Rumor: AMD Ryzen 3000 "Matisse" Launching July 7

This is just so funny the look on his face when he announces that his confirmation of his source being right and he trusts them when he announced the October Navi delay, that look suddenly is on his face now when he announces that no no it might be July (7th even to make it valid)with some sob story that TSMC is flexible enough to produce Navi. from this point on you get to hear him explain it.


Never heard so much nonsense from a "tech" youtuber and the best thing he has to walk back his earlier trusted source :).

If you don't know what you are doing online you just report everything and that is just what Paul does without any hesitation that his sources use him to get the message across when they need him, so when Radeon VII launched Navi is delayed and when the bulk of the cards are gone, you guessed it you get this message ...

Who knew :)


What's even funnier is the people that believe him. If you've been around this industry for much time and you listen to Paul's "analysis" you can't help but laugh. And I don't only mean this video. It's clear Paul has a very limited understanding not only of technical matters, but of the history of the industry.
 
Many posts here are ~"AMD is giving my wallet a hard time" :), but OTOH, are'nt we lucky to be so tempted? Nobody is forcing us. :)

Its a pretty natty biz plan AMD have executed.

Not only have they sold cpus very well against Intel - they have switched die hard long term users to a stable amd platform.

This platform stability, has often resulted in progressive sales of 3 cpus to the same customer on an AM4 mobo e.g. - an easy drop in upgrade to keep pace with advances at reasonable cost & minimal trouble. (as prices dropped and with time to save up - a zen 1400 /apu > zen 1600 > zen + 2600 perhaps? - pretty affordable and desireable perf improvements)

Each such user who onsells their superseded processor, acts as a sales agent for another punter to acquire an amd platform to add to amdS user/customer base. They have also built an amateur sales force of evangelists.

Intel made customers ~start from scratch - a new everything at high cost & hassle, often for priced in marginal gains - resulting in ~ a few sales per decade.

As above, imo there are many am4 users who have upgraded processors twice in under two years of Ryzen.

In short - each sale won from intel is far more than a simple processor sale - it is a long term transfer of revenue streams to AMD. Intel has a very hard sell ahead to regain those consumer's computer budgets.

Theys say the money for car makers isnt in selling cars, but in selling spare parts for those cars. There is an element of that in amd's durable platform policy.
 
AM4 is not TR4 and that CPU you want will only fit into TR4 with a Hammer in your hand. No, simply NO !

Maybe they will name the 3000 version threadhammer and use the spare space on the am4 package for some hbm stacks.
 
Actually i have shown it to be an issue for ALL threadrippers if you read a few more posts in the 2990wx thread.

My understanding of the issue is threads staying active under windows, but not actually doing any computing because windows doesn't quite know what to do with CPU cores not having direct access to the memory. It's not a memory bandwidth issue in most cases, it's a chip design & software communication issue.

It's an issue that shouldn't be present in AM4 Zen 2, as both chiplets(16 cores) should have access to all the system memory through the I/O die.
 
Same here - my i7-2600k still gets the job done. I was originally planning on a May upgrade, but am definitely going to wait and see what AMD and Intel launches. If I were building a system today, I'd probably go with the 2700x. Microcenter has pretty good combo prices.
I'd kind of like to wait until Navi is launched so I can figure out which GPU to go with. I've decided my max budget for a GPU is going to be $400. I'm not overly excited about what that buys today. (I have a 970 right now and am happy with it for 1080 gaming - an upgrade should get me 1440 gaming or its not really an upgrade)
I have a 1080Ti, so I know I'm leaving a little performance on the table at 1440p, but the framerates on what I play are so high, it doesn't matter. At least I won't be worried about dropping coin on a new video card as well this summer.
 
Ugh... I really hope it isn't that late a release. Previous Zen and Zen+ releases, were Mainstream in March(ish) to allow Threadripper in Summer, if I recall? If mainstream Zen2 doesn't show up until July, does that mean that Threadripper will be delayed into Q3 or Q4? I certainly hope not.
 
Ugh... I really hope it isn't that late a release. Previous Zen and Zen+ releases, were Mainstream in March(ish) to allow Threadripper in Summer, if I recall? If mainstream Zen2 doesn't show up until July, does that mean that Threadripper will be delayed into Q3 or Q4? I certainly hope not.

It appears its a mad rush just to get the july paper launch. Originally it was a "spread" till october. Now they want to kick in the teeth of both intel and nvidia, by getting established before Q4. Or this release could over extend them and drivers will goto crap, but hey if that happens we will see 33% performance improvements over six months as drivers improve.
 
Ugh... I really hope it isn't that late a release. Previous Zen and Zen+ releases, were Mainstream in March(ish) to allow Threadripper in Summer, if I recall? If mainstream Zen2 doesn't show up until July, does that mean that Threadripper will be delayed into Q3 or Q4? I certainly hope not.

Im guessing so, if the rumours of TSMC's constrained 7nm production are true, getting 8 good chiplets on a TR die is going to take quite a few wafers away from mainstream, with only 1-2 chiplets. I see them pushing TR off while they feed Server and Mainstream until production ramps enough to fill all segments. Im looking forward to the Next TR as well, 32/64 3950X please!
 
AMD really needs to bring the clock speeds up on this release.
As a gamer, I do not have much use for more than 6 cores/12 threads.
If Zen 2 can overclock to at least 4.7Ghz, it will be an instant purchase/swap for my current 2600X.
 
Intel made customers ~start from scratch - a new everything at high cost & hassle, often for priced in marginal gains - resulting in ~ a few sales per decade.
To be fair in the past AMD wasn't hardly a competition so Intel could afford to do it, since it wasn't performance comparable.
Now that AMD has stepped up and frankly started knocking it outta the park it'll be interesting to see what happens with Comet Lake.
The current rumor is that it's staying socket 1151, haven't seen any details on backwards/forwards compatibility for mobo revision.
If they actually did, that would break a pretty long streak of tick-tock new socket release for Intel.
 
My understanding of the issue is threads staying active under windows, but not actually doing any computing because windows doesn't quite know what to do with CPU cores not having direct access to the memory. It's not a memory bandwidth issue in most cases, it's a chip design & software communication issue.

It's an issue that shouldn't be present in AM4 Zen 2, as both chiplets(16 cores) should have access to all the system memory through the I/O die.

AM4 should be fine, thats why we (my post and the quoted ones) were talking about threadripper.

EDIT

and to be fair, the threadripper issue has a subcategory to it that is memory induced which is why it causes the issue on not just the 2990wx and why the 2990wx is "officially" the only one with the issue.
 
AMD really needs to bring the clock speeds up on this release.
As a gamer, I do not have much use for more than 6 cores/12 threads.
If Zen 2 can overclock to at least 4.7Ghz, it will be an instant purchase/swap for my current 2600X.


Same for me. I have a 2600X and under custom water, PBO usually has me ~4.3GHz+
I really want to see them creep towards 5GHz. I bought the 2600X thinking that thermal load would be lighter and therefore yield higher clocks, but seems I was incorrect on that and I wish I went ahead and bought the 2700X. However, now I have an excuse to get a 3x00X :)
 
Friend snapped a photo of me reading this thread with a geeky smile and a semi-Mr Evil index finger in my mouth, got a lot of shit untill I explained I was waiting 2 years for this lol.
P.s. Austrian autobahn sucks bumpy 140kmh ass.
 
My gaming PC was built in 2009 with an i7 920 OC'd to 4.0ghz. It will turn 10 next week and I was planning on building a completely new system then. It's been pretty frustrating waiting for 10nm "next year" for the last 5 years. In the meantime the focus has been on core count, which is really disappointing for gaming (nearly every game is built around a single main thread which does the bulk of the work). I was about to pull the trigger on a 9900K just for the higher clock speed, but now I'm not so sure. Waiting a decade for ~30% better single core performance is just... underwhelming. I want to hear rumors about how fast these 7nm chips can actually run!
 
My gaming PC was built in 2009 with an i7 920 OC'd to 4.0ghz. It will turn 10 next week and I was planning on building a completely new system then. It's been pretty frustrating waiting for 10nm "next year" for the last 5 years. In the meantime the focus has been on core count, which is really disappointing for gaming (nearly every game is built around a single main thread which does the bulk of the work). I was about to pull the trigger on a 9900K just for the higher clock speed, but now I'm not so sure. Waiting a decade for ~30% better single core performance is just... underwhelming. I want to hear rumors about how fast these 7nm chips can actually run!

Totally agree with this and I was in the same boat. I used to upgrade every year then every other year. I finally upgraded from an i7920 3 years ago (I didn't last 10 years like you). But I don't think the CPU really increased much performance. I think the only real difference comes from the Video card (obviously) and the new SSD's. I mostly upgraded to get the new chipset to put in better SSD's.
 
In the meantime the focus has been on core count, which is really disappointing for gaming (nearly every game is built around a single main thread which does the bulk of the work).

Vulkan is supposed to fix that. Let's wait id will pull out of the hat with the upcoming Doom. They've moved to all Vulkan pipeline.
 
Totally agree with this and I was in the same boat. I used to upgrade every year then every other year. I finally upgraded from an i7920 3 years ago (I didn't last 10 years like you). But I don't think the CPU really increased much performance. I think the only real difference comes from the Video card (obviously) and the new SSD's. I mostly upgraded to get the new chipset to put in better SSD's.

Don't get me wrong, there's no way I could have survived that long without upgrading the video card. ;)

Did you get a Skylake CPU 3 years ago? Have you played any games since then where it's obviously holding you back?

Vulkan is supposed to fix that. Let's wait id will pull out of the hat with the upcoming Doom. They've moved to all Vulkan pipeline.

That sounds really good!

[programmer_rant]Unfortunately, after a brief read, it looks like it's all limited to rendering. Rendering is already easily threadable because the inputs are finalized at the point where the work starts. Core game logic usually doesn't have that luxury since the player can take action and modify the game state at any point in time. This causes the efforts of worker threads to immediately become stale and irrelevant when handled as asynchronous jobs. On the other hand, solving this problem by keeping threads in sync via mutex locks and semaphores often introduces enough blocking overhead that you lose performance in the long run. There really is no universal solution, and I don't see one on the horizon either. Faster clock speeds please![/programmer_rant]
 
That sounds really good!

[programmer_rant]Unfortunately, after a brief read, it looks like it's all limited to rendering. Rendering is already easily threadable because the inputs are finalized at the point where the work starts. Core game logic usually doesn't have that luxury since the player can take action and modify the game state at any point in time. This causes the efforts of worker threads to immediately become stale and irrelevant when handled as asynchronous jobs. On the other hand, solving this problem by keeping threads in sync via mutex locks and semaphores often introduces enough blocking overhead that you lose performance in the long run. There really is no universal solution, and I don't see one on the horizon either. Faster clock speeds please![/programmer_rant]

I am not a graphics guy (doing a strategy game in WebGL was the closest thing I got that sort of thing :() but I thought Vulkan at least treats each core as a first class citizen. That's my way of saying I agree with you.

However, there's plenty in today's games that does not or should not run on the main thread. Not every little thing is that critical that has to be spawned/modified/destroyed at the same time, simultaneously. At least not at the scale of today's games with huge maps, lots of players/entities/etc
 
I have CPU envy but I'm hoping a 7700k is good for one more generation. 8 cores at 5 ghz and PCIe 4 or 5 is a target for my next jump.
 
I have CPU envy but I'm hoping a 7700k is good for one more generation. 8 cores at 5 ghz and PCIe 4 or 5 is a target for my next jump.

I know what you mean. I don't really "need" an upgrade but my son's 2500k system is getting glitchy and I can't decide if I want to donate my 7700k system to him and spend $500 on a 2600/2700 system for me now or buy him a 1600 system from MC for $300ish and just hold off on my own upgrade until the 3000 comes out.
 
I'm sitting with a 2700x right now and this does excite me. Then I remember there is no reasonable upgrade for my RX 580 under 300 right now.

Fingers crossed Navi is released close to this time as well.

Yup. My GTX770 is mighty long in the tooth, and yet, there is not a compelling reason to upgrade for any of the games I actually play.
And no upgrade that isn't stupid expensive for what performance I'm getting.
 
My guess is that AMD Zen2 will offer minimum 6C desktop CPUs as the 8C chiplets will have great yields and will be a waste of silicon to get them down to just 4 active cores. And thus, AMD will force the CPU market to lower price per core again after the 1st gen Zen. So, the 6C Zen2 will suceed 4C Zen1 and 8C Zen2 will suceed the 6C Zen1 for the mass market. 12C Zen2 will take the place of 8C Zen1 and if they decide to do so, they will offer later a 16C for the real high end as a halo product to dethrone Intel for good on both top spot price and cpu power leadership. Apart from the 12C and 16C models, the smaller ones will have the normal 45W, 65W, 95W TDP to make things easier for OEMs and allow the ones who already have a Ryzen CPU to upgrade easily on the existing motherboards. My 5 cents.
 
My guess is that AMD Zen2 will offer minimum 6C desktop CPUs as the 8C chiplets will have great yields and will be a waste of silicon to get them down to just 4 active cores.

AMD would use the Zen+ APUs for 4C desktop processors.

There's no reason why AMD's budget processors need to be on AMD's latest and greatest architecture.

And thus, AMD will force the CPU market to lower price per core again after the 1st gen Zen. So, the 6C Zen2 will suceed 4C Zen1 and 8C Zen2 will suceed the 6C Zen1 for the mass market. 12C Zen2 will take the place of 8C Zen1 and if they decide to do so, they will offer later a 16C for the real high end as a halo product to dethrone Intel for good on both top spot price and cpu power leadership. Apart from the 12C and 16C models, the smaller ones will have the normal 45W, 65W, 95W TDP to make things easier for OEMs and allow the ones who already have a Ryzen CPU to upgrade easily on the existing motherboards. My 5 cents.

This logic fails from what I just said above.

I don't see any reason AMD would drop its prices unless Intel drop prices of its 6C/12T and 8C/16T processors ~$200 and ~$300, respectively.

12C/24T Ryzen 9 at $499 and 16C/32T Ryzen 9 at $699 be would AMD's new price points.
 
Last edited:
This logic fails from what I just said above.

I don't see any reason AMD would drop its prices unless Intel drop prices of its 6C/12T and 8C/16T processors ~$200 and ~$300, respectively.

12C/24T Ryzen 9 at $499 and 16C/32T Ryzen 9 at $699 be would AMD's new price points.

AMD has drastically undercut Intel for the past two CPU generations. There's really no reason to believe they won't here as well. Those price points you speculate also cut into where they price Threadripper. So you're speculating that they will not only go against their Ryzen strategy from the previous two launches but also dramatically raise the price on Zen 2 Threadripper chips. Threadripper prices will definitely go up, as will top end Ryzen prices, but your speculation is lacking any real supporting logic to hold it up. I would not be surprised to see $500-$600 for top end consumer Ryzen 3 and $700-$800 for entry level TR3, but more than that is a bit hard to buy at the moment.
 
AMD has drastically undercut Intel for the past two CPU generations. There's really no reason to believe they won't here as well.

Assuming that $329 8C/16 Ryzen 7 3700X performs close to the $525 Core i9-9900K, AMD is already "drastically undercutting Intel"

Those price points you speculate also cut into where they price Threadripper. So you're speculating that they will not only go against their Ryzen strategy from the previous two launches but also dramatically raise the price on Zen 2 Threadripper chips. Threadripper prices will definitely go up, as will top end Ryzen prices,

As I said before, Threadripper is a different market. It has quad-channel memory and more PCIe lanes.

The distinction is clear.

There's no reason why there wouldn't be an overlap.

but your speculation is lacking any real supporting logic to hold it up. I would not be surprised to see $500-$600 for top end consumer Ryzen 3 and $700-$800 for entry level TR3, but more than that is a bit hard to buy at the moment.

Your speculation lacks any support.

$329 8C/16 Ryzen 7 3700X that performs close to the $525 Core i9-9900K (at 63% of the price)

$499 12/24T Ryzen 9 3800X that has 50% more cores than the $525 Core i9-9900K

$699 16C/32T Ryzen 9 3900X just to one-up on Intel when it launches 10 core Comet Lake.

That's far cheaper than Intel prices.
 
It appears its a mad rush just to get the july paper launch. Originally it was a "spread" till october. Now they want to kick in the teeth of both intel and nvidia, by getting established before Q4. Or this release could over extend them and drivers will goto crap, but hey if that happens we will see 33% performance improvements over six months as drivers improve.

Definitely reads like they're trying to rush the 3xxxx series out in anticipation of the "college computer" season. Would be interesting to know if they've scored wins with the likes of Dell/HP/Lenovo to get further penetration. Which would also mean getting more on the mobile side, which I haven't read much about (with their GPU chips). So who knows?
 
Assuming that $329 8C/16 Ryzen 7 3700X performs close to the $525 Core i9-9900K, AMD is already "drastically undercutting Intel"



As I said before, Threadripper is a different market. It has quad-channel memory and more PCIe lanes.

The distinction is clear.

There's no reason why there wouldn't be an overlap.



Your speculation lacks any support.

$329 8C/16 Ryzen 7 3700X that performs close to the $525 Core i9-9900K (at 63% of the price)

$499 12/24T Ryzen 9 3800X that has 50% more cores than the $525 Core i9-9900K

$699 16C/32T Ryzen 9 3900X just to one-up on Intel when it launches 10 core Comet Lake.

That's far cheaper than Intel prices.

Of course there would be overlap. If entry level TR3 performed close to top-end Ryzen 3 and was the same or only a little more then people would likely go for Threadripper as it has more features and would be seen as more "future proof". I know for a fact that I would, even with the higher mb and ram costs associated with it. Top-end Ryzen and entry Threadripper at those price points would definitely be in the enthusiast market space. For businesses with the need for Threadripper they'd likely go for a higher end one and for general consumers you would never recommend a $700 CPU. For content creators and streamings Threadripper would be better at that point as well. There is a reason Intel's HEDT i9s and AMD's Threadripper CPUs are in entirely different price brackets compared to the consumer chips. Both companies want to actually sell the consumer-grade chips and not push people into thinking "well the HEDT part is so close in price I might as well pick that".
 
Of course there would be overlap. If entry level TR3 performed close to top-end Ryzen 3 and was the same or only a little more then people would likely go for Threadripper as it has more features and would be seen as more "future proof". I know for a fact that I would, even with the higher mb and ram costs associated with it. Top-end Ryzen and entry Threadripper at those price points would definitely be in the enthusiast market space. For businesses with the need for Threadripper they'd likely go for a higher end one and for general consumers you would never recommend a $700 CPU. For content creators and streamings Threadripper would be better at that point as well. There is a reason Intel's HEDT i9s and AMD's Threadripper CPUs are in entirely different price brackets compared to the consumer chips. Both companies want to actually sell the consumer-grade chips and not push people into thinking "well the HEDT part is so close in price I might as well pick that".

12C/24T Ryzen 9 and 16C/32T Ryzen 9 seems more like prosumuer products for people with too much money on their hands. (i.e. people who are willing to drop $700 a pop for a processor)

Actually professionals (i.e. movie editors) would want Ryzen Threadripper with quad-channel memory and more PCIe lanes in addition to having a lot of cores.
 
My gaming PC was built in 2009 with an i7 920 OC'd to 4.0ghz. It will turn 10 next week and I was planning on building a completely new system then. It's been pretty frustrating waiting for 10nm "next year" for the last 5 years. In the meantime the focus has been on core count, which is really disappointing for gaming (nearly every game is built around a single main thread which does the bulk of the work). I was about to pull the trigger on a 9900K just for the higher clock speed, but now I'm not so sure. Waiting a decade for ~30% better single core performance is just... underwhelming. I want to hear rumors about how fast these 7nm chips can actually run!

There is one game I play that scales linearly with core count: Kerbal Space Program.

My 4c8t is always at 100% CPU usage when at a high refresh rate. I am definitely ready for more cores.

The other thing is high speed DDR4. Fast low latency DDR4 like 3200 MHz CL14 will provide a massive perfomance boost in some games like ARMA.

I think upgrading to 10nm or 7nm will provide a great performance increase. I'm waiting as well. Still on DDR3...
 
Still on a 3930K... I'm not sure if I should be looking into these AMD's or not. Still runs everything pretty well relatively speaking.
 
My understanding of the issue is threads staying active under windows, but not actually doing any computing because windows doesn't quite know what to do with CPU cores not having direct access to the memory. It's not a memory bandwidth issue in most cases, it's a chip design & software communication issue.

It's an issue that shouldn't be present in AM4 Zen 2, as both chiplets(16 cores) should have access to all the system memory through the I/O die.

The issue is with CPU's that have more than 2 NUMA nodes per processor. So, you are correct that a 16 core AM4 chip wouldn't have an issue because it would have only one NUMA node, even with two chiplets.
 
Though my initial trade wasn't "the best" .. i went from 7700k to Ryzen 1700 ... sold the 1700 and bought a 2700x .. upgraded mobo from B350 to x470 .. bought a 2200g for the B350 (for wife) , bought a mini-ITX X370 and another 2200g (for my daughter)....

So yeah .. jumping from Intel, AMD has gotten alot of my moola
 
I would say that it has the single-thread performance of the Core i7-7700K/Core i7-8700K and multi-thread performance of the Core i9-9900K.

The single-thread performance should be the same at equivalent clockspeeds...

The 3700x is expected to be better than the 9900k, so yes.
(it all depends on the final released specs though)

This is the lingering question. AMD really needs to surpass the half-decade-old Intel Skylake architecture in current Intel CPUs. Not just because that would be necessary to make the CPUs more marketable, but also because Intel maybe/actually/finally has an upgrade for Skylake coming on their 10nm node which is comparable to the 7nm node(s) that the 3000-series Ryzens will be produced on. And well, AMD will really be judged on competing with that.
 
The single-thread performance should be the same at equivalent clockspeeds...

Remember that the single-thread performance is largely determined by the boost clock.


This is the lingering question. AMD really needs to surpass the half-decade-old Intel Skylake architecture in current Intel CPUs. Not just because that would be necessary to make the CPUs more marketable, but also because Intel maybe/actually/finally has an upgrade for Skylake coming on their 10nm node which is comparable to the 7nm node(s) that the 3000-series Ryzens will be produced on. And well, AMD will really be judged on competing with that.

Intel will first launch 10nm Ice Lake on mobile in time for end of 2019 (at least that's Intel's intention)

It is not know when Intel will launch its first 10nm desktop processor.
 
Back
Top