News AMD Zen 4 Ryzen 7000 Specs, Release Date Window, Benchmarks, and More

[tek-check]

iGPU getting in the way again. Way to go Intel. Not smart at all an iGPU getting in the way of more good cores when enthusiasts do not care about the iGPU.

Yeah I know enthusiasts not the majority of Intel's market. But these iGPUs are coming on retail boxed i9 CPUs which are designed for enthusiasts. The ones that are not are sold to OEMs and not in retail boxes at Micro Center, NewEgg, Amazon, and etc and marketed towards enthusiasts. If Intel really did care about enthusiasts, they need to dump e cores and iGPU and focus on P cores or otherwise quit selling retail boxed Core i9s which are marketed towards PC building enthusiasts most of whom do not care about the iGPU nor the e-waste cores. And keep iGPU and e-waste cores for the stuff sold to OEMs where it makes more sense.

Do not blame it on iGPU. It's basic anyway. Blame it on power consumption. 12P cores would be super difficult to cool as Intel will not reduce power consumption.

 

[msroadkill612]

IMHO it explains a lot why AMD seem to have been tepid about TR - tying their desktop architecture so closely to Epyc (one solution for all goes w/ the Fabric architecture territory) means desktop is dragged inexorably upmarket - perhaps a little too much.

zen4 desktop is almost embarrassingly powerful for plain folks

 

[tek-check]

IMHO it explains a lot why AMD seem to have been tepid about TR - tying their desktop architecture so closely to Epyc (one solution for all goes w/ the Fabric architecture territory) means desktop is dragged inexorably upmarket - perhaps a little too much.

zen4 desktop is almost embarrassingly powerful for plain folks

They will need to reignite HEDT/workstation segment either with Zen4c or Zen 5. Zen4c is rumoured to be a smaller 16 core chiplet for Bergamo SKUs up to 128 cores. Two of those in desktop segment could bring 32 cores.

In 2024, they could also put three Zen 5 chiplets, so up to 24 big cores on top SKUs such as 8960XW 8970XW etc., and 1-2-3 small Zen4c 8 core chiplets, for a total of 32, 40 and 48 big-little core configurations.

I agree that 7950X with X670E is already, as the author puts it, quasi-HEDT. For enthusiasts, more I/O is needed, so next revision of AsMedia P21 chipset will have to run at PCIe 5.0 link speed to CPU. Plus, some vendors can add PCIe 5.0 PLX switch for more flexibility and peripherals. There is enough room to work with for most people on AM5 platform.

 

[-Fran-]

You are missing the woods from the trees...
Yes the marketing slide shows an extreme, a niche that not many people will use.
Intel is clearly telling you that if you don't use any CPU heavy software alongside your game then the e-cores aren't going to give you anything extra.
How is that a bad thing, they are telling you when it makes sense to get a CPU with e-cores instead of a CPU without them.

Also most live streamers still prefer CPU encoding at a somewhat higher quality level because they get a better quality/bandwidth ratio than hardware encoding, especially if they have to use the gpu for it because that reduces the FPS the GPU can get.
If you are just recording you can use hardware acceleration and a very high bitrate to get very good results but if you had to upload those at real-time you would have a big problem.

Ok, let's take a step back then, because I'm awfully confused now XD

So we're talking about the same thing (mostly) and the only difference is that you do like the E-cores as a "just in case you need them" thing and I think using that space for more P-cores would've been better instead; is that the only difference in our arguments?

If so, just to clarify, I just wish Intel would've given people the chance to choose, like they did with the 6P-core parts. As I mentioned in several places now: there's a reason the 12400 is a "fan favorite" now, right? My argument around having more P-cores is better than having the mix for a few reasons, but I'll concede they're not "a big deal" for 100% of gamers (and streamers to a degree). The reason being is that, I project, the E-cores are going to lose their edge way faster than the P-cores will in terms of performance. The move to bigger caches is going to hamper their performance heavily in the short term and you can see that in the increase in cache with Raptor Lake. But well, that's just mere speculation.

Regards.

 

[TerryLaze]

Ok, let's take a step back then, because I'm awfully confused now XD

So we're talking about the same thing (mostly) and the only difference is that you do like the E-cores as a "just in case you need them" thing and I think using that space for more P-cores would've been better instead; is that the only difference in our arguments?

No, it's about some people just saying that more "real" cores are better... instead of adding ...for my use case.

If all you do is DC then more P cores are the better thing because even at very low clocks they will give you much better performance.

If you do something real-time sensitive in the foreground (gaming etc) but also often do something heavy in the background then big little is much better for your real-time experience. The 5950x drops to 3.7Ghz if you load all cores, so even if you separate the game to one ccx and whatever else to the other ccx you still get greatly reduced FPS in gaming and an intel CPU with the same amount of p-cores wouldn't be any different if not even worse. The 12900k will retain the p-core clocks on the p-cores no matter how many cores are loaded.

If you only do something real-time sensitive then you neither need a huge amount of cores or e-cores, 6 to 8 normal cores is great for just that.

People seemingly can't understand that people choose what they need.

If so, just to clarify, I just wish Intel would've given people the chance to choose, like they did with the 6P-core parts.

Sure if intel could make a billion SKUs every year to cater to every one specifically, that would be great.
As things are intel doesn't have to make a CPU with more P-cores since AMD already makes that and it's not like intel is losing any money, in the contrary if intel would use more p-cores it would cost them more money.

The reason being is that, I project, the E-cores are going to lose their edge way faster than the P-cores will in terms of performance. The move to bigger caches is going to hamper their performance heavily in the short term and you can see that in the increase in cache with Raptor Lake. But well, that's just mere speculation.

Regards.

How is bigger cache supposed to reduce the performance of the e-cores?
Because the p-cores are going to become faster?! The e-cores will still be performance on top of that.

 

[KyaraM]

Ok, let's take a step back then, because I'm awfully confused now XD

So we're talking about the same thing (mostly) and the only difference is that you do like the E-cores as a "just in case you need them" thing and I think using that space for more P-cores would've been better instead; is that the only difference in our arguments?

If so, just to clarify, I just wish Intel would've given people the chance to choose, like they did with the 6P-core parts. As I mentioned in several places now: there's a reason the 12400 is a "fan favorite" now, right? My argument around having more P-cores is better than having the mix for a few reasons, but I'll concede they're not "a big deal" for 100% of gamers (and streamers to a degree). The reason being is that, I project, the E-cores are going to lose their edge way faster than the P-cores will in terms of performance. The move to bigger caches is going to hamper their performance heavily in the short term and you can see that in the increase in cache with Raptor Lake. But well, that's just mere speculation.

Regards.

The xx400 CPU has been a "fan favorite" since at least 10th gen, it got jack to do with not having e-cores.

The argument is that no, e-cores aren't wasted because lower clock/IPC. If you don't need them, you likely wouldn't need more p-cores, either. They are extra power for when they are needed, sime as that.

 

[-Fran-]

No, it's about some people just saying that more "real" cores are better... instead of adding ...for my use case.

If all you do is DC then more P cores are the better thing because even at very low clocks they will give you much better performance.

If you do something real-time sensitive in the foreground (gaming etc) but also often do something heavy in the background then big little is much better for your real-time experience. The 5950x drops to 3.7Ghz if you load all cores, so even if you separate the game to one ccx and whatever else to the other ccx you still get greatly reduced FPS in gaming and an intel CPU with the same amount of p-cores wouldn't be any different if not even worse. The 12900k will retain the p-core clocks on the p-cores no matter how many cores are loaded.

If you only do something real-time sensitive then you neither need a huge amount of cores or e-cores, 6 to 8 normal cores is great for just that.

People seemingly can't understand that people choose what they need.

Not quite. I can actually keep 1 CCD running at it's peak frequency while the other one at a lower speed. It doesn't have the same granularity as Intel's per-core multiplier, but in terms of multiple CCDs, you can do that. I have now a 5900X and I can keep one CCD at 5.1Ghz and the second CCD at 4.7Ghz no problem running full bore. Just to clarify, as I said, you can also do this with Intel, but with better granularity. By default, AMD will always keep 1 CCD running at full speed and throttle the second one until the power or temperature cap is met. What you said is valid for single-CCD Ryzens though. You're at the mercy of the PBO algorithm there, but for the most part, most Ryzen single-CCD CPUs can keep their max speeds with no problem all the time.

Sure if intel could make a billion SKUs every year to cater to every one specifically, that would be great.
As things are intel doesn't have to make a CPU with more P-cores since AMD already makes that and it's not like intel is losing any money, in the contrary if intel would use more p-cores it would cost them more money.

Well, on one hand they could still have AVX512 enabled with all P-cores 😆

How is bigger cache supposed to reduce the performance of the e-cores?
Because the p-cores are going to become faster?! The e-cores will still be performance on top of that.

Because the E-cores have no cache and whenever programs that rely con cache overflow to them, they'll hurt performance. That being said, and as I said, that's just speculation and there's no many programs that really need high amounts of cache to perform well.

The xx400 CPU has been a "fan favorite" since at least 10th gen, it got jack to do with not having e-cores.

The argument is that no, e-cores aren't wasted because lower clock/IPC. If you don't need them, you likely wouldn't need more p-cores, either. They are extra power for when they are needed, sime as that.

To be pedantic, no. The x400 line hasn't been even a thing before 10th gen because of the lack of Hyper Threading and full memory speeds support. The 9400 and before were 6c/6t (or 4c/4t) CPUs and even when introduced, they were already bottlenecking GPUs in MP games; AMD actually had a slight lead up to the 11400 with the 3600 and 5600X under some circumstances. Particularly, for the 12th gen, the 12400 is a great CPU mainly because of the* whole platform and pure performance.

As for the E-cores. I don't know what else to say. It's just a matter of preference at this point, so I won't keep arguing about them.

Regards.

 

[Wolverine2349]

The xx400 CPU has been a "fan favorite" since at least 10th gen, it got jack to do with not having e-cores.

The argument is that no, e-cores aren't wasted because lower clock/IPC. If you don't need them, you likely wouldn't need more p-cores, either. They are extra power for when they are needed, sime as that.

The e-cores drag the ring clock down so yes they hurt. Nevermind scheduling issues with WIN10 and WIN11 sucks big time!! The e-waste cores are absolute crap if a task gets stuck on them and will not know what to do running on something so much slower with terrible latency dragging whole system down.

There are needs and wants for more than 8 P cores. I think of the 12700K and 12900K as 8 core 16 thread chips as the e-waste cores are absolute crap in things today except being Cinebench accelerator or some productivity apps that scale to infinite threads as long as the buginess crap of hybrid arch does not get in the way.

1. Are the e-cores helpful: Depends on use case, but in most cases I would say no. Sure for a more affordable system in productivity workloads adding lots of e-cores can help for apps that scale to as many threads as you throw at them. But even better yet a Threadripper system with a bunch of strong cores would demolish that though be more expensive.

2. Majority of apps are becoming more threaded slowly, but still scale to finite number of threads. Many apps use 8 to 11 threads. and guess what, having more strong cores up to the number of threads which in this case 12-16 is sweet spot for maximum on high end enthusiast non-HEDT/Server class platform like AMD provides and can increase performance a lot and will blow performance of a ton of e-waste cores out of the water.

 

[Wolverine2349]

View: https://www.youtube.com/watch?v=frSsNi25DdE


Well sadly it is actually looking like Zen 4 may be a bit underwhelming. I hope to be wrong, but some of the comments in this video are not encouraging and its so close to release what else could change??

The fact that they mention how the 35-40% multi threaded uplift with the 7950X over 5950X is primarily because the 5950X loading all 16 cores reached thermal and power limits and had to clock lower which the AM5 platform is supposed to resolve. They mention also that the 5900X is sweet spot and already clocks higher which is why it is better for gaming and how the upgrade for parts lower than 5950X flagship will not be even close to that much performance improvement.

Even the 5950X manually or well tuned with PBO can exceed great clocks even if it exceeds the supposed power limits of AM4 142 watts (more like a guideline, but good motherboards well built with good to great VRMs are fine pushing much more than 142 watts through AM4 to a good CPU) are fine on AM4. Its not any issue if you have excellent water cooling and some have 5950X hitting 250 watts to 300 watts under fine temps with great cooling. How much better is Ryzen 7950X going to be then at stock compared to a well tuned 5950X.

And will a 7950X be able to be well tuned especially the early releases which may not be well binned?

And the non flagship parts uplift may be underwhelming. An 8% IPC uplift does not look that good and does not bode well unless you can easily hit insane 5.3GHz or more clock speeds all cores all workloads all the time on 7900X.

Trying to get overclock on air much past 4.5 to 4.6GHz and even that does not always work is so hard on Zen 3 already. A well binned CCD in a chip I can get 4.7GHz CCD1 (the good one) and 4.525GHz CCD2 (the maybe average to above average one for my 5900X and pass all stress tests including Linpack XTREME even on good air cooling.

It seems Zen 4 is more about new platform, adding AI accelerator and AVX512 support and memory bandwidth with DDR5 than actual IPC uplift and even overall performance uplift for all core workload with a manual static overclock.

We still do not know, though the arch is the least changed since Zen to Zen+ as it still has 2 6-8 core chiplets with only doubling L2 cache other than new platform socket and DDR5 and AVX512 added. If you do not need insane memory bandwidth nor AVX512 nor an AI accelerator it may be hardly an upgrade at all and in fact in latency sensitive RAM things could be worse.

My hope was for a platform that I could get more than 8 good cores with an IPC uplift equal to or better than Golden Cove with at least 5GHz or maybe more all core clock all workloads while using less power and lower temps than Alder Lake at 5GHz.

And Raptor Lake seems very underwhelming too from Intel compared to Alder Lake unless you are one of the ones who like those peasant e-cores. The top flagship i9 just has 8 more e-cores which are like Cinebench accelerators to push up Cinebench scores. While lower SKU 13700K is just the 12900K And lower SKUs just have more e-waste cores and same amount of P-cores. And the P-cores just have mild at best IPC uplift and a little better clocks.

I hope to be wrong on Zen 4 underwhelming IPC gains and wrong on Intel only in that they actually offer more than 8 P cores on Raptor Lake. There is nothing to indicate though that Intel will offer more than 8 p cores. With AMD, maybe they are sandbagging.

But not so sure in either camp. Its all marketing for flagship products that may be underwhelming no matter which camp you are in e-cores hybrid arch or not. For AMD, 5950X was power and thermal constrained so they used it to tout 7950X as having huge multi threaded uplift when in reality its because it will auto clock much higher where as lower SKUs do not. With Intel, it is flagship 13900K that has an extra 8 e-cores to increase multi threaded performance in tasks like Cinebench that scale to infinite threads. Lower SKUs from AMD like 7900X were not power/thermal constrained like the 5900X so potential underwhelming small itty bitty performance uplift for lower SKUs. And Intel 13700K is right back to e core and P core count of 12900K so itty bitty perf uplift potentially.

I wonder if these new SKUs are meant for new buyers with mild at best perf uplift and AMD and Intel are going to discontinue all sales of older gen which has almost as good of performance for far less money especially in AMD case where you will not have to do a new platform.

 

[-Fran-]

To be honest, the IPC increase on Zen4 is not "underwhelming" per se. Well, even more, we haven't actually seen proper bechmarks of it. The* rumoured ~8% increase is still "fine" if you consider the increase in clocks. I wouldn't say it's a bad thing as long as the overall performance uplift is above 15%. Sure, you most of the time want IPC, since it's a more efficient design, but what can AMD do when the whole industry is going up in power. There's only so much you can do at certain clock speeds until you increase the L3/L4 caches and/or widen the pipeline. Or to phrase it differently, AMD has going up very slowly in clock speeds to keep power under control, but now they have openly said "screw power, let's just turn it up to eleven!", which, TBH, I'm not a fan of... But what can you realistically expect? Le sigh.

Regardless, the IPC figure should be put in the same context as the speed increase, just like with Intel. If you look at Alder Lake P-cores clock per clock versus previous uArchs is around ~10% on average (for games) and no one says Alder Lake P-cores are bad, since all improvements around the platform (including clock and IMC) net around a solid 15%-30% (depending where you look), which is nice for sure.

I think there's a lot of room for the AM5 platform to grow, but I will concede that Zen4 needs to be a good showing. And, even more than the pure CPU prowess, I'm more afraid of DDR5 still being too darn expensive XD

Regards.

 

[drivinfast247]

The question is if Socket AM5 will be able to attract Socket AM4 users, and I don't think it will. Let's say it's on average 25% faster per thread than the Ryzen 5000 series. The cost of a new motherboard and RAM in addition to the CPU makes that 25% cost an awful lot. Even a Zen 2 or Zen 1 user can drop in a Zen 3 CPU into their existing setup for a large to enormous performance gain for a relatively small price.

Also assuming that AMD is going to attempt to keep the same pace as Intel, the next 3 or 4 years should bring relatively decent to large increases in performance and efficiency across all segments each year, and the last thing most people want is to buy in at the start only to see each generation bring a large performance increase, as those of us who were Socket AM4 early adopters saw.

Combine this with increasing GPU supply and more people finally being able to see what their existing Socket AM4 setup can actually do, and I predict quite sluggish sales of AM5 systems.

I don't see it being a problem as Intel has been doing this for over a decade with less of a performance increase.

 

[alceryes]

Yes, ask yourself, why did the reputable reviewers not exactly recreate what intel claimed to see if it's true or not? Shouldn't that be the first thing a reputable source would do?

If Intel shared the exact scenario (specific details) needed to reproduce the performance results they claim, you bet your boots reputable reviewers would go through that exact scenario to make sure Intel's numbers matched.

Sadly, they don't. (which is normal for these kinds of wow! performance marketing slides)

 

[KyaraM]

Not quite. I can actually keep 1 CCD running at it's peak frequency while the other one at a lower speed. It doesn't have the same granularity as Intel's per-core multiplier, but in terms of multiple CCDs, you can do that. I have now a 5900X and I can keep one CCD at 5.1Ghz and the second CCD at 4.7Ghz no problem running full bore. Just to clarify, as I said, you can also do this with Intel, but with better granularity. By default, AMD will always keep 1 CCD running at full speed and throttle the second one until the power or temperature cap is met. What you said is valid for single-CCD Ryzens though. You're at the mercy of the PBO algorithm there, but for the most part, most Ryzen single-CCD CPUs can keep their max speeds with no problem all the time.


Well, on one hand they could still have AVX512 enabled with all P-cores 😆


Because the E-cores have no cache and whenever programs that rely con cache overflow to them, they'll hurt performance. That being said, and as I said, that's just speculation and there's no many programs that really need high amounts of cache to perform well.


To be pedantic, no. The x400 line hasn't been even a thing before 10th gen because of the lack of Hyper Threading and full memory speeds support. The 9400 and before were 6c/6t (or 4c/4t) CPUs and even when introduced, they were already bottlenecking GPUs in MP games; AMD actually had a slight lead up to the 11400 with the 3600 and 5600X under some circumstances. Particularly, for the 12th gen, the 12400 is a great CPU mainly because of the* whole platform and pure performance.

As for the E-cores. I don't know what else to say. It's just a matter of preference at this point, so I won't keep arguing about them.

Regards.

Weird, then, that I find especially the 10400 praised as a great budget gaming CPU.

https://www.pcmag.com/reviews/intel-core-i5-10400

Intel Core i5-10400F Review - Six Cores with HT for Under $200

Intel's new Core i5-10400F offers a large performance jump over the previous generation Core i5-9400F because of its six-core/twelve-thread design. In this Core i5-10400F review we also test the feasibility of overclocking through BCLK, or by relaxing the PL1 and PL2 Turbo Limits.

www.techpowerup.com

And more...
Also, why are you talking about the 9400 when I'm talking about the 10400 here, and said so?

 

[-Fran-]

Weird, then, that I find especially the 10400 praised as a great budget gaming CPU.

https://www.pcmag.com/reviews/intel-core-i5-10400

Intel Core i5-10400F Review - Six Cores with HT for Under $200

Intel's new Core i5-10400F offers a large performance jump over the previous generation Core i5-9400F because of its six-core/twelve-thread design. In this Core i5-10400F review we also test the feasibility of overclocking through BCLK, or by relaxing the PL1 and PL2 Turbo Limits.

www.techpowerup.com

And more...
Also, why are you talking about the 9400 when I'm talking about the 10400 here, and said so?

I mentioned the 9400 to be clear that it is something new and not an ongoing theme. There's been good i5s and bad i5s, specially in the "x400" stepping/SKU.

Now, the problem for 10th gen and 11th gen was that Ryzen 3K and 5K were no slouches at similar price points and they, arguably, offered a better platform at a lower/similar cost. Also, price drops and late SKUs.

From their (your link) conclusion:

Compared to AMD's Ryzen 3600 and 3600X, the 10400F is slightly slower, by 4% and 6% respectively. It depends very much on the workload though, especially tasks that are easy to parallelize, like rendering, are AMD's strongest suit, and Intel has a clear lead in single and low-threaded apps, which are relevant to the majority of consumers today.

While the 10400F edged the 3600 in gaming, everything else around it was a minus, including the lack of OC in both RAM and CPU. Unlike the 12400, which is overall better, including platform cost with certain tradeoffs, the previous generations were battling new CPUs from AMD.

As I said, I was nitpicking

Regards.

 

[mamasan2000]

The question is if Socket AM5 will be able to attract Socket AM4 users, and I don't think it will. Let's say it's on average 25% faster per thread than the Ryzen 5000 series. The cost of a new motherboard and RAM in addition to the CPU makes that 25% cost an awful lot. Even a Zen 2 or Zen 1 user can drop in a Zen 3 CPU into their existing setup for a large to enormous performance gain for a relatively small price.

Also assuming that AMD is going to attempt to keep the same pace as Intel, the next 3 or 4 years should bring relatively decent to large increases in performance and efficiency across all segments each year, and the last thing most people want is to buy in at the start only to see each generation bring a large performance increase, as those of us who were Socket AM4 early adopters saw.

Combine this with increasing GPU supply and more people finally being able to see what their existing Socket AM4 setup can actually do, and I predict quite sluggish sales of AM5 systems.

I went from Ryzen 1700 to a 5600X. I saw maybe 5-10 fps difference IIRC. Then I went from a Vega 56 to a Nvidia 2080, I saw 50% more fps. This was huge! In my case, the CPU change did hardly anything. Same motherboard, Asus X470. Upgraded RAM from 3000 mhz to 3600 Mhz.
Your mileage may vary.

The main thing I would look at is DDR5. How fast kits are out there? Are they getting any faster than they have out now? How is AMDs support for DDR5? Is it like Ryzen 1s initial support of 3000 Mhz RAM(very poor)?
I don't need anything upgraded for a couple of years. And who knows how the economy goes. With some decent hyperinflation and gold prices going up and up. CPUs contain gold...who knows, in 2 years the cheapest CPU could be 6000 dollars.

 

[KyaraM]

I went from Ryzen 1700 to a 5600X. I saw maybe 5-10 fps difference IIRC. Then I went from a Vega 56 to a Nvidia 2080, I saw 50% more fps. This was huge! In my case, the CPU change did hardly anything. Same motherboard, Asus X470. Upgraded RAM from 3000 mhz to 3600 Mhz.
Your mileage may vary.

The main thing I would look at is DDR5. How fast kits are out there? Are they getting any faster than they have out now? How is AMDs support for DDR5? Is it like Ryzen 1s initial support of 3000 Mhz RAM(very poor)?
I don't need anything upgraded for a couple of years. And who knows how the economy goes. With some decent hyperinflation and gold prices going up and up. CPUs contain gold...who knows, in 2 years the cheapest CPU could be 6000 dollars.

From what I gathered, both Intel and AMD should support DDR5 5200 RAM next gen. That's not yet faster than good DDR4, but much better than DDR5 4800 Intel offers right now, and of course it is always possible to use faster RAM. Iirc, DDR5 starts to get better at either 5800 or 6000 right now, but that will of course change over time. Right now, I'm very glad to have bought my new system this year, too, when total system costs are still kinda reasonable.

 

[Forge64]

Thank you for this summary article.

My hope as a 3960X user: INTEL will eventually be able to put more than 8 performance cores in one package and be competitive again in HEDT so that AMD feels the need to release a non-PRO version of a Threadripper 7000 series. I need the cores, the PCIe lanes and the memory channels

It's not going to happen. Threadripper was a last-second panic move by AMD to take the halo HEDT product segment by force, the HEDT Threadrippers were just downmarked EPYC CPUs and banged-together enthusiast boards on downmarked workstation chipsets. AMD didn't lose money on those, but they sure didn't make as much as they liked. The very minute that Ryzen could reach the performance numbers they needed without the extra chiplets and cost of Threadripper/EPYC, that whole effort was doomed. Since you value the memory and PCIe lanes that TR threw around, AMD wants you to spend a lot more and get a nice big EPYC workstation, at 2X-3X the cost or more.

 

[s1mon7]

I'm not too worried about 16 cores on the R9 competing well against the 13900k, but boy am I worried about rumors that AMD may be trying to launch a high R5 with six cores against a 14-core i5 (even if it's a 6+8 config). And on a more expensive platform at that. If that actually happens, boy are we going to see a Rocket Lake-esque reaction - except amplified by the fact that AMD likely won't have an Alder Lake-like reputation fix launching just months later.

 

[Brian D Smith]

Thank you for this summary article.

My hope as a 3960X user: INTEL will eventually be able to put more than 8 performance cores in one package and be competitive again in HEDT so that AMD feels the need to release a non-PRO version of a Threadripper 7000 series. I need the cores, the PCIe lanes and the memory channels

As someone trying to figure some things out...may I ask why you 'need the cores'?

 

[thisisaname]

A nice article but now many things is it going to be published?

 

[Brian D Smith]

Tired? Perhaps you could take a nap?
Seriously, I love the leaks...it's 'part of the process'. It also drives eyeball traffic for the site.

What I do NOT like are articles that keep getting updated...and you are not directed to what is actually being added to/changed from the article that may have been published weeks ago.

 

[aalkjsdflkj]

What I do NOT like are articles that keep getting updated...and you are not directed to what is actually being added to/changed from the article that may have been published weeks ago.

+1

It's really frustrating to try to go through these articles to find the updated content. How hard is it to include an (UPDATE) note next to new information from the last time it was published?

 

[-Fran-]

AMD leaked by accident (allegedly) the SKUs for the first batch of Ry 7K CPUs and the 8-core part will be 7700X, which is interesting for two reasons: the 8-core SKU is the "best" full-CCD silicon for OC'ing (best PBO curves and all that; silicon lotery if you will) and they clearly received the message after their lackluster 5700X launch. That thing is basically at the same price as the 5800X nowadays; what a (bad) joke. Same-ish with the 5600 non-X. They could bin the better CCDs for either X3D or just a later release for the 7800X (like the 3800XT, but less stupid).

I mean, I just hope they've learned their lessons as this will be like the 5th time they're releasing SKUs and all previous times, except the 2K gen, they made a fool of themselves in one way or another with head-scratchers, lol.

Anyway, I would also say that the iGPU being RDNA2 and having 6-8CUs is more or less fully confirmed by now? I can't remember TBH. I mention that because the article could use some proper updates in tables and other paragraphs with the new information.

Regards.

 

[bjnmail]

Interestingly, the Ryzen 7000 IHS says the chips were made and diffused in Taiwan, whereas Ryzen 5000 chips were diffused in Taiwan but made in the US.

The 3000/5000 series will say diffused in Taiwan and diffused the USA, with the "assembled in" country up in the air, either China or Malaysia, usually, but I guess there could be a packaging facility in Taiwan too, though I've never seen one of these chips.

This is because the compute dies are diffused by TSMC in Taiwan, and the 12nm IOD is diffused by GloFo in the US, hence the two different diffusal countries. Then the complete chip package is assembled in either China or Malaysia (could be others, but those are the two I've always seen).

Now that the IO die is being made by TSMC as well, you will only see the single "Diffused in Taiwan" statement on all of the 7000 series.

 

[digitalgriffin]

Here's the Ryzen 7000 release , benchmarks, specifications, pricing, and all we know about AMD's Zen 4 architecture.

AMD Zen 4 Ryzen 7000 Specs, Release Date Window, Benchmarks, and More : Read more

Upshots
Better security
Better single thread performance
Better multi thread performance
Avx512
Graphics be roughly equivalent to vega 7 on 5700g
Should keep Intel on their toes pricing wise

Druthers
Power delivery will be insane. High cost mb & AIO
High cost DDR5

That said my son's ivybridge system is getting long in the tooth. Although it's been an extremely stable long term system @4.4GHz, frame time pacing issues are obvious on his 165Hz freesync monitor.