Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

[DisEnchantment]

Speculate at will

 

[Nothingness]

Ryzen 7 9700X im Linux-Test: Schneller als unter Windows? [Update]

Nachdem der Ryzen 7 9700X und Ryzen 5 9600X unter Windows nicht ganz das halten konnten, was AMD versprochen hat, muss sich der 9700X unter Linux beweisen.

www.pcgameshardware.de

They only test a handful of games but every game has a performance improvement in lows and averages on Linux with Proton vs Windows 11. An issue with scheduling is beginning to look more likely to me as a major culprit for the performance differential from official figures to actual reviews.

Beyond games there's GB 6.3. Interestingly in ST, Linux is about 4% faster, while it's 8% faster in MT despite poor scaling of some of the subtests. Linux scheduler looks definitely better.

As wrote @gdansk it'd be interesting to run the same benchmarks on Zen4 too and see how it goes.

 

[RnR_au]

L1 tech has 2133/6400 vs 5600 in of the 1080p game-charts
But these are also just junk EXPO timings vs EXPO timings
View attachment 105167

But as ive said many times already, if your looking for a gaming cpu --> wait for X3D

What is the memory scaling like for the 7000 series?

 

[carancho]

Probably? There are some rumours that Zen 6 DT client will be different from server.

Here's a hot take from me: I personally believe that all CPUs post-Alder Lake are a vast freaking overkill for gaming. It's really not that relevant whether CS2 runs at 600 fps or 750.

There are some games that hammer the CPU (Starfield comes to mind), but it's really mostly devs' fault.

What I'd like to see improved is single-core power consumption. It'd be hella nice to see top-end laptop chip top out at 15W @ single-core at most.

I have a >2700 Geekbench PC and still feel the browser and Excel slow. I'll happily take any ST boost I can pay for - unless it's a Mac.

 

[poke01]

I have a >2700 Geekbench PC and still feel the browser and Excel slow. I'll happily take any ST boost I can pay for - unless it's a Mac.

Keep an eye out for Qualcomms Gen 2 X Elite. It should have the best ST on a PC laptop. (well, i hope it does)

 

[carancho]

I recommend these two videos from skatterbench as beginner CO tuning guide

Crazy-good content. It's a must watch.

 

[mostwanted002]

Do we have any takers here?

 

[Markfw]

Do we have any takers here?

I just built a 9950 (except the processor) otherwise I would. Thanks for the link. Also, for this buch, I would love the see at least one Turin review.

 

[Hotrod2go]

of course he is doing it on a purpose, he´s secret Intel agent and so did HWU with their testing , I am getting feeling, that everybody want´s to hurt poor AMD

There's that.. but overall there is inconsistency in testing methodologies across a LOT of PC enthusiast sites with Zen 5. Never have I seen so many shenanigans over a product launch before.
A quick common discrepancy that comes to mind is testing against Zen 4, why don't they test that with its early release bios & agesa so as to compare with current situation with Zen 5 today? They can't compare product A with B when A has been on the market longer & thus had time to mature with low level codes.

 

[poke01]

A quick common discrepancy that comes to mind is testing against Zen 4, why don't they test that with its early release bios & agesa so as to compare with current situation with Zen 5 today?

Reviewers shouldn't have use an old agsea for Zen4, its AMD fault if Zen5 isnt up to par for gaming. Zen4 is mature and cheaper and for most gamers presents the best value right now.
Now that may change with Zen5X3D but i still think on AM5 the 7800X3D will be the value king for a long while.

 

[dr1337]

Reviewers shouldn't have use an old agsea for Zen4, its AMD fault if Zen5 isnt up to par for gaming. Zen4 is mature and cheaper and for most gamers presents the best value right now.
Now that may change with Zen5X3D but i still think on AM5 the 7800X3D will be the value king for a long while.

I get the simplistic view that reviewers should just run benchmarks and call it good. Sometimes in the moment that is all that matters.

But a lot of us run their computer for years into the future, I mean heck RAM tuning is so popular that its crazy for how technical it is. How these systems change over time is super important, remember all the x86 vulnerabilities that came out in the past few years? Products at release don't exist within a vacuum and it would be interesting to know the difference between zen 4 then and now.

Like imagine if any of the reviewers had been consistent that data from 2 years would still be relevant? But instead we constantly see new systems, new numbers, ect. For an IRL example look at TPU benchmarks for the 5800x3d, at launch they place it as being the fastest gaming CPU even beyond the 12900k at the time. However if you look at 5800x3d vs 12900k now in any of their current charts, it generally scores lower. For any of their CPU benchmarks the standards and test beds constantly change, and this is also true with a lot of other sites. Sure, people list the test bed specs in the articles, but also always gloss over the changes from last time.

I don't think we can rationally trust any reviewer to be that consistent over time, at least not here in 2024, but we can be conscious of the idea that performance changes over time.

 

[dr1337]

I wish the tech press would get more into the mindset they had 2018/2019, where people were tuning the nuts off of their rigs and analyzing things like overclocking and ram to the finest detail.

Sure Zen 5 is a dud in some ways but to me its extremely interesting from a technical perspective. Between density improvements and AMDs ability to somehow make a much wider core more efficient without increasing performance really pique's my curiosity.

 

[Ranulf]

There's that.. but overall there is inconsistency in testing methodologies across a LOT of PC enthusiast sites with Zen 5. Never have I seen so many shenanigans over a product launch before.
A quick common discrepancy that comes to mind is testing against Zen 4, why don't they test that with its early release bios & agesa so as to compare with current situation with Zen 5 today?

Sure, that would be nice I guess...but that just makes the 9000 chips look even worse on price. Like they were on release compared to zen3, pricey parts all around (cpu, mobo, ram) and early teething issues. It is all down to price/perf as usual. I mean, if power efficiency is that important to your computing purposes then by all means, pay $80-100 more to save some power. For some server or productivity apps maybe it is worth it to you. Probably a small market though because most will want to go with 12-16 core parts.

Gaming or just general office use, the 9000 chips are pointless until they drop in price. All for 3-5% in gaming performance with a few anomolies on either side, positive and negative. I know I would be more impressed if AMD had priced these chips $50 less each recognizing the reality of the current market. My 12700k was $260 last year and by TechPowerup's numbers, is only 7% slower than a 9700x in games. But it will use less power I guess, yay. I'm not seeing much better in productivity benchmarks either. Not for $100 more a year later.

If I've learned anything from AMD since Ryzen 1st gen (or even hawaii and polaris gpus), if you care about bang for your buck and not dealing with early adopter issues avoid the first gen AMD part.

 

[Hotrod2go]

Reviewers shouldn't have use an old agsea for Zen4, its AMD fault if Zen5 isnt up to par for gaming. Zen4 is mature and cheaper and for most gamers presents the best value right now.
Now that may change with Zen5X3D but i still think on AM5 the 7800X3D will be the value king for a long while.

No, I already have two Zen 4s, not investing in another one. Doesn't make sense to invest in another nearly 2 year old design when the new ones have just been released & will only get better from here on just like Zen 4 did when it first came out.

 

[Hotrod2go]

Sure, that would be nice I guess...but that just makes the 9000 chips look even worse on price. Like they were on release compared to zen3, pricey parts all around (cpu, mobo, ram) and early teething issues. It is all down to price/perf as usual. I mean, if power efficiency is that important to your computing purposes then by all means, pay $80-100 more to save some power. For some server or productivity apps maybe it is worth it to you. Probably a small market though because most will want to go with 12-16 core parts.

Gaming or just general office use, the 9000 chips are pointless until they drop in price. All for 3-5% in gaming performance with a few anomolies on either side, positive and negative. I know I would be more impressed if AMD had priced these chips $50 less each recognizing the reality of the current market. My 12700k was $260 last year and by TechPowerup's numbers, is only 7% slower than a 9700x in games. But it will use less power I guess, yay. I'm not seeing much better in productivity benchmarks either. Not for $100 more a year later.

If I've learned anything from AMD since Ryzen 1st gen (or even hawaii and polaris gpus), if you care about bang for your buck and not dealing with early adopter issues avoid the first gen AMD part.

Not pricey for me, I already have had a mid range B650 board sitting here since last year spare with nothing to do. Everyone's situation is different, taking into account the basics ( cpu, ram, board) does not apply to all users of Zen 4 already.

 

[jdubs03]

No, I already have two Zen 4s, not investing in another one. Doesn't make sense to invest in another nearly 2 year old design when the new ones have just been released & will only get better from here on just like Zen 4 did when it first came out.

Perhaps they’ll gain some performance, but probably low single digits. And won’t be worth it until it’s only slightly more expensive than their predecessors.
Sure some specific workloads see a good uplift, but those are specific server type workloads.
For gaming, or normal desktop use, what is the need to invest in an upgraded CPU? The 7000 series should be plenty quick.

 

[DrMrLordX]

That shows how sensitive the results are: use fast memory, IOD uses more power, lose some watts for the core power, get low scores in benchmarks that don't rely on memory speed. It's crazy that the power budget for the cores can vary by more than 20% just by using different RAM.

One gets the impression that they're using too much vSoC on higher-speed RAM.

Most office type work is also what home PC is used for. Web browsing, MS Office, etc.

That's not how DiY desktop CPU reviews shake out, though.

 

[Det0x]

What is the memory scaling like for the 7000 series?

Well there is this, but very few if any i know are running stock EXPO/XMP 6000/2000.. Ill ask around if anyone got

 

[Timmah!]

So, 9950x reviews today? Or was it tomorrow?

 

[Markfw]

reviews in about 42 hours from now I am pretty sure (9 am PST, pacific time, western US).

 

[Det0x]

na reviews in ~26 hours, release for sale in 50 hours
Same as with the 9700X and 9600X

 

[Mahboi]

Most of the gaming ones should be a long sigh, but I'm really really curious about how far the bars on Phoronix' benchmarks will go.
Sure, Michael greatly exaggerated how "good" Zen 5 is, it's not nearly that good, but one the server-side parts where it shines, it shines very bright. It will set a very serious bar in server for Intel to overcome.

 

[LightningZ71]

If theory holds, Phoronix's benchmarks, at least the ones that have a notable degree of memory throughput, should start to hit limits due to the RAM configuration with twice as many cores. It would be interesting to see the memory access profile of the benchmark processes and find the ones that are hitting that wall.

 

[DZero]

Most of the gaming ones should be a long sigh, but I'm really really curious about how far the bars on Phoronix' benchmarks will go.
Sure, Michael greatly exaggerated how "good" Zen 5 is, it's not nearly that good, but one the server-side parts where it shines, it shines very bright. It will set a very serious bar in server for Intel to overcome.

Zen 5 looks decent enough, at least compared to the jumps from Intel 4th to 11th generation.

 

[Mahboi]

Zen 5 looks decent enough, at least compared to the jumps from Intel 4th to 11th generation.

How so? Skylake era was ~10% a year wasn't it?
This is 22 months and looks like a 5% improvement for a lot of scalar-bound apps. AVX /large data amounts apps do great, but I wouldn't call it "decent" right now, just...interesting. Lots of potential, lots of disappointments.

 

[Mahboi]

If theory holds, Phoronix's benchmarks, at least the ones that have a notable degree of memory throughput, should start to hit limits due to the RAM configuration with twice as many cores. It would be interesting to see the memory access profile of the benchmark processes and find the ones that are hitting that wall.

If that does happen, it will be fascinating to see how EPYC handles it. If truly 16 cores start being too much for AMD, then how is the server memctrl going to take in 192 or 256 cores??