• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Discussion Intel current and future Lakes & Rapids thread

Page 675 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
IntelUser2000 said:
Supports my expectation that a delayed product not only is late but misses expectations.

Maybe it's running very close to 2GHz.

But as @lobz points out Sandra is garbage. Don't know why they even exist.
Click to expand...

According To Yuuki_AnS Even the Late QS samples(Stepping E1/E2) have some issues.

"There is a UPI Link Error on some SPR-SP processor that Reach E1/E2 QS samples"
 
pakotlar said:
What is the significance of UPI, and how bad could this error be (in terms of time to fix)?
Click to expand...
Not much info is reported by Yuuki_Ans on this, but UPI is very important because it's Intel's Infinity Fabric(coherent interconnect) so it's affecting performance. This could be the delay issue. I mean. How on else could you explain a Multi core performance on par with ROME EPYC, Lower than Milan and utterly Annihilated by Genoa?
 
Last edited:
Is it actually a quad socket part? I thought they weren't on the drawing boards.
I mean each cpu is essentially a quad NUMA, even the monolithic parts had numa suggestions because they were so big. The UPI links between sixteen pieces of silicon must be an engineering nightmare.
 
tomatosummit said:
Is it actually a quad socket part? I thought they weren't on the drawing boards.
I mean each cpu is essentially a quad NUMA, even the monolithic parts had numa suggestions because they were so big. The UPI links between sixteen pieces of silicon must be an engineering nightmare.
Click to expand...
Yeah, it's a Quad Socket System, but the Xeon 8450H could be a 4S or a 8S System(4-8 CPUs System Processor. 4S for sure) All of Intel Scalable Xeons with H and HL suffix are either S4S or S8S Systems(for 4- 8 sockets or more than 8).

Intel Scalable processors with S prefix(S4S and S8S) along with Windows Data Center edition are capable of expanding to more than a few dozen of sockets on a Coherent Single System Image.(Third-party node controller technology to expand the number of sockets beyond the Four-Eight sockets). Granted on Production sample where the UPI Links are not buggy.


1657838187771.png
 
Last edited:
nicalandia said:
Those are QS Samples with full retail number. Do you think they would be locking those just before the release?(according to Yuuki_AnS the QS don't have SDSi locks)
Click to expand...
The way I understood SDSi is that by default parts of the CPU would be locked and have to be unlocked through SDSi which needs a working driver with trusted certification to verify a given installation is allowed to access originally locked features. If the QS is supposed to be equal to launch chips and SDSi is about unlocking features, then the QS absolutely would have locked features as well, otherwise it wouldn't be a QS.

Then the big question is: what exactly is being locked there, and may that result in the low performance we are seeing there? UPI links for example have been a premium feature in the past Intel used for segmentation, so may use SDSi now.
(Did Yuuki_AnS state everything of the QS is unlocked?)
 
Some of the SKU leakers have shown E3-stepping QS. Wonder if those fare any better?

Also do we have any benches with artificial thread limits to (hopefully) keep workloads local to one tile?
 
DrMrLordX said:
Do we have any benches with artificial thread limits to (hopefully) keep workloads local to one tile?
Click to expand...

Not yet, due to Mesh of Rings on Xeons even HCC Monolithic CPUs that large benefit from what is called Cluster-on-die (COD) AND Sub-NUMA cluster (SNC). With four tiles per CPU you can imagine how complex this is for the scheduler.
 
On the topic of Raptor Lake, it seems to behave exactly as expected. It's an Alder Lake refresh with Alder Lake refresh-like clocks, IPC, etc. The fun part will be when we repeat this discussion with Redwood Cove...
 
Last edited:
moinmoin said:
I was hoping for higher power efficiency.
Click to expand...

On the same process? With more cores? With higher clocks? Really?

EDIT: Meteor Lake is currently the chip to watch out for if you are hoping for improved power consumption, and even then, I suspect Intel will push the chips 'just because'.
 
Were there any notable leaks from this "Extreme player" leaker? I watched the source video, and he only showed a motherboard and gpu box. Other videos on his channel seem to be just harware news.
 
Let's stick to the good news so far:
  • min FPS has gone up ~30% on average @ 1080p
  • power increase is presented as "peak", may actually be correlated with the increase in min FPS
Not great, not terribad. Needs more leakage, maybe a seller could put them up for sale "by accident".
 
eek2121 said:
On the same process? With more cores? With higher clocks? Really?

EDIT: Meteor Lake is currently the chip to watch out for if you are hoping for improved power consumption, and even then, I suspect Intel will push the chips 'just because'.
Click to expand...

You mean Arrowlake? Cause no reputable sources are expecting high end MTL chips.
 
Exist50 said:
On the topic of Raptor Lake, it seems to behave exactly as expected. It's an Alder Lake refresh with Alder Lake refresh-like clocks, IPC, etc. The fun part will be when we repeat this discussion with Redwood Cove...
Click to expand...
Why? isn't redwood cove a new core unlike raptor which seems to be just GLC with increased L2?
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top