AMD Announces Ryzen 7000 Series - Launching This Fall

[Guest 5150]

1 hour ago, BiG StroOnZ said:

 

I definitely agree that AMD is going to max out the Silicon capability thermally. On the last page I posted info that claimed there was a Zen 4 part flying around that could do 5.85GHz. But I agree those days of getting 20-30% overclocks are long gone. I think for the general consensus having the parts maxed out from the factory makes sense, because the majority of the people out there are not going to overclock. However, it does take out much of the fun for the enthusiast niche consumer sector. Maxing out the capabilities of a part also looks good in benchmarks for reviews. 

 

Well, AMD said that despite the AM5 platform will be maxing out much higher than AM4, they still claimed there will be lower power parts. They said to Tom's Hardware that 65W and 105W parts will still exist alongside the 170W parts. Now the 220ge and 200ge are 35W parts, so it might be some time before we see parts like that releasing. But I imagine that the 65W parts will still be able to be cooled by some sort of stock cooler solution like a Wraith Prism. 

 

Yeah, the boosting algorithms have gotten quite complicated over the years, and there is definitely some level of marketing pizzazz. However, when I was still on my desktop platform (Delidded i7-3770k) I tested both all-core overclocks and per core overclocks. Surprisingly, there was some level of performance increase with the per core overclocking in certain applications. So, it is definitely specific to what type of workloads you are doing. There were a few news outlets (quoting a tweet from Ian Cutress) that were claiming that the 5.5GHz run that they were doing was with minimal effort on a 16c/32t ES from April, and it was bouncing between 5.2-5.5GHz all core in that benchmark. Thus, it might be safe to say because of this, that there is a possibility of 5.5GHz all core overclocks with proper cooling on certain Zen 4 chips. 

It will be interesting to see some end user results honestly. But 5.5ghz looks good, doesn't mean better than the competition on the IPC front, but at least we will have decent competition moving ahead from here.

 

Looks like instead of using A-XMP AMD will go with RAMP for DDR5 overclocking, which means that XMP rated sticks will have RAMP settings for AMD platforms. I think this means nothing other than a terminology change. For the better? Not sure. Still reading up on RAMP to get a better understanding so I can make accurate replies if questions are to pop up in the future. So 6400mhz may be a pretty common OC I think, 8000mhz would be top end. Still looking at DDR5 Dies and will be watching the next couple years for new revisions, see how we make out later on DDR5 platforms, we are still new yet.

[Blademaster91]

21 hours ago, tim0901 said:

No, it's not much different.

 

But you know what would perform much better?

It is different if you were to buy a B650 board then buy an X670 board later, its spending more money instead of just spending $50 more for a X670 board. Also DirectStorage could be utilized in games.

21 hours ago, tim0901 said:

Buying a new GPU that is compatible with PCIe 5.0 does not mean you need PCIe 5.0, the same way as how you don't need a PCIe 4.0 motherboard to enjoy a 3080.

IMO, if you can afford to spend $1000 on a 3080, you shouldn't be going cheap on a motherboard. I don't see the point of spending $500 on a motherboard, though on a high end setup with something like a 7800X and a RTX 3080, at least have some room to add a pci-e 5.0 SSD in later.

[BiG StroOnZ]

49 minutes ago, Guest 5150 said:

It will be interesting to see some end user results honestly. But 5.5ghz looks good, doesn't mean better than the competition on the IPC front, but at least we will have decent competition moving ahead from here.

 

Looks like instead of using A-XMP AMD will go with RAMP for DDR5 overclocking, which means that XMP rated sticks will have RAMP settings for AMD platforms. I think this means nothing other than a terminology change. For the better? Not sure. Still reading up on RAMP to get a better understanding so I can make accurate replies if questions are to pop up in the future. So 6400mhz may be a pretty common OC I think, 8000mhz would be top end. Still looking at DDR5 Dies and will be watching the next couple years for new revisions, see how we make out later on DDR5 platforms, we are still new yet.

 

I agree, can't wait to see what the community's results are. Yeah, I also believe Intel might have the edge this battle, but the war is certainly not over. The competition is heating up for sure. I posted a story the other day that is showing Raptor Lake could be between 21-28% faster in single-threaded performance if Zen 4 is only 15-19% faster than Zen 3. Hopefully Zen 4 is a little faster than that, bringing more competition to Raptor Lake. 

 

Supposedly the Ryzen Accelerated Memory Profile (RAMP) terminology was changed yet again to EXPO:

 

https://www.techpowerup.com/294169/amd-ramp-is-now-called-expo 

https://videocardz.com/newz/amd-preparing-expo-extended-profiles-for-overclocking-for-ryzen-7000-ddr5-memory-overclocking

 

Extended Profiles for Overclocking is what EXPO stands for, and this story came out towards the end of April 2022, while the RAMP stories tend to date around January 2022. We will have to see what they actually end up going with, as I think prior to Zen 4 they were going with DOCP or EOCP on some motherboards too (besides A-XMP).

 

I'd say your 8000MHz DDR5 top end prediction is pretty accurate (for common top end speeds being actually used in builds on average in the future). As when DDR4 launched its base speed was 2133MHz, and today 3600MHz DDR4 is basically standard (next to 3200MHz). But 3600MHz DDR4 is 69% faster than 2133MHz DDR4. If we take those speeds and extrapolate, we start with DDR5 4800MHz (its base speed) and 8000MHz DDR5 is 67% faster than 4800MHz. So it's safe to say that 8000MHz kits will exist and be the norm at one point. There will definitely be speeds higher than that; as we see DDR4 kits higher than 3600MHz. However, we will have to see if there are diminishing returns beyond 8000MHz DDR5 like we see beyond 3600MHz with DDR4. DDR5 is quite new, so there's plenty of room for revision as you say.

[Guest 5150]

5 minutes ago, BiG StroOnZ said:

 

I agree, can't wait to see what the community's results are. Yeah, I also believe Intel might have the edge this battle, but the war is certainly not over. The competition is heating up for sure. I posted a story the other day that is showing Raptor Lake could be between 21-28% faster in single-threaded performance if Zen 4 is only 15-19% faster than Zen 3. Hopefully Zen 4 is a little faster than that, bringing more competition to Raptor Lake. 

 

Supposedly the Ryzen Accelerated Memory Profile (RAMP) terminology was changed yet again to EXPO:

 

https://www.techpowerup.com/294169/amd-ramp-is-now-called-expo 

https://videocardz.com/newz/amd-preparing-expo-extended-profiles-for-overclocking-for-ryzen-7000-ddr5-memory-overclocking

 

Extended Profiles for Overclocking is what EXPO stands for, and this story came out towards the end of April 2022, while the RAMP stories tend to date around January 2022. We will have to see what they actually end up going with, as I think prior to Zen 4 they were going with DOCP or EOCP on some motherboards too (besides A-XMP).

 

I'd say your 8000MHz DDR5 top end prediction is pretty accurate (for common top end speeds being actually used in builds on average in the future). As when DDR4 launched its base speed was 2133MHz, and today 3600MHz DDR4 is basically standard (next to 3200MHz). But 3600MHz DDR4 is 69% faster than 2133MHz DDR4. If we take those speeds and extrapolate, we start with DDR5 4800MHz (its base speed) and 8000MHz DDR5 is 67% faster than 4800MHz. So it's safe to say that 8000MHz kits will exist and be the norm at one point. There will definitely be speeds higher than that; as we see DDR4 kits higher than 3600MHz. However, we will have to see if there are diminishing returns beyond 8000MHz DDR5 like we see beyond 3600MHz with DDR4. DDR5 is quite new, so there's plenty of room for revision as you say.

Yah, go figure. They can't use XMP though. I think Intel owns the rights to that one lol.

 

EXPO. Ehh, RAMP looked better to me, from a marketing consumer angle. 

 

"Ramp up memory tweaks and dropin' high clocks." 8000mhz CL36.

 

We shall see the latency though. I don't anticipate low/lower than what has been accomplished already, probably worse than Intels though. I'm hoping they uncouple future IF links and give us dedicated IF overclocking. Let the RAM run free to OC without penalty!!

[tim0901]

3 hours ago, Blademaster91 said:

It is different if you were to buy a B650 board then buy an X670 board later, its spending more money instead of just spending $50 more for a X670 board. Also DirectStorage could be utilized in games.

The difference in price between a B650 and X570 is predicted to be over $100, not $50. If they were only $50 apart then sure, I could maybe justify it a bit more especially given everything else you'll get. But they aren't. Also, you can always just sell the B650 board if and when you buy its replacement to subsidise the cost.

 

RE: DirectStorage - we probably won't see any advantage going above PCIe 4.0 speeds as games are primarily designed around console hardware. A PCIe 4.0 SSD is slightly slower than the PS5's SSD while being significantly faster than the Xbox Series X/S SSD. As such, higher SSD speeds will likely go unutilised in games for the same reason that you will see underutilised CPU cores after going above 8 cores. Maybe with the following generation of consoles, but those are probably 5+ years away.

 

But lets suppose you have a game that can leverage the full capability of PCIe 5.0 SSDs - given that an x4 5.0 connection is equal in bandwidth to an x8 4.0 connection, you should still get the full benefit out of that SSD in a DirectStorage game while your GPU is running in a 4.0 x16 slot. It would take up - when running flat out - half the bandwidth, but at most that would mean we see the same 2-3% performance deficit that we see when dropping current-gen GPUs to use PCIe 3.0.

 

But DirectStorage isn't going to be hammering the PCIe bus constantly. It's going to be a very bursty workload - lots of traffic when first loading into a level, but then falling to basically nothing once you've loaded in. This is also true to an extent with the PCIe usage of GPUs - the amount of data they are sending fluctuates greatly over time. As such, you'd likely only encounter issues when both devices are hammering the PCIe bus, and how often will that be?

 

And on top of all that - DirectStorage games are still ages away. DirectStorage as it is today doesn't support GPU decompression, meaning there's very little benefit to using it as you have to send the data to the CPU anyway for decompression. This was very apparent in benchmarks we saw in March, where enabling DirectStorage in Forspoken only changed loading times by a fraction of a second... For an M.2 SSD it went down from 2.4 seconds to 1.9 seconds - that's basically unnoticable unless you've got the two side-by-side. Now of course that's an early-access title and a single data point, but it's still something to be aware of. Just having an M.2 SSD made more of a difference than DirectStorage did. It's going to be a while before GPU decompression arrives and then another while before we actually start to see games that implement the technology, so I really wouldn't worry about it at this stage.

 

3 hours ago, Blademaster91 said:

IMO, if you can afford to spend $1000 on a 3080, you shouldn't be going cheap on a motherboard. I don't see the point of spending $500 on a motherboard, though on a high end setup with something like a 7800X and a RTX 3080, at least have some room to add a pci-e 5.0 SSD in later.

Personally I disagree - I don't see the point in spending any more on a motherboard than necessary. It doesn't contribute to the performance of the system at all and so unless I'm looking for a particular feature (eg WiFi) I don't see any reason to spend more on it than I need to in order to extract the full performance from my CPU (without overclocking). There are plenty of creators out there who benchmark motherboards to test which ones are capable handling which CPUs - if a $180 B650 board can do this, why should I go higher? What tangible benefit do I get? Oh no, the VRM is 10C hotter than a motherboard that costs $50 more... big deal. Just buy a 40mm fan and point it at the VRM if you really care that much - it'll be better and cheaper.

 

For example, why should I buy this £120/$150 B550 motherboard over this £200/$250 one? The benefits of the expensive one that I can see are:

 

• 2.5Gb LAN - to some this is nice to have, but there are cheaper motherboards that offer it, or you could just buy a £26 add-in card
• WiFi/bluetooth - again, this will matter to some, but there are cheaper ways of getting it
• Two more M.2 slots (4 total) - someone might care about this, but most people - myself included - don't need more than two M.2 slots
• The 3.1 Gen 1 port is now a 3.1 Gen 2 port
• Two less USB 2.0 ports (I guess to allow for the above Gen 2 port? Maybe?)
• On-board power and reset switches - if you overclock a lot I guess that's nice, but I don't care for it
• Better (by some unquantified metric) audio
• Better (by some unquantified metric) VRM - again, nice if you plan on overclocking

To me at least, that's not worth £80/$100 at all. Why should I bother spending money on features I don't care about?

 

And FYI, all 600-series boards (with the exception of A620) have a PCIe 5.0 x4 (m.2) slot for storage, so you won't need to upgrade from B650 to add a new 5.0 SSD.

[ZetZet]

8 hours ago, tim0901 said:

Personally I disagree - I don't see the point in spending any more on a motherboard than necessary. It doesn't contribute to the performance of the system at all and so unless I'm looking for a particular feature (eg WiFi) I don't see any reason to spend more on it than I need to in order to extract the full performance from my CPU (without overclocking). There are plenty of creators out there who benchmark motherboards to test which ones are capable handling which CPUs - if a $180 B650 board can do this, why should I go higher? What tangible benefit do I get? Oh no, the VRM is 10C hotter than a motherboard that costs $50 more... big deal. Just buy a 40mm fan and point it at the VRM if you really care that much - it'll be better and cheaper.

I agree with you completely. 3070 to a 3080, that's a performance jump. So it's worth spending money. B650 to X670? Nothing happens unless you completely MAX OUT the I/O on B650 and that's not possible for most people. 

Gen5 PCIe is useless for GPUs too, unless AMD decides that their DirectStorage implementation requires it to work (or you try to "future proof", but then gen3 came out a decade ago and that's still not maxed out).