News AMD deep-dives Zen 5 — Ryzen 9000 and AI 300 benchmarks, Zen 5, RDNA 3.5 GPU, and XDNA 2 microarchitectures

[Admin]

AMD revealed the deep-dive details of its Zen 5 Ryzen 9000 ‘Granite Ridge’ and Ryzen AI 300 series ‘Strix Point’ chips at its Zen 5 Tech Day.

AMD deep-dives Zen 5 — Ryzen 9000 and AI 300 benchmarks, Zen 5, RDNA 3.5 GPU, and XDNA 2 microarchitectures : Read more

 

[Amdlova]

Want to see some benchmark at 35w and 50w power limit it's My budget power level

 

[TerryLaze]

AMD revealed the deep-dive details of its Zen 5 Ryzen 9000 ‘Granite Ridge’ and Ryzen AI 300 series ‘Strix Point’ chips at its Zen 5 Tech Day.

AMD deep-dives Zen 5 — Ryzen 9000 and AI 300 benchmarks, Zen 5, RDNA 3.5 GPU, and XDNA 2 microarchitectures : Read more

TDP / PBP / MTP

Please don't use intel specific terminology on ryzen, they don't make any sense.
It's TDP and PPT only for ryzen.

 

[TerryLaze]

Also testing was done with a ~ $200 water cooler....just saying but if you need that to run the 9950x at stock, 230w ppt, then it's gonna be a joke, at least on the intel system you would get like ~350W out of that.

 

[jeremyj_83]

Also testing was done with a ~ $200 water cooler....just saying but if you need that to run the 9950x at stock, 230w ppt, then it's gonna be a joke, at least on the intel system you would get like ~350W out of that.

So if you can get more performance out of the AMD with a lower power draw that is a negative?

 

[TerryLaze]

So if you can get more performance out of the AMD with a lower power draw that is a negative?

If you have to pay another $200 on top of the price of the CPU to get the performance that AMD claims then that is a bad thing.
And I don't know how you do math but 230W of the 9950x is not lower than 230W of the 7950x

Being able to use 330-50W with the same cooling that another CPU can only use 230-50w with is a good thing because that means that if you use less power on it you will have much better temps.
Although we don't know, the other article shows the 9950x using 320W so maybe AMD chose to show overclocking numbers for their presentation, I honestly don't know which would be worse.

 

[jeremyj_83]

If you have to pay another $200 on top of the price of the CPU to get the performance that AMD claims then that is a bad thing.
And I don't know how you do math but 230W of the 9950x is not lower than 230W of the 7950x

Being able to use 330-50W with the same cooling that another CPU can only use 230-50w with is a good thing because that means that if you use less power on it you will have much better temps.
Although we don't know, the other article shows the 9950x using 320W so maybe AMD chose to show overclocking numbers for their presentation, I honestly don't know which would be worse.

Literally none of what you are saying makes sense. On top of that AMD might have used a $200 cooler to make sure that they couldn't be called out for hurting possible i9-14900k performance. Also note that at a 170W TDP the AMD chips have a 230W PPT and that is based on AM5 specifications.

 

[evdjj3j]

Also testing was done with a ~ $200 water cooler....just saying but if you need that to run the 9950x at stock, 230w ppt, then it's gonna be a joke, at least on the intel system you would get like ~350W out of that.

Wow, I lost some IQ points reading that.

 

[TheSecondPower]

"The [Arrow Lake] chips are also said to come with the new Lion Cove P-cores and Gracemont E-cores." That should say "Skymont E-cores." Gracemont is used in Alder Lake and Raptor Lake.

 

[TerryLaze]

Literally none of what you are saying makes sense. On top of that AMD might have used a $200 cooler to make sure that they couldn't be called out for hurting possible i9-14900k performance. Also note that at a 170W TDP the AMD chips have a 230W PPT and that is based on AM5 specifications.

Wow, I lost some IQ points reading that.

It's not like it's a secret or in any way controversial that ryzen is very hard to cool.
The same amount of cooling that is required to get the PPT of ryzen at thermal throttle temps is enough to give intel 50% more power draw at 8 degrees lower temp.
https://www.anandtech.com/show/17641/lighter-touch-cpu-power-scaling-13900k-7950x/3

 

[Makaveli]

 

[PaulAlcorn]

"The [Arrow Lake] chips are also said to come with the new Lion Cove P-cores and Gracemont E-cores." That should say "Skymont E-cores." Gracemont is used in Alder Lake and Raptor Lake.

Oops, fixed!

 

[Peksha]

"Both core types have their own private L1 and L2 caches, but the 16MB of L3 cache is split into two 8MB slices, with one slice shared between the Zen 5C cores and the other slice shared between the standard Zen 5 cores (this can be seen as a single line running down the center of the eight Zen 5C cores in the block diagram)."

Isn't that 16MB L3 for 4xZen 5 + 8MB L3 for 8xZen 5c (24MB total)? Or 8+16?

 

[JarredWaltonGPU]

"The [Arrow Lake] chips are also said to come with the new Lion Cove P-cores and Gracemont E-cores." That should say "Skymont E-cores." Gracemont is used in Alder Lake and Raptor Lake.

Fixed. It's Skymont for Arrow Lake.

 

[m3city]

If you have to pay another $200 on top of the price of the CPU to get the performance that AMD claims then that is a bad thing.
And I don't know how you do math but 230W of the 9950x is not lower than 230W of the 7950x

Being able to use 330-50W with the same cooling that another CPU can only use 230-50w with is a good thing because that means that if you use less power on it you will have much better temps.
Although we don't know, the other article shows the 9950x using 320W so maybe AMD chose to show overclocking numbers for their presentation, I honestly don't know which would be worse.

And you don't have to pay that for intel 14th because air cookers they provide are so great that they handle it effortlessly?

I enjoy reading discussions with you, they often go into very interesting details. But I have never ever seen you saying something nice and good about AMD without "but". Intel's never bad, wrong, always better. And when somebody proves you wrong or misinterpreting then you simply exercise goalpost shifting. Excuse me for ad personam commentary, but it's kinda funny that you tend to bend each story to "Intel is better".

 

[jeremyj_83]

It's not like it's a secret or in any way controversial that ryzen is very hard to cool.
The same amount of cooling that is required to get the PPT of ryzen at thermal throttle temps is enough to give intel 50% more power draw at 8 degrees lower temp.
https://www.anandtech.com/show/17641/lighter-touch-cpu-power-scaling-13900k-7950x/3

Comparing power draw to CPU die temp from different companies is worthless. The 14900k is VASTY more difficult to cool than the Ryzens due to its extra power draw.

 

[JarredWaltonGPU]

Comparing power draw to CPU die temp from different companies is worthless. The 14900k is VASTY more difficult to cool than the Ryzens due to its extra power draw.

Funny thing is that I haven't heard anyone talking about AMD Ryzen CPUs failing due to exceeding power / voltage / current limits. And yet, the Core i9-13900K and 14900K are apparently becoming a serious issue a year or two after release. I had to get my 13900K chip replaced because it failed.

Anyway, forum regulars know Terry has a massively skewed pro-Intel / anti-AMD perspective. It's not worth trying to argue facts with such types in my experience.

 

[jeremyj_83]

Funny thing is that I haven't heard anyone talking about AMD Ryzen CPUs failing due to exceeding power / voltage / current limits. And yet, the Core i9-13900K and 14900K are apparently becoming a serious issue a year or two after release. I had to get my 13900K chip replaced because it failed.

Anyway, forum regulars know Terry has a massively skewed pro-Intel / anti-AMD perspective. It's not worth trying to argue facts with such types in my experience.

Sorry to hear that you had to get your CPU replaced. I do wonder if this is going to effect SPR and EMR server CPUs as well.

 

[TerryLaze]

And you don't have to pay that for intel 14th because air cookers they provide are so great that they handle it effortlessly?

I enjoy reading discussions with you, they often go into very interesting details. But I have never ever seen you saying something nice and good about AMD without "but". Intel's never bad, wrong, always better. And when somebody proves you wrong or misinterpreting then you simply exercise goalpost shifting. Excuse me for ad personam commentary, but it's kinda funny that you tend to bend each story to "Intel is better".

If you want to overclock and get up to 330-50W you have to pay, but to get the max allowed power of 253W you can get that with a $20 cooler, yes at high noise and heat but sustained.
https://www.tomshardware.com/features/intel-core-13900k-cooling-tested/2

The $20 Assassin 120 R SE sustained 5055MHz (an increase of 333MHz) with the CPU consuming an average of 245W.

 

[stuff and nonesense]

As the 4790k was a worthwhile upgrade for my old Phenom 2, the published numbers seem to suggest that the 9000 ryzens will be a worthwhile upgrade for my 3900x. Power seems to be comparable with a healthy boost to clocks and IPC. (5000 wasn’t worth it, 7000 was too much too new, 9000 a good refinement with a few more bugs knocked off?)

I hope that Intel can begin to make less vulnerable, both physical and software, processors. AMD needs competition.

 

[JarredWaltonGPU]

Sorry to hear that you had to get your CPU replaced. I do wonder if this is going to effect SPR and EMR server CPUs as well.

I didn't have to pay for the replacement... just had to deal with the problems of an unstable chip for a lot longer than I should have. (I probably should have been talking to someone about this 18 months ago! But I figured out some workarounds instead...)

As for it affecting SPR and EMP, almost certainly not. All indications is that part of the problem is motherboards pushing unsafe settings to "stock" options to boost performance. Like, there's a flag that can be set to help improve LN2 overclocking performance, but apparently that was being used by certain motherboard vendors to eke out a bit more performance even without LN2 setups. The Xeon parts should be in boards that are far more concerned with stability than a minor boost in performance.

 

[Fleshharrower]

Putting aside all the discussion on thermals, I'd need to see how Intel's new processors perform over months of time before considering them again. Intel is still trying to figuring out the stability issues with 13900K and 14900K. I was bit by this problem and spent a week in January of this year trying to figure out why my over 1 year old non-overclocked 13900K was unstable in games when it hadn't been before including several different attempts at different video cards, power supplies, memory, and more from my other PCs in the house. The issue was finally solved by down clocking by 300MHz on the CPU and 400MHz on the memory which gave me crashes still but much less frequently. I'm not going through that again until I can be certain the new chips aren't going to fry themselves out overtime.

 

[jackt]

ok ill wait 5c

 

[-Fran-]

Ugh... I don't like the fact that AMD has been hush hush with the prices. They've been forthcoming with them for a while, but now they are not. I don't want to read too much in between the lines, but I don't like it.

On the upside: well, well, well. Would you look at that. They're touting IHS thermal conductivity improvements. Where are the "but the IHS is fine" apologists now? For crying out loud... At least AMD is no Intel when it comes to improve things pointed out with close-to irrefutable proof: no acknowledgment of the bending problems with the stock ILM or news about the Raptor Lake bug telling your consumers you got their back. For friggen' shame, Intel.

Regards.

 

[tamalero]

Ugh... I don't like the fact that AMD has been hush hush with the prices. They've been forthcoming with them for a while, but now they are not. I don't want to read too much in between the lines, but I don't like it.

On the upside: well, well, well. Would you look at that. They're touting IHS thermal conductivity improvements. Where are the "but the IHS is fine" apologists now? For crying out loud... At least AMD is no Intel when it comes to improve things pointed out with close-to irrefutable proof: no acknowledgment of the bending problems with the stock ILM or news about the Raptor Lake bug telling your consumers you got their back. For friggen' shame, Intel.

Regards.

maybe waiting how bad intel is hurting with 14900k and arrow lakes pricing?