longevity issues with this kind of abuse. (Though worth mentioning that Intel is also headed in this direction with their 6 GHz Raptor Lake.)

 

[LAwLz]

22 minutes ago, porina said:

Writer of Y-cruncher benchmark had early access to Zen 4 and just released his observations along with everyone else. Part of it explains what we saw in these benchmark results.

 

https://mersenneforum.org/showthread.php?p=614191#post614191

 

 

Not sure if I should be impressed or scared by these results.

Being able to boost to over 5.75GHz at stock is really impressive, especially if it's on multiple cores.

Constantly hitting TjMax regardless of cooling and that very high voltage doesn't feel right to me.

 

 

But I won't be buying that particular SKU, and if someone who does is scared of the heat and power usage they can always cap the frequency at something lower. I am sure these chips are operating quite far outside of the peak of their efficiency curve.

[MageTank]

7 minutes ago, LAwLz said:

Constantly hitting TjMax regardless of cooling and that very high voltage doesn't feel right to me.

Ignore the reported CPU temps. Nothing has been updated to currently reflect accurate temps on these CPUs, including Ryzen Master.

 

Regardless, this is a very impressive leap in performance, interesting to see if anyone can pull off a 1:1 fabric on memory. Right now, 3:2 ain't looking too good.

[LAwLz]

9 minutes ago, MageTank said:

Ignore the reported CPU temps. Nothing has been updated to currently reflect accurate temps on these CPUs, including Ryzen Master.

Are you sure about that?

It sounds to me like swapping the cooler will make the CPU run just as hot, but at a higher frequency. To me, that indicates that it is hitting a temperature ceiling, which is probably somewhere around 100 degrees.

[Tan3l6]

As I understand, the max CPU temp for at least 7950X is 95C. But it can be changed, set higher?

[MageTank]

2 minutes ago, LAwLz said:

Are you sure about that?

It sounds to me like swapping the cooler will make the CPU run just as hot, but at a higher frequency. To me, that indicates that it is hitting a temperature ceiling, which is probably somewhere around 100 degrees.

I am very, very sure, I just cannot disclose how I am sure. 

[tim0901]

13 minutes ago, MageTank said:

I am very, very sure, I just cannot disclose how I am sure. 

Then why didn't AMD point this out to reviewers? Outlets like LTT or GN are in constant contact with AMD during the process of making these reviews confirming that the numbers they are seeing are "correct" and if this weren't the expected behaviour it surely would have been communicated and therefore mentioned in the review as being a bug.