*** AMD "Zen 4" thread (inc AM5/APU discussion) ***

[Angilion]

This is always put out of proportion. Reminder that these values are ns. High tier audio monitoring programs like latencymon work in microseconds.

1ns = 0.001 microsecond
60 ns = 0.060 microseconds

Anything below 400 microseconds is considered imperceptible for any kind of content. That's 400ns. The latency of the CPU could double to ~120ns and it still wouldn't matter. I don't remember when latencymon starts saying your system might have trouble processing real time stuff but it's around 700 microseconds if I'm not mistaken.

I'm not disagreeing with your point, but your numbers in the last line are out. 400 microseconds is 400,000ns rather than 400ns. If the degree of latency that might cause trouble processing real time stuff is around 700 microseconds, that's around 700,000ns. If it's around 700ns, that's 0.7 microseconds.

 

[GregI]

So, memory latency on Zen 3 or Zen 4 is nothing to be concerned about then?

I suppose I'm also a bit curious about what the theoretical performance differences might be.

Never has been. Reminder that Zen 2 typical latency was 70ns and it meant no difference either: https://i.imgur.com/e3uYoSK.png

There was some weird rumors that AMD had worse input latency than Intel but Gamers Nexus refuted it completely: https://www.youtube.com/watch?v=4WYIlhzE72s

Also reminder that most people on the internet right now likely have systems worse than this PC here: https://www.youtube.com/watch?v=DVBfXg8C2Ws

Now that's ~80ns. As said previously it doesn't matter. 20ns higher latency but still nobody on these laptops or whatever ever says "wow the audio feels out of sync".

@Angilion You're right. It's late ok Makes the point even better tbh.

 

[g67575]

GregI - Thanks for posting those videos.

I suspect that humans can't really perceive differences in microseconds or nanoseconds (just too small a period of time), it makes sense that in the game performance testing, they are measuring differences in milliseconds, and even these differences are small between Intel and AMD (and doesn't seem to correspond to differences in each system's memory latency).

 

[Armageus]

Thread tidied and usual suspects given a short break from this thread - try and keep it on topic and relevant to AMD's latest chips please

 

[Ross Thomson]

I'll be upgrading a 1700 cooled by a block that was designed for threadripper so will be interesting to see the temps I get.

 

[g67575]

It looks like some of the early Ryzen 7000 silicon samples are running 'overvolted' by default which ofc results in much higher temps. In the link below, a twitter user shows a 6 core 7600X at stock voltage running over 90 Celsius, apparently ~5.05ghz all core (with nearly 2x the power consumed), in Aida64 Stability test.

Then, he shows another example, where the voltage has been reduced manually, running ~56 Celsius again at ~5.05ghz all core (this time consuming ~67w). Judging by the much reduced power usage, it's probably safe to say that the voltage is running significantly lower.

AMD Ryzen 7000 CPU Undervolting Can Leave A Lot of Frequency & TDP Headroom, Also Drops Temps Signficantly

AMD Ryzen 7000 CPUs are rumored to run really hot but undervolting them can offer significant thermal, power and performance headroom.

wccftech.com

or here:

AMD Ryzen 7000 Undervolting Yields Great Results with Temperatures

AMD Ryzen 7000 "Zen 4" processors can hit up to 95 °C at stock settings, with cooling most appropriate to the TDP level. This is because the PPT (package power tracking) limits for the 170 W TDP processors is as high as 230 W, and for the 105 W TDP models, it's 130 W. After reaching this...

www.techpowerup.com

Interestingly, the CCDs and IHS (CPU lid) are gold plated on Zen 4, to increase thermal conductivity... Noctua seems to be suggesting that a single point of thermal paste in the centre, is best for AM5 CPUs...

EDIT - I'm looking at the comments on this article and unfortunately a good number of them seem dumb and opinionated (plus ça change?). Oh well, a little entertaining I suppose to watch fanatics jumping to conclusions.

 

[Grim5]

Overvolted, probably - users say under all core load these zen4 CPUs are doing 1.4v which we all know is very high for a ryzen. With my stock 5950x in cinebenc, it uses 1.0v and with pbo on it uses 1.25v, but these stock 7950x and 7600x are reporting 1.4v in Cinebench. And if you used 1.4v on Zen3 you can get a 5950x to do all core 4.9ghz as well so this makes Zen4 look less impressive if it ends up using that kind of voltage at launch


Also: don't read comments on wccftech or videocardz, there is no moderation so the comments are filled with fanboy12 year olds posting memes and fighting each other

 

[Robert896r1]

Give it a week or two and let’s see how things shake out. Right now, the memory side is quite a mess so expect a flurry of FW updates leading upto launch.

 

[g67575]

It looks like these are the kind of per core performance improvements we will see between Zen 3 and Zen 4:



This is a breakdown of the Geekbench 5 results for Zen 3 + Zen 4. 22% and 23.1% for Integer and Floating Point workloads, which seems good to me.

The Crypto score has been heavily influenced by Zen 4's usage of AVX-512, I wouldn't expect this to make much difference in most cases, so should probably be ignored.

For example, I doubt many games (if any) will make use of AVX-512 instructions (some do require AVX CPU instructions though). I would imagine AVX-512 is useful for specific tasks like rendering.

Article here:

AVX-512 Makes Ryzen 9 7950X Geekbench 5 Results Look Good — Too Good

The scores are real, the real-world performance gains aren't

www.tomshardware.com

Geekbench 5 results:
Ryzen 9 7950: https://browser.geekbench.com/v5/cpu/16969227
i7 13700K: https://browser.geekbench.com/v5/cpu/16979552

Based on these results, Zen 4 does well in Floating Point workloads, Intel's Golden Cove based CPUs do well in Integer workloads.

 

[Dave2150]

Can't wait for reviews!

Hope the launch delay is sufficient time to make the launch bios's rock solid stable. Bit concerned about the rumours that Zen4 is very hard to cool (thanks to more heat density from smaller process/design) - though this could just be a silly rumour.

 

[Freddie1980]

Looking at the Zen 4 design of the cache the arrangement looks like it's potentially regressed the IPC gains over Zen 3 that benefited from having a shared cache pool. I wonder if AMD plans to revert back to the Zen 3 arrangement for Zen 5?

 

[keyser van someone]

Can't wait for reviews!

Hope the launch delay is sufficient time to make the launch bios's rock solid stable. Bit concerned about the rumours that Zen4 is very hard to cool (thanks to more heat density from smaller process/design) - though this could just be a silly rumour.

Yeah I did wonder the same, I remember an Anandtech article about this issue on previous intel chips where the density was the reason for increased heat, but that all comes down to what the flip side gains are from reduced voltage. For me AMD's selling point is you plug and play and don't really have to tweak at all, don't worry about temps, if it runs hot oh well so be it, it's all just managing itself 1000 times a second. If a CPU runs at 90C and its rated to work up to 105C then its 15C within spec

 

[Destination]

I am very interested to see how userShillbenchmark will spin this one.
It all looks very interesting and should assist rivals to continue to push progress.
I've no plans to upgrade, my 5900x does all I need it to.
Still wanting GFX to continue to drop in price as hats where I would spend money post recession.

 

[Joxeon]

I am very interested to see how userShillbenchmark will spin this one.
It all looks very interesting and should assist rivals to continue to push progress.
I've no plans to upgrade, my 5900x does all I need it to.
Still wanting GFX to continue to drop in price as hats where I would spend money post recession.

I don't think there will be enough of a performance boost in gaming to warrant an upgrade from Zen 3 as even at 1080p according to AMD you're only looking at an average +15% for a starting cost of around £700 for board, memory and a 6 core CPU.

That £700 would be much better spent on a 4080 / 7800XT which should give over double those gains even for those already on a 3090 class card.

 

[Grim5]

I am very interested to see how userShillbenchmark will spin this one.

It all looks very interesting and should assist rivals to continue to push progress.
I've no plans to upgrade, my 5900x does all I need it to.
Still wanting GFX to continue to drop in price as hats where I would spend money post recession.

Easy actually to figure out what they'll do.

They'll say:

* cores matter and Intel has more cores
* they'll say memory latency matters and Intel wins
* they'll say amd is overpriced and other review sites are paid amd shills and Intel needs to fight AMD's marketing Shillary


Look, userbenchmark is most likely run by Intel itself - it doesn't matter how good AMD is, even if they beat Intel by 100% in every single thing and sold their CPUs for $1, Userbenchmark would still say it sucks

 

[Journey]

I don't think there will be enough of a performance boost in gaming to warrant an upgrade from Zen 3 as even at 1080p according to AMD you're only looking at an average +15% for a starting cost of around £700 for board, memory and a 6 core CPU.

That £700 would be much better spent on a 4080 / 7800XT which should give over double those gains even for those already on a 3090 class card.

They literally said, no plans to upgrade.

I've no plans to upgrade, my 5900x does all I need it to.

 

[Deleted member 258511]

Yeah I did wonder the same, I remember an Anandtech article about this issue on previous intel chips where the density was the reason for increased heat, but that all comes down to what the flip side gains are from reduced voltage. For me AMD's selling point is you plug and play and don't really have to tweak at all, don't worry about temps, if it runs hot oh well so be it, it's all just managing itself 1000 times a second. If a CPU runs at 90C and its rated to work up to 105C then its 15C within spec

Heat is my concern. What are the AMD CPU’s temperature limits? Now if it is in fact as you say “up to 105c” then now I understand why 90c is not something to panic about.

 

[keyser van someone]

Heat is my concern. What are the AMD CPU’s temperature limits? Now if it is in fact as you say “up to 105c” then now I understand why 90c is not something to panic about.

105C was plucked from thin air but looks like the 'Tjmax' temp is 95C on the 7700X which I think just means temp before throttling starts to occur for safety. But imo any temp below 95C is within spec and fine to use. At the end of the day, what are you computing that would cause such a CPU load in the first place? Usually its during a stress test but real world I only use Chrome, youtube and games, CPU usage for those is rarely over 80% on not even all 8 cores.

https://www.amd.com/en/product/12161

Zen 3 was fine ~90C. I think any beefy air cooler will be fine.

AMD’s Robert Hallock has clarified that temperatures up to 90C for the higher-end Zen 3 based Ryzen 7 and 9 parts are quite normal, and won’t affect the life-cycle of the chip.

Max Safe Temperature for AMD Ryzen 5000 CPUs? | Hardware Times

AMD’s Robert Hallock has clarified that temperatures up to 90C for the higher-end Zen 3 based Ryzen 7 and 9 parts are quite normal, and won’t affect the life-cycle of the chip. Replying to a Redditor, Hallock said that AMD views temperatures up to 90C (for the 5800X/5900X/5950X) as typical and...

www.hardwaretimes.com

 

[Deleted member 258511]

105C was plucked from thin air but looks like the 'Tjmax' temp is 95C on the 7700X which I think just means temp before throttling starts to occur for safety. But imo any temp below 95C is within spec and fine to use. At the end of the day, what are you computing that would cause such a CPU load in the first place? Usually its during a stress test but real world I only use Chrome, youtube and games, CPU usage for those is rarely over 80% on not even all 8 cores.

https://www.amd.com/en/product/12161

Zen 3 was fine ~90C. I think any beefy air cooler will be fine.

Max Safe Temperature for AMD Ryzen 5000 CPUs? | Hardware Times

AMD’s Robert Hallock has clarified that temperatures up to 90C for the higher-end Zen 3 based Ryzen 7 and 9 parts are quite normal, and won’t affect the life-cycle of the chip. Replying to a Redditor, Hallock said that AMD views temperatures up to 90C (for the 5800X/5900X/5950X) as typical and...

www.hardwaretimes.com

Thanks for the information.

Someone mentioned yesterday that they see temps above 80c in gaming and that was in a good airflow case etc. Yes it’s still under 90c but it’s coming close for my liking. My 10900k is never reaching 70c and mostly sits in the high 50c-low 60c range whilst gaming.

I’d be interested to see how much performance is affected if it did in fact reach 90c. Especially with the 4000 series GPU pumping out more heat etc and the new AMD CPU’s apparently even hotter than the 5000.

 

[keyser van someone]

Here's a cool video showing typically how an AMD 3600 CPU handles heat overload, I'm sure all the AMD CPUs do this. https://www.youtube.com/watch?v=0RYFsb99OwI

I actually think that's part of the secret sauce for AMD as if you have voltage/clock regulated on the CPU and its doing it 1000 times per second it may as well reach as fast and as close to thermal limit as it can.

Yeah the CPU is cooler in gaming vs say Cinebench, but it's the case temp that rises due to the GPU. Cranking the case fans a bit should help with that.

Yeah we all like review day I enjoy all the competition even if I might not buy the AM5 right away.