AMD Ryzen 2000-series "Pinnacle Ridge" CPUs Get Soldered IHS

[btarunr]

AMD's second-generation Ryzen 2000-series "Pinnacle Ridge" processors, which succeed the company's first Ryzen "Summit Ridge," reportedly feature soldered integrated heatspreaders (IHS), according AMD spokesperson "AMD_Robert" on Reddit. This would make the chips different from the Ryzen 2000G-series "Raven Ridge" APUs launched earlier this week, which come with a thermal paste between the IHS and the die. Soldered heatspreaders are generally known to have better heat transfer between the IHS and die, when compared to packages with thermal pastes between the two; and are more expensive to manufacture. They remove the need to "de-lid" the processor (remove the IHS). Ryzen 2000-series processors are expected to debut in April 2018.



View at TechPowerUp Main Site

 

[londiste]

Credible sources on the AMD community forums:
https://www.reddit.com/r/Amd/comments/7xhren/_/du916i5 https://www.reddit.com/r/Amd/comments/7x6g6r/_/du7yhnf

 

[ratirt]

Wonder if soldered IHS will boost frequency higher. It would have been nice to get the Ryzen 4.5Ghz or around

 

[londiste]

Wonder if soldered IHS will boost frequency higher. It would have been nice to get the Ryzen 4.5Ghz or around

All Summit Ridge Ryzens are soldered. Soldered IHS does not boost frequency, all it can do is help keep temperatures in check.

 

[trparky]

There have been talks in other places on the Internet that this series of Ryzen isn't anywhere close to what many of us wanted it to be. Those people are suggesting that we're only going to be seeing at most a 200 MHz bump in speeds, far short of what a lot of us wanted. I myself was hoping to see base clocks closer to 4 GHz (perhaps 3.8 GHz) with boost clocks closer to 4.3 GHz along with overclockablity closer to 4.6 GHz. Alas, that ain't happening.

Guess that it's going to be the 8700k for my next build in a month or two. At least one can clock that thing to 4.4 GHz quite easily even without de-lidding it just as long as you have a good cooling solution.

 

[Imsochobo]

Wonder if soldered IHS will boost frequency higher. It would have been nice to get the Ryzen 4.5Ghz or around

Amd's clock speeds are determined largely by it's node process (current 14nm)
They said 10% performance which is likely 2-3 % of ipc and other changes thus 7-8% clock speeds which should equate rathe close to 4.5ghz capability.

if the apu's are an indication we'd be seeing quite a good uplift to counter the mighty 8700K and 8600K (Rest of intel's lineup is mostly countered by existing products).
I will guess AMD with it's Pinnacle Ridge will dominate lower end market like we saw with KL vs Ryzen last spring but still struggle hard vs the top I5,I7 K sku's but at no time a bad deal.

There have been talks in other places on the Internet that this series of Ryzen isn't anywhere close to what many of us wanted it to be. Those people are suggesting that we're only going to be seeing at most a 200 MHz bump in speeds, far short of what a lot of us wanted. I myself was hoping to see base clocks closer to 4 GHz (perhaps 3.8 GHz) with boost clocks closer to 4.3 GHz along with overclockablity closer to 4.6 GHz. Alas, that ain't happening.

Guess that it's going to be the 8700k for my next build in a month or two. At least one can clock that thing to 4.4 GHz quite easily even without de-lidding it just as long as you have a good cooling solution.

Cause 8700K doesn't come in 4ghz base, it's a 3.7 ghz base.
R5 1600x is 100mhz below as base but turbo is way off, if amd increase base to 3.8 and boost to 4.4 and bring forth the apu's 2xxx improvements it's all good.
It's all that have been speculated based on amd's performance claims, the 200mhz is just guessing by people but I fail to see how amd can achieve 10% more performance without bumping clocks at least 300 mhz.
No-one knows how the 12nm node performs and it's under very close wraps it seems as no leaks are present while production have been going on for months, I wouldn't go forth with not happening but rather wait and see.
Their performance number touted certainly achieves a closer gap between Ryzen 2xxx vs CL than Ryzen 1xxx was against KL while reducing the emphasis on the "content creator vs gamer" argument and makes the trade offs smaller.
we'd have to wait and see how much it's pushed.

Just as a basis for performance talks:
https://www.extremetech.com/wp-content/uploads/2017/09/AMD-12nm.png

above 16NM which in my opinion is greater than amd's 14nm, so... anyone's bet.

 

[trparky]

rathe close to 4.5ghz capability.

Based upon engineering samples that some have gotten a hold of the answer is no, that's not going to be a reality.

 

[FrustratedGarrett]

Based upon engineering samples that some have gotten a hold of the answer is no, that's not going to be a reality.

Sources?

 

[trparky]

Edit
From a logical way of thinking, I believe he may be right about a lot of what he said in the video. A 10% improvement in clocks equals to roughly 200 MHz more and that's not at all going to be enough to really compete with the heavyweight champion that is the 8700k. The only way I figure that things will get really better in terms of clocks for AMD is after Ryzen 2.0, this generation seems more like Ryzen v1.5 than a full 2.0 version.

 

[NdMk2o1o]

That's because it's ryzen + not ryzen 2, if they do 4.2/4.3 fairly easy then I might upgrade, I say fairly easy like 3.8/3.9 is fairly easy for most first gen ryzens, for me I can do 3.9@1.335 vcore but I need 1.45 just to do 4.0 that's stable enough to run cine bench. So an extra 300/400mhz would be welcomed.

 

[Casecutter]

The only way I figure that things will get really better in terms of clocks for AMD is after Ryzen 2.0, this generation seems more like Ryzen v1.5 than a full 2.0 version.

Whoever wrote that has either S#it for brains or promoting fake crap.

Ryzen+ is not anything more that a process re-spin that includes a few small design improvements. It's not even any true % improvement, a guy like that shouldn't consider a version... it's just Ryzen+. Sure we may see clocks go up 100-200Mhz, but to buy a1600X/1700X today at their low prices, or wait to see what Ryzen+ can offer though at what will be for awhile is more MSRP is neither here nor there.

 

[Valantar]

Edit
From a logical way of thinking, I believe he may be right about a lot of what he said in the video. A 10% improvement in clocks equals to roughly 200 MHz more and that's not at all going to be enough to really compete with the heavyweight champion that is the 8700k. The only way I figure that things will get really better in terms of clocks for AMD is after Ryzen 2.0, this generation seems more like Ryzen v1.5 than a full 2.0 version.

Not to sound rude, but you need to work on your math. High-end Ryzen SKUs boost to 3.8-4.1GHz for one core (actually 4.2 for TR with XFR) with clocks dropping off a few hundred MHz from that to all cores. And, let's see, 10% of 3800 is... Oh! Not 200! Nearly twice that, in fact. Now there's a surprise, no?

As for this not being "a full 2.0 version", most of us figured that from the subtle hint that this is named Zen+, and Not Zen 2.

If Zen+ gives us a 10% clock boost, that's exactly what AMD needs for year-on-year improvement. I'd be very happy with that. I wouldn't upgrade my 1600X, but upgrading your CPU more frequently than every 3-4 (or more!) years is pretty damn dumb unless you run CPU-bound workloads constantly. The key here is that AMD needs to eat away at the small lead that Intel has, and particularly Intel's key strengths. Clock speed is one of those.

 

[Steevo]

In this thread; "Intel $360 chip is faster than this AMD $199 chip"

Well..... duh?

 

[sinnedone]

Soldering is more expensive than using TIM, correct?

If that assumption holds true than the reasoning behind this announcement is simple.


AMD was not able to achieve the clocks we were all expecting. In order to bump up the clocks they bumped the voltage tables, but that increased heat.

Solution....solder heat spreader.


Good theory? I mean they're not doing it so we can get higher overclocks. AMD is doing this so they can get higher clocks. I figure it's the same way as Ryzen. Remember when there were rumors that AMD was having problems getting Ryzen to clock past 2ghz. They're solution was to overclock themselves leaving no room for the tinkerers to achieve anything higher. I think we're going to see a repeat of that.

 

[Valantar]

Soldering is more expensive than using TIM, correct?

If that assumption holds true than the reasoning behind this announcement is simple.


AMD was not able to achieve the clocks we were all expecting. In order to bump up the clocks they bumped the voltage tables, but that increased heat.

Solution....solder heat spreader.


Good theory? I mean they're not doing it so we can get higher overclocks. AMD is doing this so they can get higher clocks. I figure it's the same way as Ryzen. Remember when there were rumors that AMD was having problems getting Ryzen to clock past 2ghz. They're solution was to overclock themselves leaving no room for the tinkerers to achieve anything higher. I think we're going to see a repeat of that.

You're quite a bit off here. Given that Ryzen chips ship at ~4GHz, they definitely didn't have trouble getting them past 2 at any stage when actual silicon was being made.

Now, it's unclear if the hard clock speed limit on Zen is caused by the arch, node or a combination of the two (although the node is not designed for very high clock speeds, so that's a hint at least), but saying Ryzen is "overclocked" out of the box is downright silly. Ryzen ships well within the efficiency bounds of its design, with only the highest-end SKUs pushing those limits on single cores. If Ryzen was "factory overclocked" that implies pushing it to its power/thermal limits, which is blatantly untrue. Heck, Ryzen is arguably significantly more efficient than KBL at stock clocks.

Of course the soldered IHS helps for this, but most Ryzen chips could run comfortably at the same clocks and ordinary air cooling at ~10-15C higher temps (which is where TIM likely would land you). As for why AMD prioritized the extra cost for a soldered heatspreader? A good question. My suggestion is image building/PR: AMD has long had a reputation for running hot and slow. The FX series was laughably inefficient, with a lot of chips using >=125W. I think AMD went for a soldered IHS simply to underscore the efficiency of their new architecture, by saying "hey, look, we have this 65W eight-core CPU, and by the way it runs cool and quiet with a small stock heatsink too!" The R7 1700 is still a huge hit, and this gave AMD a massive PR boost. If it had struggled to stay cool under load with the stock heatsink, that would have looked significantly worse, even if most users still buy better heatsinks. Worth it? That is almost impossible to say, but it plays into the massively improved reputation AMD has gained over the last year.

 

[Imsochobo]

You're quite a bit off here. Given that Ryzen chips ship at ~4GHz, they definitely didn't have trouble getting them past 2 at any stage when actual silicon was being made.

Now, it's unclear if the hard clock speed limit on Zen is caused by the arch, node or a combination of the two (although the node is not designed for very high clock speeds, so that's a hint at least), but saying Ryzen is "overclocked" out of the box is downright silly. Ryzen ships well within the efficiency bounds of its design, with only the highest-end SKUs pushing those limits on single cores. If Ryzen was "factory overclocked" that implies pushing it to its power/thermal limits, which is blatantly untrue. Heck, Ryzen is arguably significantly more efficient than KBL at stock clocks.

Of course the soldered IHS helps for this, but most Ryzen chips could run comfortably at the same clocks and ordinary air cooling at ~10-15C higher temps (which is where TIM likely would land you). As for why AMD prioritized the extra cost for a soldered heatspreader? A good question. My suggestion is image building/PR: AMD has long had a reputation for running hot and slow. The FX series was laughably inefficient, with a lot of chips using >=125W. I think AMD went for a soldered IHS simply to underscore the efficiency of their new architecture, by saying "hey, look, we have this 65W eight-core CPU, and by the way it runs cool and quiet with a small stock heatsink too!" The R7 1700 is still a huge hit, and this gave AMD a massive PR boost. If it had struggled to stay cool under load with the stock heatsink, that would have looked significantly worse, even if most users still buy better heatsinks. Worth it? That is almost impossible to say, but it plays into the massively improved reputation AMD has gained over the last year.

This seems to be mostly the node causing issues as other chips made with same process where the design was made on 22nm prior seems to have the same barriers ryzen had with power effeciency.
Mind you, two fabs, same node tech, two widely different chip sizes but that the 22nm is better than the 14nm in raw max clock is telling us something.
I can try to find the info (it was real nerdy and comparing an arm vs x86 is also iffy, but it gives strength to the node being big part of issue)

On the flip side the Ryzen chips are absolutely hilariously efficient like vega, downclock a vega to 1ghz with 1ghz hbm2, give it the voltage appropriate and find me a nvidia card worthy of competition at the wattage.
This have nothing of interest for us high end gamers and HEDT builders but if we do Epyc vs Xeon battle and vega in datacenter vs Nvidia's offerings we can see it is working very much in Amd's favor.
All of this while ultimately for us having a very inferior node at this time but the fab analysts is saying that Intel's lead is fading away quickly so ...

we'd just have to wait and see to confirm.

 

[Casecutter]

Ah, doesn't Ryzen right now have that HS solder? Why would anyone think it would (or suppose it could be changed) for the Ryzen+ ?

If at some point we hear Ryzen II will go TIM then that might be something; pointing to AMD has a new process that provides even more efficiency and less TDP. Who thinks what been termed a nothing more than a "Tock" (re-spin that includes a few small design improvements) would be able to use the lesser TIM?

 

[Valantar]

Ah, doesn't Ryzen right now have that HS solder? Why would anyone think it would (or suppose it could be changed) for the Ryzen+ ?

My thoughts exactly. Just because the low-end (<=$170, <=65W) APU series gets TIM, why should that have implications for the higher-end CPU series? It doesn't really add up, unless one is frantically looking for things that might go wrong to have something to complain over. Of course, that's not an uncommon mode of thinking in tech enthusiast circles, but it's still rather silly.

 

[NdMk2o1o]

AMD have generally chosen soldered over TIM previously so I think it's reading into nothing, they havent changed anything but hey we can speculate until the cows come home about how or why with all different theories being tossed around when I don't think it has any real bearing of significance

 

[evernessince]

Based upon engineering samples that some have gotten a hold of the answer is no, that's not going to be a reality.

Lol, what? You do know that Zen engineering samples ran at 3.0 - 3.2 GHz right? Their clock speed don't mean squat to the final product.

 

[CheapMeat]

Maybe they care about making high quality, better products? Instead of skimming and overpricing like Intel?

 

[eidairaman1]

AMD have generally chosen soldered over TIM previously so I think it's reading into nothing, they havent changed anything but hey we can speculate until the cows come home about how or why with all different theories being tossed around when I don't think it has any real bearing of significance

Less chips returned from die crushed damage since the ihs spreads the pressure over a greater area than a die can...

At least thermal tranfer will be better than using a cheap thermal interface material.

 

[Dave65]

All Summit Ridge Ryzens are soldered. Soldered IHS does not boost frequency, all it can do is help keep temperatures in check.

Someone needs to tell that to Intel

 

[Valantar]

Someone needs to tell that to Intel

Intel likely wouldn't ship KBL or CFL at higher stock clocks even if they soldered the IHS on, given that they would then have to increase the TDP rating. You can clock a 7700k or 8700k to >5GHz, but power consumption blows far past the 95W mark by that point. In other words, a soldered IHS wouldn't help Intel increase clocks if they weren't also willing to rate their chips at a higher TDP (which they don't as that would increase motherboard costs (requiring a minimum power delivery of, say, >125W rather than 95W is significant, especially as this would be required of every platform, from B310 (or whatever it ends up being called) and upwards) and make them look bad in terms of heat/power).

They did something similar with the KBL-X series, which arguably has been a massive flop, and seems to be regarded mostly as a joke. "Hey, who wants a more expensive CPU with 2% higher clocks? You just have to buy a $300 motherboard and disable half its functionality. It overclocks really high, though! Maybe even 200MHz higher!"

In other words: the reason why Intel chips OC higher delidded is because the IHS is designed for <~100W heat dissipation, and OCing brings it far beyond that. At stock clocks, the IHS and TIM are perfectly sufficient, keeping the CPU well within acceptable temperature ranges. Would soldering improve on this? Of course. But it still wouldn't allow Intel to increase stock clocks without increasing the TDP to match.

 

[ghazi]

Intel likely wouldn't ship KBL or CFL at higher stock clocks even if they soldered the IHS on, given that they would then have to increase the TDP rating. You can clock a 7700k or 8700k to >5GHz, but power consumption blows far past the 95W mark by that point. In other words, a soldered IHS wouldn't help Intel increase clocks if they weren't also willing to rate their chips at a higher TDP (which they don't as that would increase motherboard costs (requiring a minimum power delivery of, say, >125W rather than 95W is significant, especially as this would be required of every platform, from B310 (or whatever it ends up being called) and upwards) and make them look bad in terms of heat/power).

They did something similar with the KBL-X series, which arguably has been a massive flop, and seems to be regarded mostly as a joke. "Hey, who wants a more expensive CPU with 2% higher clocks? You just have to buy a $300 motherboard and disable half its functionality. It overclocks really high, though! Maybe even 200MHz higher!"

In other words: the reason why Intel chips OC higher delidded is because the IHS is designed for <~100W heat dissipation, and OCing brings it far beyond that. At stock clocks, the IHS and TIM are perfectly sufficient, keeping the CPU well within acceptable temperature ranges. Would soldering improve on this? Of course. But it still wouldn't allow Intel to increase stock clocks without increasing the TDP to match.

I believe you are conflating TDP and power draw. TDP is thermal design power, which is an imprecise rating for system integrators and manufacturers of cooling solutions and the like to tell them how hot they can expect the component to be. If soldering the IHS leads to reduced temperatures then it also should allow Intel to reduce the TDP rating.

As far as higher clocks at stock go, obviously the difference there would be negligible as you'd be looking at 100-200MHz at best without exceeding the power delivery specifications of the socket/platform. Intel specifically sells these chips with overclocking as a marketed feature, though, so there's no reason to pretend that stock is all that should matter to them or their customers here.

KBL-X was a joke because of the platform it was on. Intel could have instead specified beefier power delivery for Z270 specifically and then release 7740X as a Z270-only chip. Much less confusion and fragmentation, easier for most to upgrade I assume -- but no, someone in marketing told them that having the "most scalable desktop platform" meant something to consumers; I doubt it actually had anything to do with getting KBL to OC better.