Official AMD Ryzen 5000 Series & 7000 Series Thread

[pax]

X600 gen mobos leaking before tomorrows computex. 2 chipset io and tons of usb on some models is what comes off at first:

https://www.techpowerup.com/295068/a...ad-of-computex

https://www.techpowerup.com/295095/f...appears-online

These pictures are often used to make sure things like the solder mask and components were applied properly during production. However, it does give us a good look at the overall layout of the ASUS X670 Prime-P WiFi and the fact that the X670 chipset does indeed consist of two parts. What is also clear is that we're looking at a PCIe 5.0 x16 slot here, as these slots are SMD components rather than through hole.

Other things that are clearly visible, includes support for three M.2 slots, of which the one closest to the CPU might be PCIe 5.0, but there's really no way of telling by just looking at the connector placement. There's also a space for a WiFi module at the bottom of the rear I/O, but beyond that, it's hard to make out the proper port layout. However, there appears to be at least one USB-C port at the rear, as well as a header for another one at the front of the motherboard, next to a USB 3.0 header. The board also appears to feature 14 power phases and obviously four DDR5 DIMM slots. The chip between the two chipsets are either a Super I/O chip or possibly a PCIe redriver. In addition to the x16 PCIe 5.0 slot, the X670 Prime-P appears to be getting a single PCIe x1 slot and two PCIe x4 slots, both which appear to get physical x16 slots.

https://wccftech.com/msi-x670e-godli...zen-7000-cpus/

According to the leak, the pictures give us a first look at the design of the MSI MPG X670E Carbon WIFI while also detailing the I/O of the flagship MEG X670E GODLIKE and X670E ACE offerings. The lineup also features a singular PRO offering, the X670-P WIFI which will come with both PCIe Gen 5.0 and 4.0 support while the X670E offerings will feature all PCIe Gen 5.0 support. The board does look like it comes with a massive VRM heatsink with a new logo design & the picture also confirms a Lightning Gen 5 M.2 port on the Carbon along with the PCIe Gen 5.0 x16 slot.

2 versions of the x670- chipset. Something look out for.

[Gandalfthewhite]

yeah sounds like x670e will support pcie gen5 GPUs and expanded IO (looks like dual chipsets to handle it all) and x670 will be pcie gen4 gpu support

[Nagorak]

It's weird to me how PCI-e generations have gotten so far out ahead of what GPUs actually utilize.

[Gandalfthewhite]

that is because pcie isnt just used by GPUs anymore. its become a generalized I/O interface supporting storage , usb , expansion slots, interconnects to chipsets.

[Nunz]

that is because pcie isnt just used by GPUs anymore. its become a generalized I/O interface supporting storage , usb , expansion slots, interconnects to chipsets.

Took the words right out of my mouth.

[Gandalfthewhite]

I expect tonight's presentation to be mostly cover platform , CPUs , APUs, and a 1-2 minute preview of RDNA3 (nothing substantial maybe a tighter launch window)

[Megaman]

5.5Ghz hey........

31% faster in Blender over the 12900K

[acroig]

Link because Mega can't be bothered:

https://www.pcgamer.com/amd-zen-4-ry...-announcement/

[Megaman]

Link because Mega can't be bothered:

Word.

[acroig]

Word.

SMH.....

[Megaman]

I am the link….bish

[acroig]

I am the link….bish

[the_sextein]

MLID and many others predicted a 10-11% clock increase because of the drop from 7nm to 5nm. Turns out they were pretty close as it looks like they gained a 12% clock increase from this. This puts their clock speeds on par with Intel's fastest single thread chip at 5.5Ghz. Intel will no doubt respond with a chip that beats it by 200 or 300Mhz but still, it's nice to see AMD gain ground in this important area. The thing I find most exciting is the gameplay of Deathloop playing in the background is clearly showing this CPU hitting 5.5Ghz while playing a game that uses 8 or more cores. This could be a sign that AMD has achieved 5.5Ghz clock speeds on 8 cores or more at the same time.

The thing MLID got wrong and that I find especially disappointing is the IPC gains. It looks like they gained12% from clocks and AMD is saying 15% improvement for single thread so they only gained 3% IPC gain from the looks of it. Perhaps someone told him up to 30% gains in blender and he figured that they were going to gain 10% from clock rates so he just guessed that they had a 20% IPC gain. That's not how it works lol.

I'd be interested to see what temps these chips operate on with a large AIO.

[acroig]

Wow, AMD has a 5.5GHz all core CPU. Good times ahead.

[the_sextein]

Maybe, but 8 cores is only half the cores on the 7950X and Deathloop is not a CPU intensive game so the situation could change based on CPU load. Intel will make up some ground by catching up with AMD's expanded cache and it will probably regain the single thread crown with increased clock frequencies but I doubt Intel will be able to maintain 5.5Ghz on all 8 big cores so AMD could still pull ahead in gaming. I'd be surprised to see Intel Pass 5.3Ghz all core to be honest. Still, AMD didn't gain much from architectural improvements. If Intel makes even a 10% gain here (which they are rumored to) then they will pull ahead in most everything I use a CPU for, if I had to guess it. If the difference is + or - 3% after overclocking then I will buy whatever chip keeps the lowest temps.

[Nunz]

Falling prey to marketing schemes I see..

5.5Ghz all-core? No shot

[Mangler]

Maybe I'm a Debbie Downer, but the x670E feels like it is going to have many weird issues at launch with its dual chipset.

[Riptide]

Maybe I'm a Debbie Downer, but the x670E feels like it is going to have many weird issues at launch with its dual chipset.

I'm more concerned about cost. We keep getting further away from the $250 price point for a motherboard with these flagship high end models. What are they going to ask for on these? The same amount we pay for the CPU?

Hard to get excited any more about motherboards when the cost is getting so far out of hand.

[the_sextein]

Maybe I'm a Debbie Downer, but the x670E feels like it is going to have many weird issues at launch with its dual chipset.

It's possible but I'm sure Gamers Nexus and many others will report any problems when they review the platform and thoroughly test it.

Looks like wccf is already talking about the 15% improvement vs leaks haha.
https://wccftech.com/amd-greater-tha...nother-jebait/

[Nunz]

I'm more concerned about cost. We keep getting further away from the $250 price point for a motherboard with these flagship high end models. What are they going to ask for on these? The same amount we pay for the CPU?

Hard to get excited any more about motherboards when the cost is getting so far out of hand.

$250 for a high-end board? Maybe in 2004. Those prices haven't existed in quite a long time.

Everyone wants new features like PCIE5.0, supreme expandability, better on-board audio, high quality VRMs, 8-10 layer PCBs .. but then no one wants to pay for it.

[pax]

I think the game shown was Ghostwire Tokyo. They didnt show any fps tho but its probably cpu bound anyway.

5.5 has to be single core imo. But they did report in some places of 5.0 or more all core which is still a good boost. No info on any 3d cache. They are probably so focused on getting that 5+ ghz that we wont see it for a while. Maybe a spring refresh on 2023 or maybe even later on 3nm.

15% with 5.5 is still modest... I think this platform will hit its stride in its 2nd gen. First gen being mostly a solid productivity boost.

Some games like core speed more than cache or core count tho so be cool to see how that will pan out game wise. I wouldnt be surprised to see the 5800x3d beat zen4 in some games.

[Nunz]

I wouldn't be surprised either about the 5800X3D.

There's no chance we'll see 3D Cache on a Zen4 for some time, imo. I'm suspecting the push in core clocks is going to up power consumption, and with more power, more heat.. and heat is the main reason the 5800X3D had to have it's core clock gimped.

With the success of the chip, though, I'm sure AMD went back to the lab to figure out how to

A) make the cache more resistant to thermals+current
B) potentially reposition the cache to increase thermal transfer ability

[Nagorak]

Zen 4 sounds a bit disappointing to me based on what has been released. Only a 15-20% uplift is only going to put it in the same ballpark as the 5800X3D for gaming. Granted, outside of gaming it would be faster, but still kind of a limited improvement to justify the cost and hassle of switching over to a whole new platform and memory type.

[AllexxisF1]

I have a feeling AMD is sandbagging.

[the_sextein]

I think the game shown was Ghostwire Tokyo. They didnt show any fps tho but its probably cpu bound anyway.

5.5 has to be single core imo. But they did report in some places of 5.0 or more all core which is still a good boost. No info on any 3d cache. They are probably so focused on getting that 5+ ghz that we wont see it for a while. Maybe a spring refresh on 2023 or maybe even later on 3nm.

15% with 5.5 is still modest... I think this platform will hit its stride in its 2nd gen. First gen being mostly a solid productivity boost.

Some games like core speed more than cache or core count tho so be cool to see how that will pan out game wise. I wouldnt be surprised to see the 5800x3d beat zen4 in some games.

Yeah you are right, it was tokyo ghostwire, my mistake. I think it's possible that AMD is designating 8 cores as performance cores and 8 cores as support cores. If an app uses all 16 cores equally then it runs them all at 5Ghz but if a game is using 8 threads primarily then it probably cuts the frequency on 8 of the cores by 1.5Ghz or something and boosts the primary cores to 5.5Ghz. This would explain why the game starts at 5Ghz and quickly boosts to 5.5Ghz. They probably don't want Intel to know exactly how their CPU works which is why they only showed a single core speed instead of each core. Each core is a large core but their software is probably tweaking how the cores are used to compete with Intels 8 core config for gaming purposes. Just a thought and pure specualtion on my part.

[the_sextein]

If All PC games have to work and be optimized for 8 cores so that they work as good as possible for Intel then AMD just needs to tell the OS what 8 cores to direct the data to and they can practically turn off half of their cores and run the chip like a big 8 core CPU when it games. This would allow them to achieve higher 8 core clocks than Intel. Intel would have a higher single core clock which would win at cinebench but it wouldn't allow them to win at gaming. Thats what I think is happening. I hope anyway. I'd like to see AMD be more competitive right now.

[Nagorak]

I'm more concerned about cost. We keep getting further away from the $250 price point for a motherboard with these flagship high end models. What are they going to ask for on these? The same amount we pay for the CPU?

Hard to get excited any more about motherboards when the cost is getting so far out of hand.

I think this is mostly because the motherboard manufacturers realized there was a market for "Cadillac" boards that are well into the range of diminishing returns in terms of quality/features per buck. As far as I can tell the quality of $150-$200 boards is still the same or better than in the past, and I'd say that boards in the $100 range boards are clearly higher quality than they were a decade ago. The bottom line in my opinion is if your limit is $250 for a board, then just buy one in that price range.

[the_sextein]

Flipping burgers paid $7.50 an hour back then, now it pays $15.00. Gas prices are high and effecting the cost of everything else, Tariff's, Covid ect, it sucks pretty bad. They alienate enough people and high tech business starts to suffer because people can't afford their products anymore. If I got paid twice what I did when motherboards were $250 then I wouldn't think they were expensive or care but only minimum wage has doubled. The skilled labor just barely keeps up these days.

[Gandalfthewhite]

ive got to agree with wccftech amd didnt say 15% they said over 15% we dont know what that means yet. Hopefully we will get more information by august (if expected launch is somewhere in the late september to mid december time frame. I'm guessing late october early novemeber launch at this point.

[the_sextein]

Well, it's 15% OR greater and 12% is clock increase so IPC probably varies from 3% to some unstated peak but I get the feeling that is because the majority of their IPC improvements have to do with multi threading so they can keep up with another volley of E cores from Intel. 31% multithreading improvement with the same number of cores and a 10% clock increase is pretty solid gains for a generational jump. Better than either AMD or Intel had done for the decade leading up to the 5000 series. People were hoping for a 30% single core bump and it ain't going to happen.

[pax]

Well we saw 30% productivity so if its average over 15% its hard to say how that affects gaming. But ya they could be sandbagging and trying to get intel to segment its next gen differently.

[Nunz]

They showed the Zen3 with mediocre single-rank RAM vs the Ryzen 7000 with 6000CL30 which is pretty solid for DDR5. They did everything they could to make the gap look better.

Ryzen gains quite well with Dual-Rank especially with the lack of tREFI tweaking on Ryzen.

[the_sextein]

All I know is that game was using more than 4 cores ( It recommends 6) and the sustained boost was 5.5Ghz. I seriously doubt that AMD can push more than 5Ghz on all 16 cores. In fact, I'm kind of surprised they managed 5Ghz all core at all but a single core doesn't just run at max boost while the other cores run at 5Ghz. There is clearly something going on in this demonstration and multiple cores are running this game at 5.5Ghz. I'm pretty sure AMD did it like this intentionally. My guess is 6 to 8 cores with max boost.

[Nagorak]

Blender and Cinebench aren't work loads that really benefit from faster memory, so I don't think the memory they paired with each processor should have a substantial impact on how they stack up to one another.

All I know is that game was using more than 4 cores ( It recommends 6) and the sustained boost was 5.5Ghz. I seriously doubt that AMD can push more than 5Ghz on all 16 cores. In fact, I'm kind of surprised they managed 5Ghz all core at all but a single core doesn't just run at max boost while the other cores run at 5Ghz. There is clearly something going on in this demonstration and multiple cores are running this game at 5.5Ghz. I'm pretty sure AMD did it like this intentionally. My guess is 6 to 8 cores with max boost.

If they can do 5.5 GHz single core I don't think 5 GHz all core is out of the question. Maybe not in a heavy loads, but in gaming I see it being possible. Current Ryzen 5000 CPUs can boost to 5 GHz single core, and can easily sustain 4.6-4.7 GHz in games. A 10% drop off for max single core to multi core seems pretty standard.

Actually hitting 5.5 GHz in game is also better than what the current 5000 series line-up can do with its max boost. A 5950X can hit "5.1 GHz", but the reality is you won't see it in all but the lightest of loads (like a background workload when otherwise idle).

[the_sextein]

I think you may have misunderstood what I was trying to say. They are getting 5Ghz all core that is not in question, nor is a 5.5Ghz core clock in game. The question is how is a single core running at 5.5Ghz in a game that is clearly using 6 to 8 cores unless they have implemented a new software instruction that boosts cores based on how many threads are seeing a certain amount of load? They clearly have independent core clocks going on during this demonstration. Unless you believe they pushed all 16 cores to 5.5Ghz. This game is clearly not just using single core but we are seeing single core clock speeds of5.5Ghz on at least 1 core and it could be more, my guess is 6 to 8.

[pax]

Yea they might be sandbagging a little lol.

https://twitter.com/greymon55/status...17091682996224

Tweet
See new Tweets
Conversation
Greymon55
@greymon55
Can someone explain why the footnote is 46% faster but the slide is 31% faster?🤔

[Munkus]

Yea they might be sandbagging a little lol.

https://twitter.com/greymon55/status...17091682996224

If you read the tweets, it explains the numbers.

[pax]

If you read the tweets, it explains the numbers.

I know they chose the lower number from amd's perspective. But it did take intel 46% more time to do the task.

[KAC]

As with any ppt slides don’t trust this crap. Real world will be different.
Seems quite underwhelming overall. Will probably need to wait a couple of gens before AMD gets it shite together.

I give this reveal a -5/10. Think Intel might just come out ahead as they are already on the new platform and had one successful round.

[pax]

I think they will have a good part. Intel would have to put out a 46% faster part in MC in their next gen vs ADL just to catch up in this one test. A pretty big hill to climb in one gen. But itll be a mixed bag. Gaming probably mostly intel and productivity amd... fact that AMD didnt produce gaming benches probably means they arent as fast gaming wise this round.

Id bet on a gaming cpu again tho with that added cache.