Tuesday, February 16th 2021

Intel Core i9-11900T "Rocket Lake" Processor Allegedly Catches Up with Zen 3 in Single-Threaded Performance

When AMD announced its Ryzen 5000 series of processors based on the new Zen 3 architecture, the performance of these processors was the best on the market. Even in our own testing, we have found that AMD's Zen 3 core is the highest performing core on the market, even beating Intel's latest and greatest, the 10th generation of Core processors. However, Intel has been doing some silent work and the company has developed a new core to be used in the 11th generation "Rocket Lake" platform. Codenamed Cypress Cove, the design is representing a backport of the 10 nm Sunny Cove design, supposed to bring around 19% IPC improvement across the board.

If you were wondering if that was enough to catch up with AMD's Zen 3 IPC performance, look no further because we have Geekbench 5 performance results of Intel's 35 Watt Core i9-11900T processor. Having a base frequency of only 1.51 GHz, the CPU is capable of boosting one or two cores to the very high speed of 4.9 GHz, giving us a good example of the single-threaded performance we can expect from this CPU. In GB5 tests, the Core i9-11900T has managed to score 1717 points in the single-threaded test and 8349 points in multi-threaded results. Comparing that to something like AMD Ryzen 5800X, which scores 1674 points in single-threaded results, Rocket Lake's Cypress Cove core has managed to be 2.5% faster than Zen 3. However, in multi-threaded results, the AMD chip is unmatched as the low TDP of the Intel processor is stopping it from reaching full performance.
Source: Geekbench 5
Add your own comment

90 Comments on Intel Core i9-11900T "Rocket Lake" Processor Allegedly Catches Up with Zen 3 in Single-Threaded Performance

#1
Melvis
Nope!


And geekbench? come on honestly! .......
Posted on Reply
#2
DeathtoGnomes
Melvis
Nope!


And geekbench? come on honestly! .......
not everyone likes geekbench, what would you use to compare that is so widely popular and well known??
Posted on Reply
#3
Nephilim666
TPU I am very disappointed to see shillbench presented as evidence of performance. @W1zzard can you please vet any posts on benchmarking?
Posted on Reply
#4
Cobain
Still amazed by the amount of people that instantly jumps on any Intel related news article, always judging their products in a bad way. Is like some people can't accept the idea of their products being a good option.
Posted on Reply
#5
AleksandarK
Please note that this is only a report based on a Geekbench submission. This is not an official review and just one benchmark present. We are not judging the benchmark's credibility and we are just using it as a reference for other processors here. :)
Posted on Reply
#6
qubit
Overclocked quantum bit
Look at that, the minute there's some competition, Intel pulls its finger out and gives us real performance improvements. Who'da thunk it?
Posted on Reply
#7
londiste
11900T boost clock is reportedly 4.9GHz vs 5800X's 4.7GHz so Intel has a couple % frequency advantage.
Either way, single-core performance is within a percent or two at the same frequency. Competition is only good from our consumer perspective.
Posted on Reply
#8
XL-R8R
Some quick maths?


4.9GHz vs 4.7GHz max single core boost = 4% difference in clock speeds (being in Intels favour)


1717 vs 1674 is the max single thread score of both CPU's mentioned (being in Intels favour again)... but when we take 1717 (the score of the mentioned 11900T) and reduce it by 4% to take into account clock disparity between CPU's, we actually get a score of 1648 for the Intel part... less than the 5800X's score, at the same clock speed.

Or, in "points-per-ghz":

1717/4.9=350.40
1674/4.7=356.17


Maybe I'm mistaken, but that doesnt equate to a 2.5% lead.





Edit to ask @W1zzard for employment due to my amazing mathematical skills. Any chance? :roll:
Posted on Reply
#9
qubit
Overclocked quantum bit
@XL-R8R Unfortunately, calculations like that aren't that accurate in predicting performance, because they don't take into account constants such as cache and memory latency. This is especially true if the benchmark results are within a few percentage points of each other. If the performance difference was around 20% say, then it would be more accurate. Best way is to set the clock speeds the same and then benchmark them.
Posted on Reply
#10
_Flare
Hence the only 512kb size of the L2 Cache and not over 1 MB like in Willow Cove, i strongly suggest Cypress Cove is a Sunny Cove Backport (IceLake)
Posted on Reply
#11
Cobain
I also dont know why TPU is not writing about this subject that I already mentioned a week ago and got attacked by doing so on these forums (being called fanboy etc). It is a major thing right now on tech websites/forums and should call your attention. A lot of whea errors and faulty CPUs due to the high demand, they failing on QC.



www.overclock.net/threads/replaced-3950x-with-5950x-whea-and-reboots.1774627/page-57
community.amd.com/t5/processors/ryzen-5900x-system-constantly-crashing-restarting-whea-logger-id/td-p/423321
Posted on Reply
#12
qubit
Overclocked quantum bit
@Cobain If that tweet is accurate, then this is damning. I think AMD would go out of business if they were quite that bad, so maybe PowerGPU have been unlucky, or have an axe to grind with AMD. Or Intel paid them if we want to get into conspiracy theory territory. I don't think they did.
Posted on Reply
#13
Caring1
Apples and oranges, laptop V's Desktop CPU, and Intel as usual is focusing on single thread performance to win the speed crown, but neglects to say their current mobile CPU's are inferior in every other aspect.
Posted on Reply
#14
Melvis
DeathtoGnomes
not everyone likes geekbench, what would you use to compare that is so widely popular and well known??
Exactly as its basically a trash benchmark, use something thats consistent and there are plenty of options like what is shown in the video.
Cobain
Still amazed by the amount of people that instantly jumps on any Intel related news article, always judging their products in a bad way. Is like some people can't accept the idea of their products being a good option.
Still amazed by the amount of people that instantly jumps on any AMD related news article, always judging their products in a bad way. Is like some people can't accept the idea of their products being a good option.
Posted on Reply
#15
Cobain
qubit
@Cobain If that tweet is accurate, then this is damning. I think AMD would go out of business if they were quite that bad, so maybe PowerGPU have been unlucky, or have an axe to grind with AMD. Or Intel paid them if we want to get into conspiracy theory territory. I don't think they did.
AMD actually contacted powergpu and are trying to solve the issue. We Will get news soon. Powergpu tweeted it today. They are a big company.

Plus AMD forums are full of complains
Posted on Reply
#16
Caring1
Cobain
AMD actually contacted powergpu and are trying to solve the issue. We Will get news soon. Powergpu tweeted it today. They are a big company.

Plus AMD forums are full of complains
Pretty sure they are confusing RMA and DOA.
Consumers are returning damaged or alleged faulty CPU's, but the issue isn't the Processor, it's the Motherboard Manufacturers stuffing up voltage regulation.
Anyway, enough off topic from me.
Posted on Reply
#17
londiste
Caring1
Apples and oranges, laptop V's Desktop CPU, and Intel as usual is focusing on single thread performance to win the speed crown, but neglects to say their current mobile CPU's are inferior in every other aspect.
Intel's mobile lineup is on a newer architecture and on a manufacturing process much closer to TSMC's 7nm with 8-core CPUs rumored to be incoming.
www.techpowerup.com/278501/intel-rumored-to-launch-three-8-core-11th-generation-tiger-lake-h-cpus

Not sure what your comment has to do with 11900T and 5800X comparison though.
Posted on Reply
#18
1d10t
Wow, there's T now? I bet highest end would be 11900FTKS RGB Limited Edition :D
Posted on Reply
#19
qubit
Overclocked quantum bit
Cobain
AMD actually contacted powergpu and are trying to solve the issue. We Will get news soon. Powergpu tweeted it today. They are a big company.

Plus AMD forums are full of complains
Thanks for the clarification, that sounds serious indeed.
Posted on Reply
#20
Selaya
londiste
Intel's mobile lineup is on a newer architecture and on a manufacturing process much closer to TSMC's 7nm
[ ... ]
Unfortunately, node density will do nothing about power consumpion/heat generation (quite on the contrary, actually) ...
Posted on Reply
#21
ARF
Cobain
Still amazed by the amount of people that instantly jumps on any Intel related news article, always judging their products in a bad way. Is like some people can't accept the idea of their products being a good option.
It is fair and square.
You probably know that Intel is still on the 5-6-year old 14nm process, and its processors are slower, have less multi-threading, are more power hungry and are not reliable because of security vulnerabilities.
So, given all these inferiorities, it is a tough task to remain positive towards Intel.

Hell, even Intel's CEO lost his position recently, what more can or shall we add?
Posted on Reply
#22
Frick
Fishfaced Nincompoop
Nephilim666
TPU I am very disappointed to see shillbench presented as evidence of performance. @W1zzard can you please vet any posts on benchmarking?
Sometimes I feel like the word "allegedly" these days means "factually".
Posted on Reply
#23
londiste
ARF
It is fair and square.
You probably know that Intel is still on the 5-6-year old 14nm process, and its processors are slower, have less multi-threading, are more power hungry and are not reliable because of security vulnerabilities.
So, given all these inferiorities, it is a tough task to remain positive towards Intel.
14nm is from 2014, so 6 years and counting.
This leak seems top indicate Rocket Lake is only slightly slower compared to Zen3 which is made up by slightly higher frequency.
Less multi-threading and power-hungry are still accurate enough.
Rocket Lake being Ice Lake/Sunny Cove derivative, it should have most of the security vulnerabilities fixed.
Posted on Reply
#24
Frick
Fishfaced Nincompoop
qubit
Thanks for the clarification, that sounds serious indeed.
They have since removed the tweet and various retailers (swedish link) report no more CPU failures than normal. I am much more inclined to believe there's something going on with some motherboards. 15% of CPUs being DOA would be unheard of.
Posted on Reply
#25
qubit
Overclocked quantum bit
Frick
They have since removed the tweet and various retailers (swedish link) report no more CPU failures than normal. I am much more inclined to believe there's something going on with some motherboards. 15% of CPUs being DOA would be unheard of.
Exactly, that would be business-killing failure rate. Dodgy mobos sound like quite a probable reason for this and the fact that the tweet has been taken down now is telling.
Posted on Reply
Add your own comment