Friday, August 17th 2018

Intel Confirms Soldered IHS for 9th Gen Core Series

Soldered integrated heatspreader has been a longstanding demand of PC enthusiasts for Intel's premium "K" mainstream-desktop processors. With AMD implementing it across all its "Summit Ridge" and "Pinnacle Ridge" Ryzen AM4 processors, just enough pressure for built on Intel. The company, in a leaked slide, confirmed the feature-set of its upcoming 9th generation "K" Core processors, which highlights "STIM" (soldered thermal interface material) for this chip. It shows that STIM could be exclusive to the "K" series SKUs, namely the i9-9900K, i7-9700K, and i5-9600K.

The slides also list out the clock speeds and cache sizes of the three first 9th generation desktop SKUs, confirming that the Core i7-9700K will indeed be the first Core i7 desktop SKU ever to lack HyperThreading. The TDP of the 8-core chips don't seem to breach the 95W TDP barrier Intel seems to have set for its MSDT processors. The slides also seem to confirm that the upcoming Z390 Express chipset doesn't bring any new features, besides having stronger CPU VRM specifications than the Z370. Intel seems to recommend the Z390 to make the most out of its 8-core chips.
Source: VideoCardz
Add your own comment

93 Comments on Intel Confirms Soldered IHS for 9th Gen Core Series

#51
Vya Domus
hat said:
Not sure how much we should care about HT these days. It would be nice if somebody with a processor with a lot of cores (8 or more) could run a few tests to see what the difference is with it on and off.
The usefulness of HT decreases with the number of cores to the point where it might even hurt multithreaded performance. This isn't exactly news but it's largely unknown to most people.

https://www.pugetsystems.com/labs/hpc/Hyper-Threading-may-be-Killing-your-Parallel-Performance-578/

Single core performance can be hurt too by hyperthreading.
Posted on Reply
#52
altcapwn
The 9th gen seems awesome, but I just can't give my money to Intel anymore. They just abused too much of their monopoly and I would feel stupid to do like nothing ever happenned. It encourages that kind of attitude towards consumers; just like EA and Activision do with their shitty games and microtransaction.

I mean, yeah the goal is to do money thanks to capitalism, but can't they respect us a bit at least and not take us for plain morons?
Posted on Reply
#53
notb
Vya Domus said:
Doubling of core counts.
Both 6C and 8C CPUs must have been in works for years. Intel might have delayed them to lower costs, but there isn't that much revolution.
But sure, if AMD wasn't so busy ruining itself, Intel might have given us those 6 cores a year earlier or something.
Launching a new generation of CPUs without changing the socket or chipset.
???
It's a second generation using the same platform, so pretty much what Intel has been doing for a while.
Next gen will use a different socket.
Going back to solder.
This I didn't expect. I actually believe it's a reaction to users' expectations more than an actual engineering necessity.
Of course most clients don't care what's under the heatsink, but at least all the internet/review TIM criticism will be gone.
All in less than a year, interesting how quickly things have changed.
So now what? You'll become an Intel fanboy now or are we doomed to years of "thank you AMD"? :)
Posted on Reply
#54
yotano211
nemesis.ie said:
Actually autistic people are often very logical, so did you mean to say "not-autistic"?
My son is autistic, little man can do college level math, but sucks at English. Still proud of him.
Posted on Reply
#55
cucker tarlson
yotano211 said:
My son is autistic, little man can do college level math, but sucks at English. Still proud of him.
They can generally do exceptionally well in one field, in others they are usually not that devoted,and frankly you're better off NOT pushing them too hard in what they do not want to do. They are very goal oriented people. Dealt with some kids like that,they're usually so kind and sincere in everything they do.

btw I wonder what's next, 10900K ? Y2K ?

@RejZoR please school us on this one ;):laugh:
Posted on Reply
#56
chodaboy19
Will intel ever increase the number of PCIe lanes for non-HEDT? I'd like to get Optane PCIe SSD but don't really need to invest in an X-series CPU/platform...
Posted on Reply
#57
yotano211
cucker tarlson said:
They can generally do exceptionally well in one field, in others they are usually not that devoted,and frankly you're better off NOT pushing them too hard in what they do not want to do. They are very goal oriented people. Dealt with some kids like that,they're usually so kind and sincere in everything they do.

btw I wonder what's next, 10900K ? Y2K ?

@RejZoR please school us on this one ;):laugh:
y2k was 18 years ago
Posted on Reply
#58
cucker tarlson
chodaboy19 said:
Will intel ever increase the number of PCIe lanes for non-HEDT? I'd like to get Optane PCIe SSD but don't really need to invest in an X-series CPU/platform...
And what is stopping you exactly ? You can do full speed optane raid on z370.
Posted on Reply
#59
efikkan
RejZoR said:
6 and 8 core CPU's without HT in 2018. Clock or not, this is incredibly poor…
RejZoR said:
And double the threads are way superior to whatever you clock half as much threads for loads that require them. Given how basically everything is multithreaded without limitations these days, having just physical cores and no SMT is just ridiculous. Unless it's a budget CPU where cuts need to be made and is understandable. It sure as hell isn't on a product that will cost 400€ for sure.
SMT(like HT) made sense back when we had one Pentium 4 core, but as CPUs have gotten much more efficient, more cores, and much more complex, the costs of implementing SMT are getting too high. SMT is a trade-off, it utilizes idle cycles for other threads, at the cost of lowering single thread performance and increasing latency. This is why it may hurt e.g. gaming, but it actually sucks for any synchronous task. Modern desktop CPUs are highly reliant on an efficient prefetcher and cache, SMT hurts both. SMT adds scheduling overhead for the OS kernel, and with increasing core count the benefits are diminishing, especially with uneven workloads. You can of course disable SMT in software, but it's still wasted die space. The only place where SMT makes sense any more is in specialized server workloads, where only total efficiency matters, and there is a massive amount of threads (e.g. enterprise Java code or web servers). I haven't heard any indication of Intel dropping SMT, but I think they should. They can start with the consumer lineup. Another issue with SMT is that most IPC improvements will make SMT less efficient, since SMT relies on utilizing idle cycles.
Posted on Reply
#60
cucker tarlson
I know for sure know 2018 is the year from which one on I'll no longer buy a HT CPU ever.

Thank you AMD :love: Now ditch SMT and release a CPU with 8 core CCX on 7nm.
Posted on Reply
#61
HTC
Everyone knows when Intel's "goo pasted" chips get delidded, they get better thermals and therefore are able to have better stock / turbo speeds and be OCed higher (not always, but most of the time). That is the REAL reason for Intel to go with soldering.

AMD's "threat" is closing in and Intel, in an effort to remain in the lead for a longer period while they come up with a proper reply to Ryzen's performance AND scalability, are pulling "all the stops". This move will guarantee their continued lead in single threaded performance (which is already significant) and hope to stay that way by the time Zen 2 products are launched.

For those that wish to discuss the recent security flaws, there's a topic for that already in the news section: suggest we use it when discussing that specific issue.
Posted on Reply
#62
phanbuey
cucker tarlson said:
I know for sure know 2018 is the year from which one on I'll no longer buy a HT CPU ever.

Thank you AMD :love: Now ditch SMT and release a CPU with 8 core CCX on 7nm.
AMD's SMT is actually good though.
Posted on Reply
#63
Vya Domus
phanbuey said:
AMD's SMT is actually good though.
Nah, it has more or less the same deficiencies with scalability but at the very least they don't charge extra for it.
Posted on Reply
#64
Upgrayedd
TheinsanegamerN said:
So you want an HDET chip without the HDET price, when the disabled version will not draw any extra power, as has been established time and time again?

Protip: stop being poor and begging for such a chip. If you dont want to pay intel prices, buy AMD. Intel isnt going to make an extra special die cheap for you guys that dont want to pay for it, stop asking for it.
It used to be affordable. Its not anymore. And its HEDT. The percentage of people of people that will NEVER use the iGPU on their unlocked 5GHz CPU far outweighs the percentage who use it even just once I imagine. I'll even take another 5775C with that L4 cache.

You got some controlling vibes coming off you. Swear you tried to tell me what to do.

Why just roll over and not ask for it? Why not just roll over on non-soldered chips too? You can have goo paste ones since you don't really mind and are drowning in so much cash that you just buy em by the lot from silicon lottery who will fix them for you. Since you're so rich and all, bet my johnson is bigger, after you lose maybe I could finally afford HEDT.
Posted on Reply
#65
Hood
Upgrayedd said:
Since you're so rich and all, bet my johnson is bigger, after you lose maybe I could finally afford HEDT.
So now we need a thread for member's dick pics, so anyone losing an argument in a thread can just link to the TPU Dick Database. Or better yet, everyone could make their avatar from their dick pic, with a ruler/yardstick for scale. Epeen gone wild! If you have a small dick - no posting allowed, come back after you work out with your peter pump. But what about our female members? Tit pics?
(disclaimer - this post was sarcastic, I do not advocate that TPU be turned into a showcase for photo-shopped genitalia!)
Posted on Reply
#66
cucker tarlson
phanbuey said:
AMD's SMT is actually good though.
https://www.purepc.pl/procesory/test_procesorow_amd_ryzen_5_2600x_vs_intel_core_i5_8600k?page=0,46

it's hit or miss, according to the results I posted

+14%
+0,8%
-1,9%

+2,3%
-1,5%
-1,7%

+6,5%
+23%
+16%
+11%

it had helped to gain 10-15% in 3 cases out of 10, in just as many performance was lost,though only marginally. You could see over 20% in only one, 3 out of 10 had either minimal or very small gains.

It does on AMD what it does on Intel,meaning it only helps in very specific conditions. Ideally you'd rather have more physical cores than SMT. It may seem SMT is better but that's just cause it helps Ryzen more as this chip is more multi thread oriented with lower single core performance and higher latency. Intel has better single core performance,therefore you achieve slightly lower CPU load on physical cores,which in turn translates to HT making less difference on Intel cpus than SMT does to AMD ones.
Posted on Reply
#67
notb
Vya Domus said:
Nah, it has more or less the same deficiencies with scalability but at the very least they don't charge extra for it.
How exactly is Intel charging extra for HT?
Do you even slightly understand how CPUs are produced and positioned? :)
HTC said:
Everyone knows when Intel's "goo pasted" chips get delidded, they get better thermals and therefore are able to have better stock / turbo speeds and be OCed higher (not always, but most of the time). That is the REAL reason for Intel to go with soldering.
Every CPU will transfer heat faster after delidding. You mean replacing the conductor, right?
How can they have better stock speeds after that? Stock speeds are fixed.

We've seen overclocked 7700K pulling 150W and working perfectly well with the factory TIM.
There's enough headroom for 6 and 8-core CPUs (without OC, obviously).
Ryzen's performance AND scalability, are pulling "all the stops"
When will people learn that the "scalability" is simply a cost lowering idea. It hurts the performance.
We all should be hoping that Intel can keep designing CPUs how they used to (and selling it at higher prices).
Posted on Reply
#68
nemesis.ie
You lost me at the "and selling it for higher prices".
Posted on Reply
#69
HTC
notb said:
Every CPU will transfer heat faster after delidding. You mean replacing the conductor, right?
How can they have better stock speeds after that? Stock speeds are fixed.

We've seen overclocked 7700K pulling 150W and working perfectly well with the factory TIM.
There's enough headroom for 6 and 8-core CPUs (without OC, obviously).
I worded it wrong: what i mean was with the big temps reduction they get from the soldering VS TIM, they can increase the stock voltages and this will lower the temps and also increase the OCing headroom.

They could have a version of the 8600K soldered and with a slight base clock jump and make a "new chip" out of it ...

... Oh wait ...

That's what the I5-9600K is ...
Posted on Reply
#70
Dimi
Xzibit said:
[IMG]https://www.techpowerup.com/forums/proxy.php?image=http%3A%2F%2Fwww.techpowerup.com%2Fimg%2F1TySoeF6d4ZlTxO2_thm.jpg&hash=519da3e491e936bb4c76180a2ac183b7[/IMG]

read the fine print asterisk



Wait and see until one of the OCers delids one
There is an asterisk at the OPTANE listing ....
Posted on Reply
#71
phill
I'm not seeing a real reason to be upgrading my 5960X any time soon... I have no money but that's aside from the point.. No Intel CPU's for me for the foreseeable future...
Posted on Reply
#72
Midland Dog
RejZoR said:
6 and 8 core CPU's without HT in 2018. Clock or not, this is incredibly poor. I have a 6 core CPU with 12 threads. Sure, it doesn't clock to 4.x out of the box, but come on, with manual overclocking, I have it at 4.5GHz and 12 freaking threads. TWELVE! The IPC gain is nowhere near as significant as some say it is on newer CPU's, it's basically just gains from higher clocks. 5820K is 4 years old. One would expect things from HEDT to finally trickle down to normies ranges by now. I guess Intel just can't get out of their skin...

9600K should be 6c/12t
9700K should be 8c/16t
9900K should be 10c/20t

With HEDT far beyond that starting at 16c/32t. Sigh.
i5 = 4/8
i7 = 6/12
i9 = 8/16 would have been better
Posted on Reply
#73
RejZoR
And i3 with just 4 physical cores, no HT. No more 2c/4t nonsense. Ryzen 3 cut that crap from the beginning.
Posted on Reply
#74
Hood
i2 = 2c2t
i3 = 2c4t
i4 = 4c4t
i5 = 4c8t
i6 = 6c6t
i7 = 6c12t
i8 = 8c8t
i9 = 8c16t
i10 = 10c10t
...and so on in the same progression, right through the HEDT lineup
All unlocked, all soldered
The even-numbered CPUs all designate actual core/thread count, the odd-numbered CPUs are all hyper-threaded - once you know these 2 simple facts, you can figure out where any CPU belongs in the heirarchy.
It's clean, decisive, ends any confusion, and all possible permutations are covered - meaning Intel will never go for it!
Posted on Reply
#75
StrayKAT
Hood said:

It's clean, decisive, ends any confusion, and all possible permutations are covered - meaning Intel will never go for it!
They used to be extremely simple. What the hell happened?
Posted on Reply
Add your own comment