Intel Core i9-10980XE Cascade Lake-X CPU Can Hit Lofty 5.1GHz Overclock Across All 18 Cores

[BiG StroOnZ]

2 hours ago, Jurrunio said:

lifespan's concerned? Or just finishing a benchmark?

Doesn't really matter, the website is heavily active. How the algorithms work to finalize the numbers, probably unknown for obvious reasons. But despite this, the numbers are accurate based on plenty of confirmatory data and large sample sizes. If you have a better source, please let me know.

 

2 hours ago, Fasauceome said:

I didn't claim that there's no difference, just that the difference itself is contained within the range of CPUs that the product stack is in, and I showed that a 10 core is gonna behave really similarly to an 18 core despite almost half the cores.. Is the 9600K a better overclocker than the 9900K?

 

also @Jurrunio is right, HWBot won't give accurate "average" OC data

The average OC data is plenty accurate and used as a source all over the internet quite frequently. Specifically because of its accuracy based on the sample size and variety of cooling methods. As I said above, if you have a better, more accurate source; please do tell.

 

Right, the 9900k is the better clocker than the 9600k. However, again, I didn't claim all the time less cores means better clocks. This is why I was careful with the words chosen when I said "in theory." Meaning possible, but not always. 

[Deli]

No AIO can cool this. Intel says standard watercooling probably means four 360mm rad link together, and with 12x 5000rpm Delta fans at full speed.

 

Good for heating up the house in the winter.

[BiG StroOnZ]

15 minutes ago, 2SidedPolygon said:

I wouldn't get too excited. Firstly, a comparison against last-gen silicon is useless. The 3900X performs better in most of the Geekbench scores I found (https://browser.geekbench.com/v4/cpu/search?utf8=✓&q=ryzen+9+3900x, I'd guesstimate around 15%-20% based on these numbers), and with significantly higher power efficiency, more than likely. Regarding those overclocks, I wouldn't be inclined to trust Intel's own statistic, and I certainly wouldn't be inclined to claim that if the overclocks are X% faster, then the performance is X% better.

 

And let's not even get started on the price tag. This isn't a mainstream processor, this is an Extreme chip. I'd expect some eye-watering numbers, even in light of the recent 7nm shortages.

 

So don't get ahead of yourself, 10th gen isn't shaping up to be Intel's big break.

Yeah, I went heavily into comparable results with the 3900x in the OP. Then I also compared those numbers with possible overclocks with the Cascade Lake-X parts (as the bench results I found with the 3900x weren't @ stock while the Cascade Lake-X results were at stock).

 

I wouldn't trust their own statistic blindly either. But I think a variance of 300MHz still makes the part promising.

 

I believe the source lists the prices - for Intel, definitely good. Not comparable to AMD in that aspect, but this is Intel we're talking about.

 

As far the last comment, not sure what this means. Intel doesn't need a big break. They're doing fine.

[Rune]

I want a 12 core that can do 5ghz with at least coffee lake-ish ipc. That'll be my next upgrade.

[Vishera]

Just now, Jurrunio said:

that's the problem, results in HWbot aren't meant for 24/7 use, in other words not what a general user should come to expect. Minus all the silicon binning going on.

I agree,when you take data from the largest overclocking community's website,it will have a lot of exotic cooling in the results,making the claims of "Average Overclock" invalid.

[2SidedPolygon]

1 minute ago, BiG StroOnZ said:

Yeah, I went heavily into comparable results with the 3900x in the OP. Then I also compared those numbers with possible overclocks with the Cascade Lake-X parts (as the bench results I found with the 3900x weren't @ stock while the Cascade Lake-X results were at stock).

 

I wouldn't trust their own statistic blindly either. But I think a variance of 300MHz still makes the part promising.

 

I believe the source lists the prices, for Intel definitely good. Not comparable to AMD in that aspect, but this is Intel we're talking about.

 

As far the last comment, not sure what this means. Intel doesn't need a big break. They're doing fine.

Heavily? You linked two benchmarks that didn't give an entirely clear picture of performance, whereas I linked a metric ton of them, running at stock speeds.

 

The chances of it being enough to push it over the edge performance-wise are mediocre, and the chances of it making it a remotely decent value proposition are slim to none.

 

No, this is the market we're talking about. If the product is irrelevant because of its price, well, its irrelevant because of its price. Approaching double that of the AMD part is madness.

 

They're really not. How often do you see Intel getting recommended for anything other than no-compromises gaming builds, or maybe some use cases with Adobe products? The only place I can think of them as being very competitive in right now is notebooks.

[BiG StroOnZ]

5 minutes ago, 2SidedPolygon said:

*snip*

Yeah, I see where this is going... but unfortunately, I'm not going down this road with you, as it will lead to nothing fruitful.

 

Everything you said is right, enjoy your night. 

[pas008]

17 minutes ago, Deli said:

No AIO can cool this. Intel says standard watercooling probably means four 360mm rad link together, and with 12x 5000rpm Delta fans at full speed.

 

Good for heating up the house in the winter.

Nice standard setup

[2SidedPolygon]

3 minutes ago, BiG StroOnZ said:

Yeah, I see where this is going... but unfortunately, I'm not going down this road with you, as it will lead to nothing fruitful.

 

Everything you said is right, enjoy your night. 

Eh, guess I'll take it

[Rolling2405]

1 hour ago, Deli said:

No AIO can cool this. Intel says standard watercooling probably means four 360mm rad link together, and with 12x 5000rpm Delta fans at full speed.

 

Good for heating up the house in the winter.

What type of clocks do you think we could get on a more "normal" CPU such as Intel I9 10900x? 

[porina]

 

6 hours ago, Arika S said:

Also what is deemed "standard liquid cooling"? 

In overclocking circles, that just means not cooled below ambient. It could still well need a decent custom loop, not a cheap AIO.

 

6 hours ago, BiG StroOnZ said:

Average overclock

Don't use hwbot for that. The CPU clock one can be anything from stock to no stability suicide runs. Averaging that is pretty meaningless. If you really want to use hwbot data, look at a popular benchmark like Cinebench R15 and limit it by cooling class for a better idea.

 

3 hours ago, Thomas001 said:

What type of clocks do you think we could get on a more "normal" CPU such as Intel I9 10900x? 

Based on my 1st gen 7800X that was hitting 4.8 comfortably, 4.9 right on the limit of my cooling (3x120mm rad), and I did manage 5.0 with a chiller. I don't think I had a good sample either as others were reporting hitting 5.0. As long as you can cool it, comparable clocks should be possible.

 

I seem to have missed a quote here, someone said Cascade Lake-X is essentially the same as Skylake-X, and we don't know that. The fundamental architecture is, but they have updated the AVX-512 instruction support so it is different silicon. Skylake-X was made on same process as Kaby Lake (14+) and Intel have stopped reporting on process updates since then. It is quite possible they moved onto Coffee Lake equivalent or even beyond.

[RuffRuffmcgruff]

Wow, that's pretty cool.

[schwellmo92]

10 hours ago, Thomas001 said:

If this is true I may consider upgrading from my Intel 9900k to a Intel 10900x.

Buy a new CPU, motherboard and RAM sticks for an additional 2 cores?

[schwellmo92]

Isn’t Cascade Lake arch very similar to Skylake arch? So we should expect single-digit IPC gains and maybe a small frequency bump? I wouldn’t expect performance to be radically different from the previous HEDT lineup, even if Geekbench shows as such (Geekbench is well known to be a poor indicator of performance anyway).

[porina]

6 minutes ago, schwellmo92 said:

Isn’t Cascade Lake arch very similar to Skylake arch? So we should expect single-digit IPC gains and maybe a small frequency bump? I wouldn’t expect performance to be radically different from the previous HEDT lineup, even if Geekbench shows as such (Geekbench is well known to be a poor indicator of performance anyway).

I'd expect 0% IPC gains outside of the new VNNI instruction, but would be happy to be wrong on that note.

[MeatFeastMan]

Before anybody gets their hopes up, don't be fooled. Comparing it to a 2920x is good and all..but the 3900x beats the 10920x, and the 9920x is about on par:

https://browser.geekbench.com/v4/cpu/14707930

https://browser.geekbench.com/v4/cpu/14778435

 

Geekbench tells us absolutely NOTHING about multi-core performance.

 

The 10920x is exactly what everybody thinks it is. A 9920x with 300mhz more and maybe a few percent better IPC at best. There's nothing to see here other than a price drop. It's Skylake-X with a wig, some face paint and a hat.

[RejZoR]

Core i9-10980XE, codenamed "SCORCHED EARTH"

[Brainless906]

Yah and my fx9590 will do 5.0ghz on all cores too!

[Sauron]

15 hours ago, BiG StroOnZ said:

Intel’s Mark Walton has confirmed that the processor has plenty of overclocking headroom to hit 5.1GHz with liquid cooling

I don't know if I'd take this at face value straight from Intel's mouth... if all or even most chips could reach that they would be sold at a higher stock speed. Also what @2SidedPolygon said.

[melete]

7980XE overclocked well so this all makes sense. 

[DildorTheDecent]

>2920X

 

I forgot AMD made this chip.

[MageTank]

2 hours ago, Sauron said:

I don't know if I'd take this at face value straight from Intel's mouth... if all or even most chips could reach that they would be sold at a higher stock speed. Also what @2SidedPolygon said.

Yeah... Where have we seen this before? https://www.techspot.com/review/837-intel-core-i7-4790k-devils-canyon/page8.html

Quote

At Computex, Intel president Renee James said the chip could be pushed to 5GHz in air-cooled systems and the company's Francois Piednoel tweeted that he got 4.6GHz without a fan, so he'd have to offer some training to those who couldn't manage to hit 5GHz.

Unfortunately, I may have to take Piednoel up on his course in Devil's Canyon overclocking because our review sample wouldn't budge past 4.7GHz at 1.36v and it wasn't for a lack of trying. Intel's Australian PR company demanded to have the processor back within 7 days, so without a chance for future testing we weren't afraid to give the 4790K everything it could handle.

http://m.hardocp.com/news/2017/01/19/intel_core_i77700k_cpu_5ghz_overclock_chances

 

I cannot imagine 18 cores at 5ghz is going to be easy. Even if they managed to pull it off, it tells us absolutely nothing about stability or practicality under a realistic load. The 9980XE's got extremely hot, even despite being soldered when clocked higher than 4.5ghz. Direct die 7980XE's with liquid metal and a large custom loop was topping out around 4.7ghz on the high end, and stability was still questionable depending on the load. I would not take these words seriously until multiple sources can confirm it's possible.

[porina]

17 minutes ago, MageTank said:

I cannot imagine 18 cores at 5ghz is going to be easy. Even if they managed to pull it off, it tells us absolutely nothing about stability or practicality under a realistic load. The 9980XE's got extremely hot, even despite being soldered when clocked higher than 4.5ghz. Direct die 7980XE's with liquid metal and a large custom loop was topping out around 4.7ghz on the high end, and stability was still questionable depending on the load. I would not take these words seriously until multiple sources can confirm it's possible.

I didn't know of FP then, but on the few occasions I cross paths with him on twitter, doesn't leave me with a good impression. Talks a lot, no substance to claims.

 

Think I saw somewhere that Intel's claims on the 9900KS hitting 5 GHz on all cores relies on it being a lower intensity workload. Presuming there will be limits hit. AVX is probably right out. I'd assume something similar here. Maybe Cinebench R15, but even that isn't the lightest load even without AVX.

 

I don't know how the LCC and higher CC dies scaled, but my 7800X was an easy 4.8, on the limit at 4.9 (ambient water), and managed 5.0 with chilled water. Recently got a 7920X to try under the chiller but not got there yet. On air taken it to 4.5 GHz running hwbot x265 so I'm hopeful of getting much more than that with better cooling. Solder aside, was there any difference with the 9000X series?

[MageTank]

3 minutes ago, porina said:

I didn't know of FP then, but on the few occasions I cross paths with him on twitter, doesn't leave me with a good impression. Talks a lot, no substance to claims.

 

Think I saw somewhere that Intel's claims on the 9900KS hitting 5 GHz on all cores relies on it being a lower intensity workload. Presuming there will be limits hit. AVX is probably right out. I'd assume something similar here. Maybe Cinebench R15, but even that isn't the lightest load even without AVX.

 

I don't know how the LCC and higher CC dies scaled, but my 7800X was an easy 4.8, on the limit at 4.9 (ambient water), and managed 5.0 with chilled water. Recently got a 7920X to try under the chiller but not got there yet. On air taken it to 4.5 GHz running hwbot x265 so I'm hopeful of getting much more than that with better cooling. Solder aside, was there any difference with the 9000X series?

Not that I've seen. The samples we've tested showed no difference, but our sample size was small (2 7980XE's and 2 9980XE's). I imagine the 9000 series might have a better binning, but that's difficult to test without a larger sample size. From what I could tell, they were the exact same chips but soldered. Cascade Lake looks to be a bit more interesting mostly because the cache hierarchy of the smaller HCC chips appears to be different, leading me to believe they changed something under the hood with physical location of the IMC on the die. I am hopeful that this circumvents the write bandwidth penalty that these chips suffered from on the previous 7000 and 8000 series of HEDT chips. AMD made a similar change with their Zen 2 EPYC processors, so it's not outside of the realm of possibilities.

 

I just don't see the 5ghz advertisement holding true, not at ambient temperatures under realistic loads.

 

 

[mr moose]

4 minutes ago, MageTank said:

I just don't see the 5ghz advertisement holding true, not at ambient temperatures under realistic loads.

 

 

Gotta read that fine print,  I am sure he insinuated under his breath when no one was looking that there were important conditions.