News AMD delays its Ryzen 9000 launch due to unspecified quality issue — new launch in August; chipmaker pulls back all units shipped globally for quali...

[munkee_zero]

Just wondering , are the 5900XT and 5800XT also affected by the delay?

 

[stuff and nonesense]

In comparable generations e.g. 5000 / 10xxx, 7000 /12xxx there will be cross overs single thread and multi thread. AMD faster in some, Intel in others. It all depends on what benchmarks are chosen.

There are lies, damn lies and benchmarks.

 

[lordmogul]

If they need the time to make the new chips run solid, they should have the time.
It's not like AMD needs them immediately, they're doing fine with the current current chips.

 

[bit_user]

Everything that is MT, I predict 9700x ~= 12700k and 9600x =~ 12600k. The 13700k / 13600k will fly past them both. Come back after the reviews and let's see if I'm right. I won't hide, I hope you won't either.

So, let's try to examine this, a little more closely. I've taken current street pricing data from PCPartPicker. I got leaked Ryzen 9000 prices off a Reddit post - take with a grain of salt, but it gives us something to work with.

AMD Model	AMD C/T	AMD Price (leaked)	Intel Model (A)	Intel Price (A)	Intel C/T (A)	Intel Model (B)	Intel Price (B)	Intel C/T (B)
9600X	6/12	$229​	i5-12600K	$175​	10/16	i5-13600K	$260​	14/20
9700X	8/16	$299​	i7-12700K	$250​	12/20	i7-13700K	$330​	16/24

So, if we restrict it to MT performance, as you said, then I doubt the Alder Lake options can eke out many wins - especially anywhere Zen 5 can benefit from AVX-512. They match the Ryzen 9000 models on P-cores, perhaps at a clock speed disadvantage, but have an extra 4 E-cores to fall back on. What I think they lack in in MT performance, their price advantage could probably make up for. The i7-12700K is currently available just $20 more than the leaked price of the 9600X, and should have a pretty easy time winning that contest on MT performance.

If we look at the Raptor Lake options, they're both currently at a price disadvantage to the 9700X. MT performance seems like it's going to be more competitive between those models in your matchup.

Of course, people rarely consider just one aspect of a CPU, when making a purchasing decision. Some care about gaming performance, others care about efficiency. It's on these fronts where I think your Raptor Lake contenders will face their most serious challenges.

 

[bit_user]

And no, premiere pro isn't multithreaded. Here ya go, can you see that all CPUs regardless of corecounts perform identically? Unless you are telling me 7600x = 7950x no, premiere pro isn't multithreaded.

I would assume you never read the link to ADOBE about Premier Pro. If you had you would have seen this in the first two paragraphs instead of trying to yet another Intel homer statement.

"People who say that Premiere Pro does not support multi-cores don't understand how multi-core support works. What you're asking about is optimization for high core counts, which is different.

Premiere uses multiple cores. Cores are independent CPU processors that can be assigned tasks. Premiere breaks this down with tasks for decoding codecs, applying CPU-driven effects, encoding to specific codecs, etc. Some of these threads may use a core more than other cores, and right now, an "optimal" number of cores is somewhere around the 8-12 mark."

Whenever there's a mismatch between expectations and data, there's usually an explanation. Often, it's staring you right in the face. Is this case, the Premier Pro benchmark that TechPowerUp ran was just Object Tracking. As Harold observed, that doesn't appear to be well-threaded. As Jeremyj quoted, Premier relies mostly on functional parallelism, in which case the simplistic pipeline used by TechPowerUp wouldn't have done much to exploit it.

TL;DR: find a different/better Premier Pro benchmark that better represents real world usage by a experienced operator.

I am going to assume that you don't speak German

ComputerBase is a good site, though. You can use Google Translate to read it, although that breaks some of their interactive charts.

When we're this starved for good data, I think we have to take what we can get, whenever we can find it!

 

[bit_user]

Sure it can be MT, but I'd love to where your "almost always faster" statistics come from since the preeminent testing for things like that seemingly contradicts your assertion:

https://www.pugetsystems.com/labs/a...review-2369/#Video_Editing_Adobe_Premiere_Pro

Again, it usually pays to look at the fine print. Right below that chart, here's what they say about it:

"Premiere Pro (and video editing in general) is a very strong area for Intel with its “Core” line of processors. They have a significant advantage due to the Quick Sync technology found on the models that include an iGPU, which can be used for hardware decoding and encoding of H.264 and HEVC codecs. In the case of Premiere Pro, Quick Sync tends to give higher performance than using the GPU for decoding, ..."
​

So, it's a valid real-world test, but shouldn't be taken as a proxy for general MT performance.

 

[bit_user]

@bit_user Weird you don't chime in to correct him

I wasn't following this thread, at the time.

Now I need to find an example of a game that uses more than 1 thread.

FWIW, I divide up workloads into two main categories:

1. Lightly-threaded, like productivity apps and many games.
2. Heavily-threaded, like rendering, software compilation, workstation/engineering apps, and certain other games.

By that understanding, MT doesn't mean simply "more than one thread". It means most of the cores are busy most of the time. I'd say average CPU utilization should be at least 70%.

I think ComputerBase and Toms use the second classification for selecting which benchmarks to include in their multithreaded composite score.

 

[bit_user]

Last time I checked the conversation wasn't about the 7950X/13900K it was about the 7600X/7700X compared to the 13600K/13700K. Pretty easy explanation right there.

That's not a position I ever endorsed and I think it's at odds with my take on TheHerald's matchups. I think it's not supported by pricing, either, since the current street price of the R5 7600X is about $210 and the R7 7700X is selling for about $294.

However, you have a point that maybe it's what Puget Systems was talking about, because their article does subsequently match them up like that. In fact, they even pair the R9 7900X with the i9-13900K, rather than matching it against the R9 7950X.

 

[bit_user]

according to PCpartpicker the 12700k is 233$ right now,

You need to be a little careful with the prices it claims. I actually had to go back and fix two of the prices in my table (before you saw it), because two of them were from like Amazon marketplace sellers who almost certainly aren't official resellers. That has implications on warranty coverage, especially if they turn out to be selling OEM tray, rather than retail-boxed.

So, you have to take the next step of checking out the lowest prices and make sure they're officially sold by the online store and not a 3rd party seller. Those are the prices we should use, in order to keep things fair and consistent.

 

[bit_user]

If the 9600x loses to the 13600k in ISO power MT,

Not sure what you mean by iso-power, since the 9600X is a 65W part (presumably with 88 PPT). The i5-13600K has a 125W TDP and 181 W PL2. If it's truly a MT workload, then there's no way they're running iso-power. Therefore, you'd better use a non-K part.

[if] the 9700x loses to the 13700k, doesn't that mean that at least in that segment AMD is at least 2 years behind in efficiency?

Well, again the 9700X is a 65 W part. So, you should really use the i7-13700 (non-K), as your point of comparison. BestBuy currently has it for $320 ($10 cheaper than the K), which also helps close the price gap.

As for whether it reveals a 2-year gap, that's quite a reach. You're picking a single point in the product lineup and suggesting it reflects an overall technology gap? This is the same kind of thing I think Helper was talking about and that we've repeatedly discussed about making sweeping declarations by extrapolating from like a single datapoint. No, I don't agree that's what it tells us.

Let's just wait and see how it all unfolds. I'll look at how all of the lineup compares, on multiple facets, before I draw my own conclusions.

 

[bit_user]

If the 13600 (non k) is more efficient than the 9600 (when both are run within their specific limits, no mobo shaenanigans for either) doesn't that mean that amd is 2 years behind in efficiency? Apply the same to the 9700 / 13700.

The last thing I'm going to say about this, until we know the real street prices and have actual performance & power data is that I also think you can't quite claim the current Raptor Lake product are reflective of Intel @ 2 years ago (or even 22 months, since Raptor Lake launched in late October, not mid August). The reason being that it doesn't account for the current maturity of their manufacturing node, which is enabling Intel to sell them at their current pricing.

On the point of price-parity, we should remember that Intel has a pricing advantage from having in-house fabs. So, part of their advantage vs. AMD is business-level, not technical. However, to the end user, the "why" doesn't really matter.

Okay, I'm done speculating, here. We only have just under 2 weeks until the first Ryzen 9000 series drops. No amount of speculation beats having real data.

 

[Zerk2012]

Funny how it only took 11 post to go from AND recall/delay to AMD VS Intel.
It's always the same 5 or 6 people.
Then we get 3 more pages of off topic crap fights.

 

[SonoraTechnical]

Funny how it only took 11 post to go from AND recall/delay to AMD VS Intel.
It's always the same 5 or 6 people.
Then we get 3 more pages of off topic crap fights.

I hardly participate here as can be seen by low post count. However, I can name 4 of those people already. They just ruin every thread for everyone. Good lord.

I'm delighted to see AMD be proactive on maintaining quality at launch. Good on them!

 

[ex_bubblehead]

I hardly participate here as can be seen by low post count. However, I can name 4 of those people already. They just ruin every thread for everyone. Good lord.

I'm delighted to see AMD be proactive on maintaining quality at launch. Good on them!

And if those 4 or 5 people keep it up they'll find themselves on the outside, looking in. Bans are NOT off the table here In other words KNOCK IT OFF CHILDREN!

Thank you to those that stayed on topic and didn't get tied up in the offtopic bickering, you know who you are.

UPDATE: Off topic bickering deleted & thread cleaned up as best as possible. Back down to 2 pages. And now a general bit of advice to everyone. If you can't keep your comments on topic then keep them to yourself.

 

[stuff and nonesense]

You need to be a little careful with the prices it claims. I actually had to go back and fix two of the prices in my table (before you saw it), because two of them were from like Amazon marketplace sellers who almost certainly aren't official resellers. That has implications on warranty coverage, especially if they turn out to be selling OEM tray, rather than retail-boxed.

So, you have to take the next step of checking out the lowest prices and make sure they're officially sold by the online store and not a 3rd party seller. Those are the prices we should use, in order to keep things fair and consistent.

I have in the past bought equipment for builds. Part of the spec was traceability. Third party reseller is a nice way of saying “grey market”. These parts may be, if you are lucky, warranted by the seller. They won’t be warranted by Intel/AMD etc. unless it’s a one off fix from the bottom of their hearts.

Normally it’s not a big problem but right now if I were buying intel I’d pay the extra and buy from a franchised reseller.