Official AMD Ryzen 5000 Series & 7000 Series Thread

[Megaman]

Seems quite underwhelming overall.

Cue 3D cache.

You heard it first here, by me, on Rage3d

[KAC]

Yes when it comes we will see. I wanted to upgrade at end of the year. Guess this 5900X has still some more to give.

[Nagorak]

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.

Yeah, I misunderstood what you were saying. Good point.

[pax]

Coreteks thinks the sandbagging was more on gaming tho CB20 had a 13% uplift from zen2 to zen3 vs the 31% from zen3 to zen4 so that was impressive as well.

[yt]kTuV7l42ncs[/yt]

Looking forward to this now also since he says zen 5 is only 2 years later in 2024... I thought it might show up only a year later.

[the_sextein]

AMD's power point slide says 5Ghz+ boost. So it will be able to do 5Ghz at the minimum which would be in 16 core mode. It doesn't specify how much more than 5ghz it will go. That means all 16 cores will hit 5Ghz and from there the boost can go higher. A game that uses 6 to 8 cores was running at 5.5Ghz. What do you think the single core boost is going to be? If AMD hit 6Ghz first I'd sh!t myself. If Intel gets stuck at 5.2Ghz for 8 core mode and AMD can push 5.5Ghz that will give AMD at 300Mhz X 8 boost. Or to put it in other words, a 2400 clock cycle beat down.

We will see 15% or greater single thread performance which probably depends heavily on how high it's boosting. So 15% per thread in all core mode but it could be much higher in single core mode. I think they are just pissing with people when it comes to their wording in this presentation. They don'ts state single core regarding boosts and they specifically call out single THREAD when talking about 15%+ increases. They also mention AI instructions but give no details. They are teasing everyone because they can't let the cat out of the bag right now or Intel would have time to optimize against their chip specifics.

[Nunz]

Damn you're chomping up marketing BS like it's a buffet

Wouldn't be the first time AMD lied on a marketing slide, won't be the last.

[the_sextein]

Nah, if anything I think their marketing it holding back on purpose. It's basically not making any "greater than" claims and is giving vague minimal constraints so people get and idea of worst case scenario. No handpicked game numbers like last time, no talk of specific IPC improvements. I'm just giving AMD the benefit of the doubt. I don't think it's any secret that 5nm vs 10nm is going to give some power and heat benefits. The IPC war is close and it looks like clocks are as well. Intel uses smaller cores to get around their power and heat problems but that locks them into 8 cores in gaming mode where their all core clocks are completely hand tied. With only half of AMD's cores needing to operate at a high level when gaming.... I think we will see AMD's advantage resurface with this generation. It just required a little more sophisticated clock assignment to work around the issue.

[pax]

Wondering if the doubling of L2 cache in Zen 4 will be as good as the increase in L3 in the 5800x3d... then add the core clocks to 5.5 ghz those games that prefer core speed over cache could mean AMD has a lock on gaming.

[the_sextein]

https://wccftech.com/amd-ryzen-7000-...a-2-igpu-more/

Some new info regarding clocks, overclocks and L2 cache. So he is claiming 5.2Ghz to 5.5Ghz on all 32 threads during that game demo without OC. I watched it and it boosted to 5.5 Ghz within seconds and pretty much stayed within 100Mhz the entire time. Of course there will be flux on some of the threads but that's pretty impressive if true. If an app only uses a single thread, chances are it will boost even higher and then we get overclocking on 360mm AIO's. I wouldn't count 6Ghz single core out of the equation just yet.

[pax]

Coreteks said core dependency wasnt high on that game. He measured a max 19% so having high clocks may be easy in this case.

So clocks could vary a lot game to game depending how they use their cpu cores.

[the_sextein]

For sure, they picked a modern title that goes lite on the CPU to make it look impressive but in my opinion it is impressive. I wouldn't be shocked if you could OC one of these to a stable 5.5Ghz in most games based on this info and that was an old sample so the new chips at launch will be even better.

[the_sextein]

For me, moving up from an I9 9900K. This would mean going from 8 cores at 5Ghz to 16 cores at 5.5Ghz. If games use more than 8 cores over the next 5 years it's got me covered and gives a huge advantage over Intel. Even if it doesn't, Intel will struggle to hit similar clock frequencies on half the cores unless they really do something to compete against 5nm. It's got a 50% IPC increase over my 4 year old chip and 4X the L2 and L3 cache while using the same amount of power. Add in PCIE 5.0 on everything and DDR5, this would do nicely as an upgrade this fall.

I'm not counting Intel out of course but this looks really promising to my eyes and I'm optimistic about the launch. It's not like the 3000 series where they had the same IPC and 700Mhz slower clocks a year late. This is coming out early with similar IPC and higher clock speeds on double the cores while obliterating the 12900K's 380watt,100C power and temps. AMD has come a long way over the last 5 years and it's really made me respect them.

[pax]

Itll be fun to see how it benches at release. Im sure with good cooling it could do allcore 5 ghz at the very least. And of course the worst case benching tests from reviewers using such as prime95 are often far above many typical gaming scenarios. Im seeing that with the 5800x3d. Its rarely going above 67c with my uber silent fan curve which only ramps at 70-80c. GB would go above that at 83c.

[the_sextein]

Yeah I'm really looking forward to seeing the reviews. My I9 9900K is similar regarding temps at 5Ghz all core. It plays games at about 65C. Cinebench is about 83C and Prime 95 is about 91C. I really don't care about prime 95 but I would like to see all core loads in cinebench drop about 5C from where they are. I doubt that will happened on a chip as beastly as the 7950X but the fact that it's even capable of keeping similar temps and power with double and quadruple everything is pretty neat in my book. Hopefully they will have designed a interface that is slightly more efficient at cooling the chip this time. You never know.

[Gandalfthewhite]

yeah given that it sounds like raptor lake is just adding E cores.... zen5 and meteor lake look like when things are going to get extra spicy

[acroig]

yeah given that it sounds like raptor lake is just adding E cores.... zen5 and meteor lake look like when things are going to get extra spicy

Next two CPU releases promise to be a great fight. Everyone wins..... provided there's plenty of stock.

[Nunz]

yeah given that it sounds like raptor lake is just adding E cores....

That's not the case at all.

[the_sextein]

As impressed as I am with what AMD has shared so far I think the 13900K poses a pretty serious threat. Intel isn't dropping down to 5nm but they are improving the 10nm wafers they are working with which should improve power draw and allow a clock boost. They are also increasing their cache and I would assume improvements are probably being made to the IPC of both the large and small E cores. So adding 8 more is going to be pretty harsh competition regarding multi threading. Still, AMD looks like they have made a fine chip that might come out on top. The clock increases that AMD has been able to achieve with 5nm are beyond what I expected.

[Nunz]

We don't know what the actual clocks will be on the AMD chip. Again.. nothing definitive from marketing slides.

[the_sextein]

The slides say 5Ghz+ all core which would be under full load worst case scenario. I don't think you will get a definitive number outside of the that because it will vary based on the chip and cooler and application core usage from there. I would assume 5.5Ghz on all 32 threads was a best case scenario for a modern game in order to make it look as good a possible for the presentation. I wouldn't expect better than that outside of single core mode which no modern apps are even single threaded anymore so it's little more than a pissing match these days. With a good AIO and some overclocking you might be able to achieve that 5.5Ghz clock in more demanding titles but common sense tell me all core under full load will probably stick within 200Mhz of the base clocks.

[Nunz]

I think above 5Ghz all-core on 16cores is a MASSIVE stretch unless the chip will be allowed to suck down 250-300watts, and with the thermal issues 5000 series chips see regarding lack of dissipation due to high-density, I really don't see that being possible. Cooling isn't the issue, a single 360rad can handle 300watts, but pulling the heat off the die is.

I think 5Ghz all-core on 16c/32t will be achievable but under high power consumption. At stock limits, I don't think many will see 5Ghz unless under light load/low current. 5.5Ghz at 1-2 cores is possible, and we already have that on Intel 7 12900KS. That's not really a revelation especially with how hard AMD bins the top-dog chip (5950X cores are binned quite well)

5.5Ghz on 16c/32t? I'd say you're approaching "are you doing crack" territory.

[the_sextein]

Well, I guess we will see but https://wccftech.com/amd-corrects-it...or-am5-socket/

I doubt that AMD was straight up lying about the clocks during the interview with the nerds. I believe them when they say that it was running at 5.5Ghz all core. I saw it on screen with my own eyes but it is a light load on 6 cores which is not the same as a full load on 16 cores obviously. My I9 9900K sucks 220 watts under full load so it doesn't surprise me that they are pushing 250 this time, It's still a very efficient design. Still, pushing up the ranks isn't as hard when the silicon isn't fighting you. Intel has hit a brick wall and I believe 5nm is going to give a larger advantage allowing all core clocks to rise. It would be smoking crack levels if Intel were to push 5.5Ghz all core on even 8 cores let alone 16 but lets not forget how far ahead AMD is when it comes to design and silicon. Intel has been half assing it for a long time and their small core design was a quick bolted on idea after they got rid of trust fund stock market timmy. They are going to need a top to bottom modern design on cutting edge silicon designed by people that were actually hired based on skill level if they want to stay up with AMD at this point. I think the 13900K is going to be very close but Intel will need to scrap it and start over after this if they hope to stay alive. Personally based on what I am seeing, I think intel will be defeated this round when it comes to gaming and will probably trade blows or lose by a small amount in multithreading. Intel's big cores are completely locked up as far as I can tell and it would take a miracle of engineering to bring them up past 5.3Ghz all core on 10nm.

I expect to see 5.8Ghz single core with 6Ghz overclocks from Intel and 5.2 to 5.3Ghz 8 core for gaming. Increased cache will help them keep up minimums but it's really going to come down to IPC for Intel this time. If they can't increase the IPC enough to keep up with AMD's clock increases from 5nm, they will be toast and frankly, I don't think they have enough left in the bag to do it. I will be pleasantly surprised if they do manage it but excuse me for doubting them given how bad 11900K flopped and how ho hum the 12900K was. The 12900KS is a limited and binned chip that was way late and double the power draw of AMD but Intel managed to push ahead by a steady 10%in gaming before AMD vollied at 3Dcache chip at the last second. lol, Honestly, I think they are riding on fumes right now. This is going to be a close fight but Intel has been pushed to the breaking point for years now and this isn't 2019 anymore, if games start pushing beyond 8 cores they are finished. They need to change their ways and soon.

AMD will deliver 5ghz on 32 threads at 100% load. Games that use 20% to 60% will probably flux from 5.2Ghz to 5.5Ghz. If you can get a 200Mhz OC out of this chip you could see 5.2Ghz all core which would certainly push the power and heat limits of current tech. OC averages of 5.5Ghz on video games would not surprise me at all. OC 5.8Ghz to 6Ghz single core would also not surprise me. I suspect Intel will be locked at 5.3Ghz on 8 cores tops, and they will be defeated unless they improve IPC enough to reverse AMD's gains.

[pax]

If you look at the chiplets on the 7000 package lisa su showed they are pretty big. Might indicate not just the bigger L2 cache but a layout design that allows for better heat dissipation. Normally we see chips drop in size from one gen to the next even when transistor budgets are increased.

[the_sextein]

When the 3000 series came out AMD was in a similar situation with 7nm vs Intel's 14nm and the IPC was neck and neck just like it is now. They lost because their design was a fail and couldn't clock up to the same levels as Intels because of heat. The 3700 was the same speed as the 3800X because the cache design didn't pay off ect, They have improved their ability to cool the chips with the 5000 series. They have had a silicon advantage for a long time and power draw has been solid for just as long. AMD's heat problem has been their limiting factor for a very long time. It's gotten better and better to the point where the Silicon advantage will actually be an advantage this time instead of a crutch. They are also releasing on time instead of late which marks a nice change and shows they are back on their feet again. Intel continued to push ahead every year but without challenge they have fallen into retardation as well. They made some drastic decisions to try and jump start their business but they still need to make some changes if they don't want to be left behind. I don't doubt that AMD has once again improved it's cooling efficiency with it's brand new design and tweaked the way it's cores clock based on load and thread count. They have had to alter their entire design to adjust to the new cutting edge silicon. Intel is the one who has run into heat problems this time and instead of fixing the problem they worked around it with a creative hack. Unfortunately the heart of their chip is stuck in a rut and they need to do the same thing AMD did if they want to keep up.

[Megaman]

So the goal post has been shifted with regards to Max power from the socket. AMD now claims 230w max power up 60w from their previous claims. 170w is the TDP though.


https://www.guru3d.com/news-story/am...p-to-230w.html

[the_sextein]

Lisa Su is the one who said it wrong during the presentation and then it was repeated. On the one hand it's using more power than they originally stated but on the other hand it means they are using enough power to realistically deliver on the clock and performance claims with sustained boosts.

It's using about 10watts more than my overclocked 9900K. It's still 150watts below the 12900KS which means even if Intel manages a monstrous 100watt power reduction with their tweaked 10nm, Intel will still use considerably more power despite running at much lower clock frequencies.

[Nunz]

I doubt that AMD was straight up lying about the clocks during the interview with the nerds. I believe them when they say that it was running at 5.5Ghz all core.

I think you're foolish to believe that. There's intentionally much less information than required to make an actual inference or determination of what you seen.

I saw it on screen with my own eyes but it is a light load on 6 cores which is not the same as a full load on 16 cores obviously.

You seen numbers without knowing what those numbers referred to.

My I9 9900K sucks 220 watts under full load so it doesn't surprise me that they are pushing 250 this time, It's still a very efficient design. Still, pushing up the ranks isn't as hard when the silicon isn't fighting you.

I'd hope that a 5nm processor is more efficient than a 14nm chip made 4-5 years ago..

Intel has been half assing it for a long time and their small core design was a quick bolted on idea after they got rid of trust fund stock market timmy.

Huh? Big-Little will be the direction AMD is going within the next two generations of chips. How is Big-Little a "quick bolted on idea" when it's literally the direction all CPUs are indicating to be going? I'm not sure you know what you're talking about here.

They are going to need a top to bottom modern design on cutting edge silicon designed by people that were actually hired based on skill level if they want to stay up with AMD at this point.

Yep, the crack has been smoked.

I think the 13900K is going to be very close but Intel will need to scrap it and start over after this if they hope to stay alive. Personally based on what I am seeing, I think intel will be defeated this round when it comes to gaming and will probably trade blows or lose by a small amount in multithreading. Intel's big cores are completely locked up as far as I can tell and it would take a miracle of engineering to bring them up past 5.3Ghz all core on 10nm.

Most 12900K can do 5.3Ghz all-core (P-Core) if you have sufficient cooling - 5.2Ghz is pretty much guaranteed, even on an AIO and a shitty P-Core bin, if you understand how to OC or can figure out how to turn on OCTVB and set a +2 Boost Bin. It seems to me you don't know much about ADL, which is fine, but don't keep talking about it like you understand it. With the efficiency improvements Intel is pushing for Raptor Lake, 5.3Ghz should be perfectly achievable. Wattage/Voltage get up there on the 12900K for 5.3Ghz, but with efficiency gains and Intel stating they have specifically worked on dropping voltage requirements for Raptor Lake, I think 5.3Ghz all-core won't be that difficult to see coming, especially since it's only 8c/16t P-Cores. The E-Cores.. I'm not sure how that will work. They are kind of a PITA with the limitations they impose on Ring clock, and I haven't seen Intel talk about a new solution for that, but I suspect that won't come until Lunar Lake. I'd have rather they kept 8 E-Cores or maybe go to 12, and give us 10 or 12 P-Core .. but either way, Raptor Lake will be a large jump in efficiency, which is really the only place Alder Lake struggles, at least once you start pushing voltage/current through the chip.

I will be pleasantly surprised if they do manage it but excuse me for doubting them given how bad 11900K flopped and how ho hum the 12900K was.

The 11900K was a chip that should have never been made. We know that. 12900K is ho hum? Interesting. I disagree.

The 12900KS is a limited and binned chip that was way late and double the power draw of AMD but Intel managed to push ahead by a steady 10%in gaming before AMD vollied at 3Dcache chip at the last second.

Yes, the 12900KS was not made for the regular consumer. It's a high binned/halo chip meant for OCers. You're paying for binned cores, and despite some of my buddies with 12900K/KF that are better binned, there's not as low of a floor on the 12900KS.

lol, Honestly, I think they are riding on fumes right now. This is going to be a close fight but Intel has been pushed to the breaking point for years now and this isn't 2019 anymore, if games start pushing beyond 8 cores they are finished. They need to change their ways and soon.

I disagree. They beat AMD to the punch on the future of chip design (Big-Little), managed to satisfy the market with options by offering two memory controllers on Alder Lake to ease the transition to DDR5, and they have a great architecture to work from. Alder Lake was a very strong, stable start after years of Intel's missteps. Saying they are "riding on fumes" .. I think you're terribly mistaken. They're designing great chips+chipsets right now.

AMD will deliver 5ghz on 32 threads at 100% load.

100% load doesn't mean anything. What wattage? Current? Instruction set .. etc? I can hit 100% load in one application and draw 100w, launch another and get 100% load and draw 220w. If you're talking about games .. yeah, might hold 5Ghz 16c/32t, but I suspect some games will cause power throttling unless it's opened up all the way; with that being said, I'm sure motherboard manufacturers, just like they do with Intel chips, will be removing power limits OOB.

Games that use 20% to 60% will probably flux from 5.2Ghz to 5.5Ghz. I think If you can get a 200Mhz OC out of this chip you could see 5.2Ghz all core which would certainly push the power and heat limits of current tech. OC averages of 5.5Ghz on video games would not surprise me at all. OC 5.8Ghz to 6Ghz single core would also not surprise me. I suspect Intel will be locked at 5.3Ghz on 8 cores tops, and they will be defeated unless they improve IPC enough to reverse AMD's gains.

This is where I find it surprising you think Intel would be locked at 5.3Ghz on 8 P-Cores when most people ADL chips can already do that. Hell, my 12700K in low-thread games can do 5.5Ghz stable and I can run 5.2Ghz all-core if I pump the voltage up, but gaming at 4K means I'm 99.99999999% GPU bound even in a lot of CPU-centric games, so I just run 5.3x4/50x8. My 12700K is also an atrocious bin. I run more voltage for 5Ghz than a lot do for 5.2Ghz, but that's OK. It's just a placeholder until Raptor Lake

[the_sextein]

I saw numbers knowing exactly what they were referring to. I understand that if a chip can run at 5Ghz all core under full load then it could realistically run at 5.5ghz all core under light load. Running Prime 95 with 100% load on all threads vs running daz 3d at 100% load on all threads isn't the same. It's not always the clock frequency that is the problem, it's how much data and what type of data being processed that matters. Both use 100% load on all cores at the same frequency but one runs 10C more hot and uses 60 watts more power. Gaming runs all cores at the same 5Ghz and runs at 65C hardly using anything. I could push my I9 9900K as far as the socket would allow without problems in gaming. It's the all core full load with AVX 512 that really crippled the 9900K.

Big/ Little isn't Intels problem it's that they tacked on the little cores without updating the the heart of their chip enough and if they wanted to realistically keep up they should have considered more large cores and better silicon.

The new AMD compared to my 9900K chip is 2X the cores and threads with faster clocks and 4X the cache with a 10 watt increase. I'd say its way better than a chip from 4 years ago. If you compare that to Intel's 380watt 8 core chip it's pretty impressive.

The 12900K came out late because the 11900K failed and it didn't take back multithreading and beat a 6 month old chip by an insignificant 4% in gaming despite using way more power.

As for why I think Intel will stay around 5.3Ghz with 8 P cores is because their new HEDT designs are not blowing past that figure and their power consumption is already out of control. They aren't making these chips for liquid nitrogen, they need to work with a standard AIO that people actually use.

I have nothing against Intel as a company but they appear to be trailing behind and making some questionable decisions in my opinion. If they deliver an IPC miracle I would be more than happy to buy from them but right now based on the realities of tech limitations, I believe AMD is going to deliver better gaming and comparable multithreading while AMD's 8 extra big cores can be used for gaming in the future if it's needed. I believe AMD can deliver this with lower power usage than Intel. I could be wrong but we will have to wait and see. Throwing your chips behind a companies success or failure adds a little fun to the conversation and we have nothing better to do than speculate at this point lol. GPU limitations are going to change drastically when these chips hit the scene, so hiding behind a GPU bottleneck won't be an option like it was the last couple CPU launches.

[Nunz]

I saw numbers knowing exactly what they were referring to. I understand that if a chip can run at 5Ghz all core under full load then it could realistically run at 5.5ghz all core under light load.

What is "light load"?? You keep repeating this. Do you mean when less threads are used, it will clock higher? AKA, 4 cores go to 5.5Ghz and the rest stay at 5.0Ghz? That is NOT "all-core". That is per-core and/or turbo ratio (at least in Intel terminology).

It's the all core full load with AVX 512 that really crippled the 9900K.

Weird you say this since the 9900K does not support AVX512 instruction, at least not officially from anything I've seen, or what I can remember from when I owned one.

Big/ Little isn't Intels problem it's that they tacked on the little cores without updating the the heart of their chip enough and if they wanted to realistically keep up they should have considered more large cores and better silicon.

You are being extremely vague. What is "updating the heart of their chip enough"? Gracemont, high-efficiency cores that are faster than the cores in your 9900K is not "updated enough"? Golden Cove cores with the highest IPC on the market at release? That's "not enough"...? Ring re-designed (thanks Rocket Lake!) to correct the horrendous architecture issues of Comet Lake? I really don't understand how you could think Intel hasn't done major work between Rocket Lake to Alder Lake.

The new AMD compared to my 9900K chip is 2X the cores and threads with faster clocks and 4X the cache with a 10 watt increase. I'd say its way better than a chip from 4 years ago. If you compare that to Intel's 380watt 8 core chip it's pretty impressive.

10 watt increase.. uhh, where? That's "TDP", which as we all know, is not equivalent to actual power consumption. Intel's 8-core chip pushing 380watt.. where? Do you realize how much voltage you have to push into a 12900K to be in the 380watt arena? That's no where near stock, even with AVX512 instruction.

The 12900K came out late because the 11900K failed and it didn't take back multithreading and beat a 6 month old chip by an insignificant 4% in gaming despite using way more power.

12900K is more power efficient than my 10900K and an 11900K I briefly played with in a buddies rig. What is "way more power"? In what usage? If you are going to say it's using more power in CB23 and Blender, then yeah sure (except the 10900K which was an actual pig).. but you're taking a CB23 workload wattage and then using that to explain how the 12900K pulls more power in games. That's not how that works. Remember that the power usage you use in gaming is not the power usage you see in CB20/23 and Prime95. I see ~90-100w during gaming at 5.1Ghz all-core @ 1.32v load. How is that "way more power"? I see 5900X's regularly using 80-90watts, or more, in real-world gaming scenarios just with PBO enabled.

As for why I think Intel will stay around 5.3Ghz with 8 P cores is because their new HEDT designs are not blowing past that figure and their power consumption is already out of control. They aren't making these chips for liquid nitrogen, they need to work with a standard AIO that people actually use.

How is HEDT relevant here? More P-Cores as ADL-X will have is going to force core clocks down. That's normal, just like we've seen on every single HEDT release from Intel previously. I fail to see how this relates to what we'll see on Raptor Lake. ADL-X, which from everything I've read, will be using ADL Golden Cove cores for P-Cores. Raptor Lake is using Raptor Cove - the arch that Intel has been tweaking for significant voltage reduction, and a rumored 25% reduced power draw.

[the_sextein]

A gaming load is a light load that uses less than half the power and runs 30C lower which makes it obvious why it would clock higher during gaming. This isn't hard to understand.

Intels 8 core heart has been clock locked since the 9900K. All improvements to the 10900K and onward have not been ignored but then again, I never said Intel has done nothing, just that they shouldn't have tacked on little cores to a an 8 core design that needed a complete redesign and better silicon.

The updated power for the socket is 230watt on the new AMD 7950X with 170what TDP.

Intels 12900KS has topped 380watts in power testing vs AMD's 220 watts over at techpowerup.com. 160watts more power for the same multi threading and 10% more gaming performance at 10C higher temps and 2 years late.

HEDT is relevent because it's Intels highest next gen tech and it isn't going over 5.2 Ghz on 8 core. You will also notice that none of the leaks or rumors have mention going over 5.2Ghz because they all know better lol. I feel that I am being generous with 5.3Ghz on 8 cores. It's not likely they will even reach that without blowing past 110C and 400 watts on 10nm.

[Nunz]

A gaming load is a light load that uses less than half the power and runs 30C lower which makes it obvious why it would clock higher during gaming. This isn't hard to understand.

It's hard to understand when you keep calling it something it's not. All-Core is not what you're describing.

Intels 8 core heart has been clock locked since the 9900K.

I dunno, you'd be hard-pressed to find people hitting 5.4-5.5Ghz in low-thread scenarios, but that's pretty common with ADL, so..

just that they shouldn't have tacked on little cores to a an 8 core design that needed a complete redesign and better silicon.

The cores were completely redesigned, that's the whole "Big-Little" thing .. and what's the point of more than 8 cores? I've heard you talk so much about consoles being 8c/16t back when Ryzen was touting all their cores .. now all of a sudden they should throw more than 8? I think 8 works fine, especially with high-efficiency cores to handle background tasking. Load up cache, not P-Cores.

Intels 112900KS has topped 380watts in power testing vs AMD's 220 watts over at techpowerup.com. 160watts more power for the same multi threading and 10% more gaming performance at 10C higher temps and 2 years late.

I see 360watts at stock in P95. That's irrelevant to gaming or (most) real-world workloads. Cinebench pulls 298w. More power, yes, but your "160watts more power" claim is based off of Prime95.. lol. I'm looking at the TPU 12900KS review right now under the "power consumption" page. Again, when you blast super high current workloads through the chip, yeah, it's going to get hot. Looking at what my chip draws in gaming .. 90-100w is efficient considering how fast it is, and that's with the Gracemont cores enabled. Disable them and I could lower that consumption even further, even while raising the Ring clock up.

HEDT is relevent because it's Intels highest next gen tech

No it isn't. HEDT coming out is "Alder Lake-X" which is rumored (because nothing is confirmed yet) to use Golden Cove P-cores - the same cores you see on a 12900K - which do not have any of the previously mentioned voltage reductions or efficiency improvements Intel has touted for Raptor Lake. HEDT is irrelevant for a discussion about Intel's future mainstream release competing against AM5, because ADL won't be competing against AM5 - Raptor Lake will.

and it isn't going over 5.2 Ghz on 8 core. You will also notice that none of the leaks or rumors have mention going over 5.2Ghz because they all know better lol. I feel that I am being generous with 5.3Ghz on 8 cores. It's not likely they will even reach that without blowing past 110C and 400 watts on 10nm.

I mean, look at X299. Plenty of people OC'd those chips up to 5Ghz all-core on an 18-core 10980XE. The HEDT ADL-X platform is not indicative or foreshadowing of what we'll see with Raptor Lake, but hey.. believe what you want.

[the_sextein]

So running all cores at a given frequency is not all core? Claiming that AMD can't do 5.5Ghz on all cores while gaming and that they are straight up lying but agreeing that 5Ghz all core under full load is possible with 250watts or more? Doesn't make alot of sense dude.

I could play 5.3Ghz all core on my I9 9900K when gaming and keep the temps in the low 70's with stability. Beyond that would require more power than the socket would provide without burning up.

If Intel completely redesigned their 8 core chip then they did a piss poor job because it's been stuck between 5Ghz and 5.2Ghz all core for 4 years despite a drop down to 10nm.

If HEDT for next gen isnt' next gen but rather their old gen then ok but we will have to wait and see. I'm 90% certain you won't be seeing 5.5Ghz clocks on all 8 cores from Intel. If you have seen the power usage then you know why. I don't care if you can manage it in a game. Run Daz 3d on your CPU and tell me how much power it uses lol. Prime 95 is even worse.

[the_sextein]

I pooped all over AMD's extra cores back 3 years ago because the new consoles with 8 cores were about to hit the scene and it was sure fire bet that games would not use more than 8 cores for the 4 year life cycle of the product. Also, for the few things that we can actually use more than 8 cores on, a 2 or 4 extra cores at lower clocks isn't significant enough for me to care. It's not 2019 anymore, consoles could come out within the next 4 years that could use more than 8 cores and now they are packing 16 cores at decent clock speeds so we will actually see significant improvements in apps that can use more than 8 cores. Different times. Like I said, I would still buy Intel if they perform an IPC miricle because multi threading is still second rate on my list of importance but it is more important now than it was then. Meaning is has minor importance vs being useless sales gimmick.

[Nunz]

So running all cores at a given frequency is not all core? Claiming that AMD can't do 5.5Ghz on all cores while gaming and that they are straight up lying but agreeing that 5Ghz all core under full load is possible with 250watts or more? Doesn't make alot of sense dude.

Because I don't think it will be all-core. It'll be per-core or a ratio clock, where if only 4 cores are loaded, 4 cores will go to 5.5Ghz, or if 8 cores are loaded, 8 cores will go to 5.3Ghz. If all 16 are loaded, it will go to 5Ghz. Once it hits power limits and/or temp limits, it'll drop down from there. Makes plenty of sense

AMD calls it PBO but it's really not all that different despite their marketing. It's also sort of a form of OCing, I guess. Depends how you look at it. Is stock considered PBO disabled, or enabled? Interesting discussion on that I suppose.

I could play 5.3Ghz all core on my I9 9900K when gaming and keep the temps in the low 70's with stability. Beyond that would require more power than the socket would provide without burning up.

You had/have a good chip then. Unfortunately, not many did 5.3Ghz - I was locked at 5.2Ghz max, if I remember correctly? I had to pump a lot of voltage too. Fortunately, it's much more common now, and that's with significant IPC improvements and added instructions, like AVX512 .. which was not supported on the 9900K.

If Intel completely redesigned their 8 core chip then they did a piss poor job because it's been stuck between 5Ghz and 5.2Ghz all core for 4 years despite a drop down to 10nm.

Yes, because clock speed is everything. It's not. AMD has proven that for years!

If you have seen the power usage then you know why. I don't care if you can manage it in a game. Run Daz 3d on your CPU and tell me how much power it uses lol. Prime 95 is even worse.

I've seen the power usage because I actually own the chip. You keep changing the goal posts. Are we talking about power consumption in gaming loads, or Daz3D/CB23/P95?

5.5Ghz all-core on a 12900K is possible right now on most chips if you are only worried about being game stable and don't care about voltage (which you shouldn't if it's just a gaming workload). If you're talking about Daz3D and compute workloads.. then yeah. Does Daz3D use AVX512? If so, the 12900K with AVX512 instruction enabled is going to take your 9900K and yeet it into a ****ing volcano even if you ran it at stock vs your 9900K @ 5.3Ghz.

But that's a completely different scenario, and considering with the AMD chip, we are talking about 5Ghz all-core in games with 5.5Ghz light load boost .. in games .. I'm not sure why it's even being brought up.

[Nunz]

Before I bow out, so it doesn't seem like I'm on some "death to AMD" train here, despite my belief the hype-train is getting too nuts over marketing slides..

I want to see AM5 crush it. More competition, better hardware, and hopefully better pricing. I'm sure the 7000 series is going to perform extremely competitively, even if I am skeptical on the clock speeds.

I'm curious how long it will be before we see Zen4 X3D, which I think will be pushed much quicker considering the success story of the 5800X3D.

I was shocked to see AMD say "nope" to DDR4 support, though. I expected them to roll first gen X670 as DDR4 with their "X670E" variants going DDR5. DDR5 pricing has come down significantly though which is great. You can grab a 32GB kit of great Hynix memory (good bin will get you into the 7000s with the right chip/board, at least on Alder Lake right now) for $300 or so. Expensive still, but 32GB of RAM that is quad-channel on only two sticks is beauty

also.. my $$ invested in AMD stock lolol

[the_sextein]

Oh no problems nunz, I respect your opinion but I'm a stubborn man. If you have looked at my past comments then you know I'm not an AMD fanboy. I'm just convinced that Intel is clock and power limited. I'm not saying clocks mean everything, I've said it multiple times that Intel has to deliver major IPC gain if they want to win.

I also thought they used a system like Intel that more intelligently scaled the boosts based on the number of cores being used but AMD clearly stated that all 32 threads were running at 5.5Ghz during that demo. I don't think it's impossible at all. If it can run 5Ghz under full load than it probably could run 5.5Ghz in a game but I would think boosting a certain number of cores being used would be a superior solution so I'm kind of surprised they didn't go that way to be honest.

[Nunz]

AMD PBO is a little different. It's similar to Intel Boost with TVB (Thermal Velocity Boost) enabled. TVB came out with the 10900K, wasn't available on the 9900K. TVB kind of brought some parity to both, all it does is add more factors to what can influence the clock-speed. Back before TVB, it was simply load. If 4 cores are loaded, run xx ratio. If 8 cores are loaded, run xx ratio. With TVB, you add temperature and voltage control, so now..

4 cores are loaded, run 54x ratio
If temperature hits 60c (ex), run 53x ratio
If temperature hits 70c, run 52x ratio

It also adds voltage control, so as temperature increases, the chip will demand slightly more voltage to compensate for the temperature increase (because hotter the chip, the more voltage it needs). Cooler the chip, it'll actually lower the voltage. It's a really interesting concept that, if utilized correctly, allows a good OCer to do some ridiculous ****.

There's a guy on OCN who can run 58x single core, 56x 4 cores, and 55x 8 cores. In games, his temps are below 70c, so with TVB, he actually gets 56x on all 8 cores. He uses the boost binning feature, another part of TVB.

Either way, things are much more complicated with OCing these days. The simple "all-core" OC is a really inefficient way to overclock and leaves a lot of performance on the table, especially for gaming, since we rarely see every core loaded.

[the_sextein]

lol yeah I settled for 5Ghz all core on my 9900K because of that. Daz3d pushes my temps up to 85C and Prime 95 breaks into the 90's. For gaming it could handle it but it would run out of power and shut off if I tried 5.2Ghz all core in Daz studio without an avx offset.

[the_sextein]

Regarding PBO, I don't know, they said it was a sample from April that didn't utilize the 230watt socket power or have any overclocking. It's possible that they didn't have it enabled yet or that this was a golden sample but I doubt AMD really wants to put unrealistic expectations on it's products because I think the backlash will be pretty bad for them if they do. Who knows.

[pax]

Robert Hallock is seeding the rumor mill again but looks like they really are sandbagging the next part:

He also said AMD "was conservative on that 15% ST number in 4 or 5 unique ways"

https://www.reddit.com/r/Amd/comment...e_performance/


https://www.youtube.com/watch?v=DbWhHkhBGvM&t=3203s

[yt]DbWhHkhBGvM[/yt]

At the 53 min mark