News AMD announces Zen 5 Ryzen 9000 processors launches in July — four new Ryzen 9, 7, and 5 processors with a 16% IPC improvement

[CelicaGT]

>I missed the article about the laptop processors on Toms, which bit_user helpfully linked.

That's another issue with the THW layout, that "main" news pieces are quickly pushed off the front page by filler pieces, even on the same day. The reader has to hunt for what scrolled off--which few people bothered to.

Ironically, Ars has the opposite problem. Its content volume has dropped drastically. I check on it about once a week.

Ars gets about 2 or 3 new articles a day on average, there's also lots more third party contributions which can really vary in quality. I usually check in daily still, but I don't always read everything. Tom's I let go a few days between check ins, as I find there is a larger portion of fluff pieces which I find uninteresting (Ars gets them too). The downside is, I tend to miss some of the good articles due to the layout here. I prefer a straight up timeline for the main page, with Topics grouped along the menu bar as Ars does. On that note, I wish like Ars, Tom's would offer a paid subscription tier so I can miss the ads. I'd love to legit support the site without dealing with ad tracking...

 

[CelicaGT]

Same with Anandtech. Their layout allows for more to show on the front page, but the volume is so low that you could go a month between visits with very little risk of missing anything. They do still cover some things Toms doesn't, but I'm now down to a weekly visitor.

TBH I'm flat out surprised AT is still going. When Anand was still at the helm it was my go to tech site, with Tom's at a close second. I do check back there once in a while but it seemed mostly a grave yard with obscure networking and mobile CPU articles. Looking today, it does seem like there's more going on there than before.

 

[jordanbuilds1]

imo, seeing that amd just released the same core count chip on the am4 plat, unless your in dire need of a new cpu, theres no real point to an upgrade. also if you are on an am5 plat and looking to upgrade, same gen chips should be much cheaper than what these will prob be priced at. the only use cases i can really see for these chips are the one i mentioned above and maybe a later on upgrade from am5. am4 again released the same core count chips on that plat, and older am5 chips judging that its fairly new is still really good value to new builds. also from what ive heard, the 14900ks is still ahead of all these chips so theres that for people who do cpu intensive tasks. but it comes down to this: price and preformance, we still have yet to get our hands on these things so we might aswell wait and see.

 

[kerberos_20]

AMD is intent on continuing with AM4 for budget, simply because DDR4 is cheaper. 7000 series will be phased out once 9000 comes online. For bargain hunters, there should be buying opportunities when inventory clearance hits.

well technicaly, they are using slower CCXs in multi CCD options, so they can sell high core CPUs with one low end CCX in it...dont forget about epycs, they need to fill them aswell

 

[TheSecondPower]

I'm impressed. Zen 5 is supposed to be the biggest redesign since Zen 1 so I was expecting a bit more of an IPC increase. I also had low expectations when I saw that TSMC N4 would be used instead of N3. But 16% IPC with TDPs that are in the neighborhood of 30% lower sounds impressive with only a minor node change. It looks like the focus this generation was on improving power efficiency for mobile and server, but on desktop we get the power savings and 16% more performance.

If Intel's Lion Cove cores are this much better than Golden Cove cores, then Arrow Lake on Intel 20A should be quite impressive. Although I think Intel would be talking up Lion Cove a lot more if it was this good.

 

[kerberos_20]

just noticed, those comparison slides with 14900k vs 9950x....on intel side its running intel default power plan..which intel said isnt recommended power profile to run intel cpu in spec, but just tempfix for instability

so maybe not so great upgrade afterall?

 

[Co BIY]

That's another issue with the THW layout, that "main" news pieces are quickly pushed off the front page by filler pieces, even on the same day. The reader has to hunt for what scrolled off--which few people bothered to.

Crazy to see truely premium Tom's content, like in-depth multi-page GPU/CPU reviews that took days or weeks of effort to produce , get bumped off the "Front page" by paragraph length copy/paste articles that took 15 minutes to produce.

 

[KnightShadey]

Yeah, Harold likes to cherry-pick benchmarks. He does not look at things holistically or integrate diverse datapoints into a cohesive picture. The problem is that if a single benchmark could tell the whole story, we'd only need one!

All we really need is Bungholiomarks. 😈 😉

I will say, somewhat related regarding THG/Anand discussion, my brief return to both (and discovering HarOCP is no more [except the remnants in the forum]) was in search of deeper-dives to get beyond the presentation/slides PR-speak (which are accurate within a strict view), but seems like that will have to wait a few days/weeks. This feels like the old paper launches with the lack of detail. 🤬

Also to your previous point, I agree, the mobile segment is the bigger leap this release , where there are major difference in options/limitations vs smaller incremental desktop/workstation changes that are easier to ride out on AM5 and with add-ons/swaps. This is even worse when you're and edge case of an edge case use-model.

Hopefully we get closer to the metal & its potential implications before product availability drastically changes (ex last of the 7945HX/3D before the long wait to StrixHalo ) and before budget changes and gets refocused . 🤞

 

[bit_user]

in search of deeper-dives to get beyond the presentation/slides PR-speak (which are accurate within a strict view), but seems like that will have to wait a few days/weeks. This feels like the old paper launches with the lack of detail. 🤬

Well this just went up mere minutes ago:

https://www.tomshardware.com/pc-com...-ipc-gain-for-e-cores-16-ipc-gain-for-p-cores
​

If you want even more detail, check out ChipsAndCheese:

Thoughts on Skymont Slides

Recently a set of Skymont slides were published, apparently without any embargo or confidentiality marks.

chipsandcheese.com

Intel’s Lion Cove Architecture Preview

Intel’s P-Core’s lineage can be traced all the way back to the P6 architecture that was originally found in the Pentium Pro.

chipsandcheese.com

There should be some good deep dives after Hot Chips, but that's not until late August.

 

[thestryker]

I will say, somewhat related regarding THG/Anand discussion, my brief return to both (and discovering HarOCP is no more [except the remnants in the forum]) was in search of deeper-dives to get beyond the presentation/slides PR-speak (which are accurate within a strict view), but seems like that will have to wait a few days/weeks. This feels like the old paper launches with the lack of detail. 🤬

With the rise of techtubers and more entertainment focused tech media all hardware companies have been less transparent. As @bit_user said don't expect much of anything until either post Hot Chips or near product launches whichever happens first.

 

[TheHerald]

Again not true at all. Read ALL the benchmarks and you will see that the 7950X at 65W TDP more times than not is faster than 13900k at 105W TDP. Every benchmark has the 105W TDP 7950X being faster than the 13900k at all TDPs EXCEPT stock (253W).

Ah, I knew it was this one.

Only one problem, the 7950x at 65w TDP was drawing 90 watts, not 65. I have the CPU, it tdoes NOT score anywhere near 31k in cbr23 at 65w. It's more like 21-22k

Regarding the 5950x

Spoiler

 

[bit_user]

Ah, I knew it was this one.

Only one problem, the 7950x at 65w TDP was drawing 90 watts, not 65.

That was Y-cruncher, not sure how applicable it is to the other benchmarks. More importantly, it's peak power. They never said how long that 90W peak was sustained.

They go on to mention something that sounds awfully suspicious:

"For example, restricting the socket and 7950X to 125 W yielded a measured power consumption that was still a whopping 33% higher. By comparison, the 13900K exceeded its set limits by around 14% under full load."
​

So, the first part is suspicious, because 33% seems too close to the 35% PPT ratio to be a coincidence. Whether it was operator error or maybe a bug in the BIOS they were using, it sure seems like they restricted TDP and not PPT.

The second thing to point out is that even Intel is going over by nearly half as much. So, neither are operating according to their graphs, but they sadly didn't provide us with any average power (or total energy) figures that we could use to know for sure.

 

[TerryLaze]

That's not what IPC means. This is probably where you went off track.

You can call it whatever you want, the bottom line is that you will only be 16% faster than the previous gen when using at least the same amount of power as the previous gen.
If the ppt is lowered it will eat into the speed improvement, no matter if it is called ipc or lollypop.

 

[bit_user]

You can call it whatever you want, the bottom line is that you will only be 16% faster than the previous gen when using at least the same amount of power as the previous gen.

That doesn't follow from the data provided. Again, what's 16% is the typical ISO-clock performance, not ISO-power.

 

[TheHerald]

That was Y-cruncher, not sure how applicable it is to the other benchmarks. More importantly, it's peak power. They never said how long that 90W peak was sustained.

They go on to mention something that sounds awfully suspicious:

​

"For example, restricting the socket and 7950X to 125 W yielded a measured power consumption that was still a whopping 33% higher. By comparison, the 13900K exceeded its set limits by around 14% under full load."​

​

So, the first part is suspicious, because 33% seems too close to the 35% PPT ratio to be a coincidence. Whether it was operator error or maybe a bug in the BIOS they were using, it sure seems like they restricted TDP and not PPT.

The second thing to point out is that even Intel is going over by nearly half as much. So, neither are operating according to their graphs, but they sadly didn't provide us with any average power (or total energy) figures that we could use to know for sure.

But I already know how fast the 7950x is at actual 65w in CBR23. It scores exactly 22k. I don't need some graphs made on excel to tell me that, I've actually tested it

 

[Ogotai]

I've actually tested it

you keep saying you have tested all of these things, but i dont recall you ever posting screenshots or anything to show it. your claims are pointless. show proof, or drop it.

cherry picking other graphs and such, proves nothing when others have shown you are wrong, or habe shown you are misinterpreting what you are reading.

 

[TheHerald]

you keep saying you have tested all of these things, but i dont recall you ever posting screenshots or anything to show it. your claims are pointless. show proof, or drop it.

cherry picking other graphs and such, proves nothing when others have shown you are wrong, or habe shown you are misinterpreting what you are reading.

I've shown proof - not just screenshots but videos - numerous times of eg. 12900k beating 7800x 3d in gaming, 14900k capped at 90w in games being faster then the 3d etc. People just don't believe it cause "random youtuber" or what have you, they prefer graphs made on excel than actual in game footage. What can you do man.

If you think that the 7950x score 31k at 65w in CBR23 okay, im not going to try to change your mind. Believe away

 

[Colif]

Same with Anandtech. Their layout allows for more to show on the front page, but the volume is so low that you could go a month between visits with very little risk of missing anything. They do still cover some things Toms doesn't, but I'm now down to a weekly visitor.

funny how both sites owned by the same company. One day they might combine and make either best or worst front page design.

 

[bit_user]

funny how both sites owned by the same company. One day they might combine and make either best or worst front page design.

I hope not. It would undoubtedly compromise the coverage Anandtech still does that Tom's doesn't. Also, I find Anandtech has a better search engine and it's easy to find their articles going back 2 decades or more. I find myself referring to old articles on their site, with some regularity.

 

[Colif]

that is a possibility. Every site is different so who can say. Its not the same writers everywhere. Easier to search Google than our search.

 

[KnightShadey]

With the rise of techtubers and more entertainment focused tech media all hardware companies have been less transparent. As @bit_user said don't expect much of anything until either post Hot Chips or near product launches whichever happens first.

I agree for the desktop part which is more of a typical Computex heads-up launch prior to later true launch and availability a month or so later; but Strix laptops will be available next month (funny enough ASUS' QR-code went live right after the presentation, then died before day's end a coupla hours later).

I kinda expected the typical tick-tock media release of Launch event, NDA expires same day (or shortly thereafter) to media dump the pre-release deeper dives not coverable in a 1-2 hr orchestrated presentation (depending on # of products) then shipping parts whenever. Oh well.

 

[jeremyj_83]

You can call it whatever you want, the bottom line is that you will only be 16% faster than the previous gen when using at least the same amount of power as the previous gen.
If the ppt is lowered it will eat into the speed improvement, no matter if it is called ipc or lollypop.

There is no data to support your claim. The 16% gain means that at the same clock speed on average Zen 5 will be that much faster than Zen 4. However, if the uArch is more efficient in addition to node efficiency that can mean it doesn't need the same amount of power to hit the same clock. Therefore you can lower the PPT and still get the same clocks. We also don't know of Zen 5 is able to boost more cores to higher clocks, if the boost profile allows for longer boosts, etc... All of those play a role in absolute performance increase in addition to IPC gains. Like how Zen+ was only a 3% IPC gain but 10-15% faster than Zen 1? That was due to the boost behavior.

 

[KnightShadey]

Well this just went up mere minutes ago:

​

https://www.tomshardware.com/pc-com...-ipc-gain-for-e-cores-16-ipc-gain-for-p-cores​

​

If you want even more detail, check out ChipsAndCheese:

​

Thoughts on Skymont Slides

Recently a set of Skymont slides were published, apparently without any embargo or confidentiality marks.

chipsandcheese.com

​

Intel’s Lion Cove Architecture Preview

Intel’s P-Core’s lineage can be traced all the way back to the P6 architecture that was originally found in the Pentium Pro.

chipsandcheese.com

​

There should be some good deep dives after Hot Chips, but that's not until late August.

Yeah, that's more like it. Though need to look at it all later. Doubtless it raise as many questions as it answers, especially since it looks optimistic for intel, but better starting point that the Z5 information-wise.

Thanks. 🤙

 

[jeremyj_83]

I've shown proof - not just screenshots but videos - numerous times of eg. 12900k beating 7800x 3d in gaming, 14900k capped at 90w in games being faster then the 3d etc. People just don't believe it cause "random youtuber" or what have you, they prefer graphs made on excel than actual in game footage. What can you do man.

If you think that the 7950x score 31k at 65w in CBR23 okay, im not going to try to change your mind. Believe away

Are you a PROFESSIONAL REVIEWER? No. Of course people aren't going to believe you. You could easily change a setting or two to show results that YOU want. Just like I'm sure I could figure out a way to make my almost 11 year old 4770k faster than a 7800X3D in gaming. People who do this professionally, while they might have bias toward AMD or Intel, have to be as objective as possible. If they aren't then they are out of a job.

 

[jeremyj_83]

But I already know how fast the 7950x is at actual 65w in CBR23. It scores exactly 22k. I don't need some graphs made on excel to tell me that, I've actually tested it

Random forum blog poster vs professional reviewer. I'll trust the professional.