Intel Core i9-10900K benchmarks have leaked, and it's still slower than the Ryzen 9 3900X

erek

[H]F Junkie
Joined
Dec 19, 2005
Messages
10,785
Find no interest in Intel CPUs right now to be honest. Anyone preferring Intel right now?

"Intel's 10th-generation Comet Lake-S processors may narrow the massive gap that exists between AMD and Intel in the desktop world right now, but it may not last for long. Keep in mind that AMD CEO Lisa Su has said that Ryzen 4000 processors for desktop will be coming this year.


If the Intel Core i9-10900K only manages to come 7% short of beating the 3900X and only beating it in single-core by around 10%, that doesn't bode well for Intel whenever Team Red manages to launch its next desktop platform. Word on the street, according to an AdoredTV leak, is that the Zen 3-based Ryzen 4000 lineup is going to see a 15% boost in IPC performance. If that's paired with higher clockspeeds on AMD's next platform, Intel's single-core lead could vanish.


And now that we've seen AMD bring the Zen 2 improvements over to mobile, there's a lot of pressure on Intel to come up with something truly exciting. We said it in another piece touching on our brief testing of the AMD Ryzen 9 4900HS (more on that coming very soon), but we'd love to see Intel come up with its own Ryzen moment.


Intel Comet Lake-H has just arrived and Comet Lake-S is likely right around the corner, so we're incredibly interested to see whether or not it can shake up AMD's stranglehold on the processor world.

And if it does, you can bet we'll be diving into that when the time comes. "


https://www.techradar.com/news/inte...d-and-its-still-slower-than-the-ryzen-9-3900x
 
Only comes 7% short with a 16.7% core deficit, and 10% faster in single core. I don't see how one can conclude it's "slower" than the 3900X from those numbers.

Yeah, I won't support Intel as a business ever again, but multicore benchmarks where core counts aren't equal are stupid metrics. I still want to know single core IPC.

Oh and power draw.
 
Only comes 7% short with a 16.7% core deficit, and 10% faster in single core. I don't see how one can conclude it's "slower" than the 3900X from those numbers.

Yeah, I would say it's effectively on par but consumes vastly more power and I am guessing costs more?

Yeah, I won't support Intel as a business ever again, but multicore benchmarks where core counts aren't equal are stupid metrics. I still want to know single core IPC.

Oh and power draw.

Being that it boosts to 5.3GHz, it seems like the IPC hasn't really changed much. It has a pretty substantial boost clock advantage vs. the 3900x (around 15% at 5.3 vs 4.6)
 
Yeah, I would say it's effectively on par but consumes vastly more power and I am guessing costs more?



Being that it boosts to 5.3GHz, it seems like the IPC hasn't really changed much. It has a pretty substantial boost clock advantage vs. the 3900x (around 15% at 5.3 vs 4.6)


Right, it's the same architecture as all the previous Lakes. Zen 3 will be out three months after this, and will have better per-core performance (as well as having more cores at the top-end).

Not even slightly exciting. People have been exceeding 5.0 GHz on Skylake since the 7700k :rolleyes:
 
People have been exceeding 5.0 GHz on Skylake since the 7700k

Do they still need to be delidded to hit that? When I built my current system the fact that the 8700k didn't make sense unless you had a healthy OC which required good cooling and delidding was the deciding factor though the extra cost to go with the 8700k didn't help either.

At this point I'm actually amazed at how many incremental improvements they've been able to squeeze out of the 14nm process but they really need to get a smaller node working correctly or their single core performance lead will turn into another loss. I've been enjoying the increased competition and it would be a shame to see the situation just end up reversed rather than staying competitive.
 
Do they still need to be delidded to hit that? When I built my current system the fact that the 8700k didn't make sense unless you had a healthy OC which required good cooling and delidding was the deciding factor though the extra cost to go with the 8700k didn't help either.

At this point I'm actually amazed at how many incremental improvements they've been able to squeeze out of the 14nm process but they really need to get a smaller node working correctly or their single core performance lead will turn into another loss. I've been enjoying the increased competition and it would be a shame to see the situation just end up reversed rather than staying competitive.

Intel will remain competitive - they just don't think you Poles are worth the effort. Ice Lake SP launches in Q3, with up to 38 cores. They don' have enough node room for everyone to use Ice Lake, but they have enough room for servers.

It will be just in time to compete with AMD's Zen 3 servers. Of course AMD has a reason to launch Zen 3 by August (Intel's servers showing first signs of life in years)
 
Yeah, I won't support Intel as a business ever again, but multicore benchmarks where core counts aren't equal are stupid metrics. I still want to know single core IPC.

Oh and power draw.

It’s not a stupid metric if the two products are supposed to be competing in the same price category (although it seems Intel still tends to be quite a bit higher). If AMD is pounding the crap out of Intel at a lower price point and more cores is part of the equation, then the correct conclusion is that Intel should lower their prices or add more cores to compete at that price point, not that the comparison is stupid.
 
Only comes 7% short with a 16.7% core deficit, and 10% faster in single core. I don't see how one can conclude it's "slower" than the 3900X from those numbers.
It's got a 16% core deficit and a 15% clock advantage... seems like a wash for the most part, except scaling cores isn't normally a linear increase in performance, while increasing clocks tends to be more linear. So it comes 7% short with basically a wash on performance metrics, while burning a good amount more power to do so. Never thought I'd be talking about how much more power Intel draws than AMD just to come up short... but here we are.
 
Only comes 7% short with a 16.7% core deficit, and 10% faster in single core. I don't see how one can conclude it's "slower" than the 3900X from those numbers.

And it has a GPU. AMD's still stuck with 4 core APUs on the desktop side. Where are the high core count Zen 2 desktop APUs? mITX builds with an actual usable expansion slot for 10GbE or an HBA you either have an underpowered 3400G or Intel, not a hard choice there.
 
And it has a GPU. AMD's still stuck with 4 core APUs on the desktop side. Where are the high core count Zen 2 desktop APUs? mITX builds with an actual usable expansion slot for 10GbE or an HBA you either have an underpowered 3400G or Intel, not a hard choice there.

oem's are probably getting priority for those chips.. but my guess is they were going to launch in May until the human malware decided to show it's ugly face and ruined it.
 
It’s not a stupid metric if the two products are supposed to be competing in the same price category (although it seems Intel still tends to be quite a bit higher). If AMD is pounding the crap out of Intel at a lower price point and more cores is part of the equation, then the correct conclusion is that Intel should lower their prices or add more cores to compete at that price point, not that the comparison is stupid.

Pounding the crap??? It's 7% faster here in a synthetic bench mark (geekbench).

The only real world software that extra Ryzen cores, will typically beat fewer faster Intel cores, is at is Rendering/Encoding. Things that people tend to walk away from. So it changes you video encoding job from 22mins, to 20 mins. Sure it's quicker, but in the real world that kind of difference is irrelevant. You aren't going to sit there and wait for the 20minute encode to finish. "Oooh, glad I was here watching that encode for 20 minutes to notice the 2 minutes I saved." ;)

Now that changes if encoding is part your money making business, then you go ahead a write off a 64 core thread-ripper...
 
Oh I think I found out who manages the userbenchmarks website... Lol.
Whoever gets it, gets it.
 
And it has a GPU. AMD's still stuck with 4 core APUs on the desktop side. Where are the high core count Zen 2 desktop APUs? mITX builds with an actual usable expansion slot for 10GbE or an HBA you either have an underpowered 3400G or Intel, not a hard choice there.
I also wish they would do a bit more with the APU's, but I get it. The niche use case of something looking for a powerful desktop processor with a really crappy integrated GPU is slim (although, I'm in line if it ever comes out!) I mean, an 8/16 with a couple of navi CU's would be great for a small server build. The thing is, it's really not much more to just buy a 3600x and slap a really cheap pcie x1 in there. If you're building a small server and you need space for drives, you'd be hard pressed to find an ITX case that meets your needs without a PCI slot. If you're looking for a small media player, then you really don't need that many cores. Like I said, pretty niche group it'd be targetting. How many people here do you think bought a 9900k and used the iGPU day to day? Very limited market, although I am one of the few that would be interested in it to replace my dated server eventually, but really I can just throw one of my spare rx 560's in it or something if I needed to use it for installation or plex transcoding or something and it'd probably be much cheaper that way anyways.
 
Pounding the crap??? It's 7% faster here in a synthetic bench mark (geekbench).

The only real world software that extra Ryzen cores, will typically beat fewer faster Intel cores, is at is Rendering/Encoding. Things that people tend to walk away from. So it changes you video encoding job from 22mins, to 20 mins. Sure it's quicker, but in the real world that kind of difference is irrelevant. You aren't going to sit there and wait for the 20minute encode to finish. "Oooh, glad I was here watching that encode for 20 minutes to notice the 2 minutes I saved." ;)

Now that changes if encoding is part your money making business, then you go ahead a write off a 64 core thread-ripper...

Ryzen 3000 series also launched in July of 2019. It has pounded the crap out of Intel’s current offering ever since, while coming in at a lower price point. This is what I’m referencing.

With regard to the current article, Intel’s next generation processor, which has yet to be released, is STILL lagging Ryzen. Yes, it appears Intel is closing the gap with a processor launched last year (still lagging markedly, but by a less embarrassing margin), but that doesn’t change what I said. Price-performance should be the primary purchase decision making factor. If Intel maintains the current pricing model, you’re still going to pay a lot more to get less. This doesn’t somehow invalidate the comparison by virtue of the Ryzen chip being better equipped and delivering more while costing less

Further, based on current roadmaps, Intel is not positioning this chip to compete with the almost one-year old Ryzen 3000 offering, it’s supposed to be competing with Ryzen 4000. If Ryzen 4000 is the kind of upgrade from 3000 and 3000 was from 2000, this will continue to be a situation where Ryzen is continuing to pound the crap out of Intel for less money. That’s just reality.

Intel is going to eventually be able to solve this. They have most of the OEM contracts, far more money, and currently employ the guy who designed Ryzen, but that doesn’t change the fact that, with the current release, and with Intel’s next generation if this leak is reliable, they will continue to lag AMD in raw performance for the foreseeable future. Whether or not AMD does this by offering more cores is simply irrelevant. I, as a consumer, see no reason to buy any of Intel’s current or next generation offering, dollar for dollar, based on this information.
 
It’s not a stupid metric if the two products are supposed to be competing in the same price category (although it seems Intel still tends to be quite a bit higher). If AMD is pounding the crap out of Intel at a lower price point and more cores is part of the equation, then the correct conclusion is that Intel should lower their prices or add more cores to compete at that price point, not that the comparison is stupid.
No you're not wrong, I don't think anybody buying these processors are doing it for single core IPC advantages obviously looking at price per multi-threaded workflow is a pretty decent metric. I'm just genuinely more interested in single-core IPC as I would like to see if AMD has been able to surpass Intel with single-core IPC

And also the power envelope necessary to do that. I don't run multi-threaded workflows ever, so for me it's a useless metric... obviously for a Content producer that's not the same case. I just don't agree that they're getting crushed right now based on plus or minus 15% in a benchmark that allows for higher core # advantages.

Price parity is very important, but that doesn't seem to be Intel's main concern right now so.
 
Running your OS is a multi-threaded workload..........

Nearly everything today is multi-threaded, but multi-threaded is a meaningless designation on its own.

In order to actually have a significant benefit beyond 8 cores, you need a very specific kind of multi-threaded workload: An Embarrassingly Parallel workload.

Generally, for most users, these are only Rendering/Encoding tasks, which when it gets down to things a significant portions of computer users actually do, it's Video Encoding.

So for people who aren't doing a lot of Video encoding, they extra cores are usually wasted.
 
Nearly everything today is multi-threaded, but multi-threaded is a meaningless designation on its own.

In order to actually have a significant benefit beyond 8 cores, you need a very specific kind of multi-threaded workload: An Embarrassingly Parallel workload.

Generally, for most users, these are only Rendering/Encoding tasks, which when it gets down to things a significant portions of computer users actually do, it's Video Encoding.

So for people who aren't doing a lot of Video encoding, they extra cores are usually wasted.

This is some great gamer science here but reality is people use computers normally, without thinking about closing one program in order to open another comfortably. Having an abundance of threads to run whatever you want whenever you want is what you should be after unless you treat your PC like a console, with its limited resources...

And yet the OS will run fine on two.

Fine? You call a fresh install of Windows 10 home running with 50-100% CPU usage at all times Fine?
 
I also wish they would do a bit more with the APU's, but I get it. The niche use case of something looking for a powerful desktop processor with a really crappy integrated GPU is slim (although, I'm in line if it ever comes out!) I mean, an 8/16 with a couple of navi CU's would be great for a small server build. The thing is, it's really not much more to just buy a 3600x and slap a really cheap pcie x1 in there.

But then you've used up one (and in my case, my only) PCIe slot for video card that could have been integrated into the CPU. If I could ever find Asrock's X570D4I-2T board for sale, that's what I'd get to get around the gpu issue: https://www.asrockrack.com/general/productdetail.asp?Model=X570D4I-2T#Specifications

If you're building a small server and you need space for drives, you'd be hard pressed to find an ITX case that meets your needs without a PCI slot.

Sure, they have a single PCI slot, which currently houses a LSI 9260-8i to handle all the drives.

If you're looking for a small media player, then you really don't need that many cores. Like I said, pretty niche group it'd be targetting. How many people here do you think bought a 9900k and used the iGPU day to day? Very limited market, although I am one of the few that would be interested in it to replace my dated server eventually, but really I can just throw one of my spare rx 560's in it or something if I needed to use it for installation or plex transcoding or something and it'd probably be much cheaper that way anyways.

That's the beauty of using an Intel solution, you get quick sync out of the box and Plex takes full advantage of it.
 
In order to actually have a significant benefit beyond 8 cores, you need a very specific kind of multi-threaded workload: An Embarrassingly Parallel workload.
So for people who aren't doing a lot of Video encoding, they extra cores are usually wasted.

so if that's the case then i'd say grab a 37 or 3800X and dump the $200+ you'll save on a better graphics card. Plus you won't be using an Embarrassingly Unsecure pc platform.
 
No you're not wrong, I don't think anybody buying these processors are doing it for single core IPC advantages obviously looking at price per multi-threaded workflow is a pretty decent metric. I'm just genuinely more interested in single-core IPC as I would like to see if AMD has been able to surpass Intel with single-core IPC

And also the power envelope necessary to do that. I don't run multi-threaded workflows ever, so for me it's a useless metric... obviously for a Content producer that's not the same case. I just don't agree that they're getting crushed right now based on plus or minus 15% in a benchmark that allows for higher core # advantages.

Price parity is very important, but that doesn't seem to be Intel's main concern right now so.

Intel has an advantage predominantly in gaming, but again, if you break things down to dollar per frame, that advantage evaporates. You will get a higher FPS yield on an i9 than a 3900X, for example, but you’re also paying disproportionately more for that performance.

Intel is probably less concerned about price right now because they don’t have to be. Most of the big OEMs are pretty well Intel only, and AMD needs to make the price lower as part of a market share play, but from a strictly engineering standpoint, Intel is currently way behind, which also means the risk is that some of those OEMs will begin to build Ryzen systems of consumer demand shifts in that direction. I think that’s when you’ll see Intel get more aggressive with price, but sales for them are still strong overall.
 
erek

Nope no interest at all unless someone sells me an x299 and 7820x for 300 total for my sons rig.
 
Fine? You call a fresh install of Windows 10 home running with 50-100% CPU usage at all times Fine?
If the machine's only purpose is to run a browser or a spreadsheet, sure.

Also, you're definitely overestimating resource usage there. Modern cores, two is plenty to get a lot of work done. Obviously you'd want a few more, but we're speaking to what's necessary.

Plus you won't be using an Embarrassingly Unsecure pc platform.
Given that there are no secure platforms, it might even be more ideal to get a platform where the vulnerabilities are known ;)

Intel has an advantage predominantly in gaming, but again, if you break things down to dollar per frame, that advantage evaporates. You will get a higher FPS yield on an i9 than a 3900X, for example, but you’re also paying disproportionately more for that performance.
You can say the same about GPUs, monitors, keyboards, mice, headsets... etc. used for gaming too. Yet those parts are still desirable because they do something that others cannot.

Intel is probably less concerned about price right now because they don’t have to be. Most of the big OEMs are pretty well Intel only, and AMD needs to make the price lower as part of a market share play, but from a strictly engineering standpoint, Intel is currently way behind, which also means the risk is that some of those OEMs will begin to build Ryzen systems of consumer demand shifts in that direction. I think that’s when you’ll see Intel get more aggressive with price, but sales for them are still strong overall.
Agreed; AMDs production limitations alongside Intel's market inertia, production capacity, and still strong performance for the vast majority of workloads mean that Intel doesn't have to lower prices to avoid accumulating stock.
 
I initially got a 3800x and realised I want more cores for folding and occasional encoding besides gaming. So I moved up to a 3900x, then 3950x. The 3950x basically allows me to fold or run other CPU intensive tasks while gaming if I want to. Nothing seems to faze it.

If the 10900K loses to 3900x in multithreaded benches, it basically costs more for less performance while consuming more power and also requires a new platform. ST advantage for gaming? If that extra 5% FPS over the 3900x really matters, the 9900k should be similar. I just don't see the point of these new CPU's

Given that there are no secure platforms, it might even be more ideal to get a platform where the vulnerabilities are known ;)

And yet every few weeks new Intel vulnerabilities keep getting discovered which also reduces performance in most instances when fixed.
 
No you're not wrong, I don't think anybody buying these processors are doing it for single core IPC advantages obviously looking at price per multi-threaded workflow is a pretty decent metric. I'm just genuinely more interested in single-core IPC as I would like to see if AMD has been able to surpass Intel with single-core IPC

And also the power envelope necessary to do that. I don't run multi-threaded workflows ever, so for me it's a useless metric... obviously for a Content producer that's not the same case. I just don't agree that they're getting crushed right now based on plus or minus 15% in a benchmark that allows for higher core # advantages.

Price parity is very important, but that doesn't seem to be Intel's main concern right now so.

AMD already surpassed Intel's IPC with Zen 2. Intel's only advantage now is raw clock speed.
 
And it has a GPU. AMD's still stuck with 4 core APUs on the desktop side. Where are the high core count Zen 2 desktop APUs? mITX builds with an actual usable expansion slot for 10GbE or an HBA you either have an underpowered 3400G or Intel, not a hard choice there.
Have to agree with you on this one.
This is something I've been looking forward to as well, and while Intel has the higher-core with iGPU options, AMD is kind of limited by using just quad-core APUs.

If they start using 8-core CPUs with an iGPU, that is an instant win for just such a setup and scenario - here's to hoping it will be sooner than later.
 
That line about a thing with known vulnerabilities being somehow better, I must say:

There's no such thing as a _secure home_, so I hope that you don't bother locking up whenever you leave home, also I hope that everyone that knows where you live can tell with a glance that you left everything open.
 
I was initially going to replace my 8600K (dud that does only 4.75GHz at like 1.31v as one core is very bad) with a 9700K hopefully pushing to 4.9~5GHz with 2 more additional cores as a simple drop-in upgrade without need to swap other parts when price drops but Intel has been STUBBORNLY keeping prices so high on 9000 series with no interest to what AMD is offering that it looks like they won't be adjusting 9000 series at all and rather let it die quickly and focus on the 10k series. But the stupid decision to once again require new motherboard for such a modest change was the final nail in the coffin.

Now I'm rather just waiting for Ryzen 4000 desktop parts seeing how well the laptop CPUs do and hopefully even Intel's only hope they have clinged onto forever is that single thread advatnage will be lost as I do mostly play games (with a 1080p 240Hz monitor mind you) so therefore I've been somewhat reluctant to go AMD until now even if we're talking perhaps some 5% or 7% here and there difference. Now it seems even that gap will be gone I think many "gamers" won't either have any problems going AMD that previously hold onto Intel for that reason.

With my overclocking luck I'm just glad both Intel and AMD seems to squeesh everything possible out of the CPUs these days. :p It's crazy to think we've gone to a situation when manufacturer will make faster stock parts than what a lot of samples would overclock to manually just 3 series back.
 
Last edited:
Footage from i9-10900K testing site!

tenor.gif
 
It's the difference between knowing where an attack might come from and having no clue.

Security researchers have had 5 years to tear into Skylake.

It's a pretty heavy spin, attacks and exploits are now a PLUS, because you know about them. Ignoring the ones that haven't been found, or the obvious tradeoff that was made for years to sacrifice security for speed.

Also, they've had three years to tear into Zen, so I'd expect the ratio of exploits to be 3:5.
 
It's a pretty heavy spin, attacks and exploits are now a PLUS, because you know about them. Ignoring the ones that haven't been found, or the obvious tradeoff that was made for years to sacrifice security for speed.
It's absolutely a spin, but understand that these newer iterations of Skylake also include hardware mitigations.

Also, they've had three years to tear into Zen, so I'd expect the ratio of exploits to be 3:5.
Sure, but Zen represents some fraction of a fraction of the total installed base. Skylake is literally everywhere.

There's still significantly more value in targeting and protecting Skylake versus Zen, but that trend reverses if / when the Zen install base grows to something more significant.
 
Back
Top