AMD breaks the world record for 16-core CPUs in Cinebench with a 7950X on LN2

[takaozo]

Lol I just seen the AM5 board prices, good luck to early adopters -(^^)-.

 

[kiwigraeme]

AMD breaks the world record for 16-core CPUs in Cinebench with a 7950X on LN2 - That's a world record? Honestly? It sounds like TS is just taking AMD advertising and printing it as an article.

Intel's Core i9-13900K tops the single thread performance charts

The recently published single-thread benchmark score of 4,883 leaves little doubt regarding the new Intel flagship's capabilities. The 24-core i9 processor's eight performance and 16 efficiency cores...

www.techspot.com

Posted nearly at same time - TS knows these chips are about to hit so interest is high .

As for fuzzies - sure probably many of us have more love for AMD - given Intels indifference for so long to it's customers - saying that 95% ( grab from nowhere ) would happily buy either Intel or AMD if it's best for us - who bought AMD before just before Zen came out ?- from memory if you want an APU system for business or building a simple system for your grandma - ie a CPU/GPU all in one .
For myself Intel will need to be 5% plus better - maybe more - as have option of replacing the CPU
What stood out to me was when I saw scores elsewhere for new AMD chips
Was 7600x could beat a 3950x in some tests

ie maybe same for Intels new ones - someone now spending for a mid range chip can beat flagships on last models for some applicatons ,
Maybe same will happen chip GPUs ie a RTX 4060 will equal a RTX 3080 ( and same for RNDA 3 vs 2 )
if so this will be the best thing going forward

 

[Avro Arrow]

Well, this is a situation that I'll never encounter but it does speak to the extreme potential of the new architecture. Sure, it's useless for us because who's going to use LN2 to cool their CPU on a regular basis (I might be surprised but I seriously doubt it)?

What these tests do is offer proof of capability. What I see from this otherwise useless feat is that the R9-7950X at ~6.47GHz defeats the i9-12900KS operating at 7GHz by 8,000 points. Now, since we don't know what these points mean or what their values are, we have to get fancy.

Since the R9-7950X scored 48,235 it means that the i9-12900KS scored about 40,235. Since the R9-7950X ran at 6.47GHz and the i9-12900KS ran at 7GHz and the points are a common denominator, we should just need simple division to get and IPC score that's directly comparable:
CBR23IPC7950X = 48235 ÷ 6.47
∴ CBR23IPC7950X = 7455.18

CBR23IPC12900KS = (48235-8000) ÷ 7.00
∴ CBR23IPC12900KS = 5747.85

To get the performance Δ between them, we just divide again:
Δ = CBR23IPC7950X ÷ CBR23IPC12900KS
Δ = 7455.18 ÷ 5747.85
∴Δ = 1.297

So, in Cinebench R23, Zen4's IPC is superior to Alder Lake's IPC by ~30%. I'd say that's a devastating curbstomp. Sure, these are both OC numbers but that doesn't matter because they're both OC the same way with LN2 and this is the max that they're capable of. Once the difference in clock speed was accounted for, we got a great apples-to-apples comparison.

Now, I don't know what the power usage was for the 12900KS, but Ryzen has repeatedly shown itself to be much more power-efficient than Intel's Whatever Lakes. In this situation, that is almost assuredly the case because the R9 was pulling 6.47GHz at 1.4V while the i9 was pulling 7GHz at 1.5V. The stock TDP for the R9 is 101W while the stock TDP for the i9 is 125W despite the i9 being made of 8 performance cores and 8 "efficient" cores.

At the R9 / i9 level, performance of this kind and power efficiency are king. Gaming results might matter to a few enthusiasts who are crazy enough to buy an R9 or i9 for gaming but it's not enough to make any significant difference in the sales of these chips.

The way it looks right now, Alder Lake is in big trouble from Zen4 because this is a serious win for AMD.

 

[Avro Arrow]

LN2 and world records it's just pure marketing BS and takes advantage of human psychology on client side.
"my truck it's much bigger than yours"......"have you seen that guy, his truck it's the biggest"
Same tactics applies to Crapple users.

Well, considering what I was able to glean from the comparison to the Intel i9-12900KS, that Zen4's IPC in Cinebench R23 is 30% faster than Alder Lake, I'd say that this article was very revealing.

AMD breaks the world record for 16-core CPUs in Cinebench with a 7950X on LN2 - That's a world record? Honestly? It sounds like TS is just taking AMD advertising and printing it as an article.

Are you new? Overclocking world records are always covered by the tech press. I remember when the Phenom II X4 940 broke the world overclocking record. Then the FX-8350 did the same thing. It was meaningless but it was a fun fact on a slow press day. At least with this article, information could be gleaned from the comparison between the R9-7950X and the i9-12900KS.

AMD must build the hype for tomorow, what do you expect?

The "hype" shows a 30% actual IPC advantage in Cinebench R23 over Intel's Alder Lake. That's actually valuable information, not just hype.

Lol I just seen the AM5 board prices, good luck to early adopters -(^^)-.

Where did you see them? I'm not going to be an early adopter (I'll be quite late actually) but I'm still curious. All that I've been able to find are the top-level X670E motherboards and while they are expensive, they're no more expensive than their AM4 counterparts were. In the beginning, it's always the boards like the ROG, Taichi and Godlike that are out first. AnandTech reports that MSi's X670 boards will start at $290USD but there's no mention of the B650 or A620 models yet. As more and more board-makers release more and more models, the prices will fall like rain just like they always have.

 

[mbk34]

Are you new? Overclocking world records are always covered by the tech press.

I understand testing for CPUs for performance. I understand having records for single core performance and I understand having records for all cores on a single CPU. What I don't understand is why there'd be a record for a specific number of cores (16) - would you be influenced by a world record for 7 cores or 3 cores? The only think that matters to consumers is single core performance for certain types of games or all core performance. It just makes no sense to me to have a record for a specific number of cores. Perhaps you can explain?

 

[Avro Arrow]

I understand testing for CPUs for performance. I understand having records for single core performance and I understand having records for all cores on a single CPU. What I don't understand is why there'd be a record for a specific number of cores (16) - would you be influenced by a world record for 7 cores or 3 cores? The only think that matters to consumers is single core performance for certain types of games or all core performance. It just makes no sense to me to have a record for a specific number of cores. Perhaps you can explain?

Well yeah, they're getting a bit ridiculous with the number of cores. I really don't know why they're doing that. If I had to guess, I'd say that they got bored and added categories like this because it's a lot harder to get 16 cores to whatever speed than it is only four. That's just a guess though and your guess is as good as mine.

You make a really good point, one that I hadn't considered. I really don't know the answer.

 

[Endymio]

What I don't understand is why there'd be a record for a specific number of cores (16) - would you be influenced by a world record for 7 cores or 3 cores?... It just makes no sense to me to have a record for a specific number of cores.

The simple answer is that most people who purchase high core-count CPUs often do so for one or two specific software applications. Each of these scales differently by core count. For some, doubling the core count from 8 to 16 may double performance; for others it may have little impact (in a very few cases, it may even decrease performance).

In brief: if an 8-, 10- or even a 32-core CPU "has the same score" as a 16-core one, that synthetic benchmark doesn't mean it will run a specific package just as fast.

 

[kiwigraeme]

I understand testing for CPUs for performance. I understand having records for single core performance and I understand having records for all cores on a single CPU. What I don't understand is why there'd be a record for a specific number of cores (16) - would you be influenced by a world record for 7 cores or 3 cores? The only think that matters to consumers is single core performance for certain types of games or all core performance. It just makes no sense to me to have a record for a specific number of cores. Perhaps you can explain?

I think your original point is the only valid pont - whether this is news worthy on TS - and given the respnse = it seems it is - saying that it was over the weekend - so less competition.

I don't follow these world records - but as as far as I know they - I could hunt down some 30 year old PCs about to be scrapped - and test the hell out of the CPU with newest methods hoping for a WR . I.e there are probably 1000s of records in the OC world - but like the olympics there is the gold medal for 100m people remember and then some standing single shot target pistol medal.
Plus they probably have rules - like if you disable 4 cores on a 8 core CPU - is it a 4 core record ? .

Every once and awhile we see a story somewhere - someone made a pentinum scream a little bit more

When you get audiophiles crazy about "conditioning" their power - I wonder for OC how crucial pure clean voltage is with no "noise".

I take the audio stuff with a sack of salt - if a $300 thingy is going to make the amazng diff they claim - was like removing a curtain he cried, from muddy to crystal clarity - then those expensive amps would aready have it .