Intel Demoes i9-11900K Against Ryzen 9 5900X

[SavageNeo]

seems like intel will sell the 11900k at 5900x' price point. makes no sense. 4 less cores with single digit gains at gaming? That is nuts.

Also waiting for reviews to show the power usage and temps. Will this be an oven or actually decent power usage?

if 11900k and 5900x are at the same price point, I would always go with the 5900x. Also cheaper motherboards and cooling.

 

5800x is single CCD so in some games it should perfrom better than 5900x since it has 2 CCDs.

 

Anyway I think that this 11th gen will be bad. Hope that the 10nm cpus will be better. Kinda dissapointing tbh.

[EpiCheeseTime]

I hope I'm not the only one that started reading it as "de-mode" then realized its "demo-ed"...

 

Or "de-moes" rather than "demo-os"...

 

Is it time to go to bed yet...

[D13H4RD]

I'm quite sure Rocket Lake is that "stopgap"-esque release. I can't really see 10900K users (and especially Ryzen 5900X users) looking at this and be interested.

 

The one people are really interested in is Alder Lake, because that is by far and away a pretty darn big change not just for Intel but potentially for x86 processor design. Rocket Lake's significant in the sense that it uses backported Sunny Cove cores, of which the 14nm backport likely led to limitations that made the top-end SKU limited to 8C/16T.

[leadeater]

1 hour ago, TetraSky said:

Temperature difference?

 

High

 

1 hour ago, TetraSky said:

Power Consumption ?

 

A lot

[AndreiArgeanu]

If the 11900K is priced similarly to the 5900X it's a really big oof for Intel, I'd take 4 more cores instead of the single digit improvements that I won't notice. Hopefully their next generation cpu's will actually be better, especially the power consumption, if it continues like this, on the same 14nm process, Intel might end up with consumer cpu's that surpass AMD's FX 9590 220W TDP.

[Valentyn]

Me with my 5950X that boosts to 5.2Ghz for single core.

[SavageNeo]

7 minutes ago, Valentyn said:

Me with my 5950X that boosts to 5.2Ghz for single core.

When you have SSD and you accidentally restart your pc

[SteveGrabowski0]

20 minutes ago, AndreiArgeanu said:

If the 11900K is priced similarly to the 5900X it's a really big oof for Intel, I'd take 4 more cores instead of the single digit improvements that I won't notice. Hopefully their next generation cpu's will actually be better, especially the power consumption, if it continues like this, on the same 14nm process, Intel might end up with consumer cpu's that surpass AMD's FX 9590 220W TDP.

Shit seems like it would be DOA if you could actually get your hands on a 5900X these days. About the only thing going for it would be that it'll be in stock everywhere.

[Valentyn]

8 minutes ago, SavageNeo said:

When you have SSD and you accidentally restart your pc

When I finally moved to all NVMe storage, and my only SATA SSD/HDDs are external raids.

[LAwLz]

8 hours ago, leadeater said:

High

 

A lot

Intel bad. 

 

AMD good. 

[Moonzy]

I'm interested in getting 11th gen, if they're priced right of course

 

Patiently waiting for reviews

[leadeater]

48 minutes ago, LAwLz said:

Intel bad. 

 

AMD good. 

Well it's not like the high end Intel K SKUs aren't already hotter and use more power, these have a even higher TDP and PL2 so it's not like this isn't true.

 

Not that this shouldn't stop a silly joke anyway 

 

Spoiler

Last time I ever cared about power or heat: Never 

 

[porina]

5 hours ago, comander said:

At the same timings, using dual rank RAM can give up to 20% better performance (extreme case is 7-zip... even AOTS benefits by 12%).

This has been known for many years, although it only seemed to be "rediscovered" at Zen 3 launch for some reason. Note Intel CPUs also benefit from this. I first encountered this years ago when by chance I got dual rank ram on my first 6700k system, and couldn't work out why my second 6700k system which has single rank ram performed much worse. Ideally any testing would be like for like.

 

5 hours ago, comander said:

I think a lot of what happened with reviewers is that the notion of "more RAM doesn't matter" resulted in set ups with a limit number of modules (or lower capacity/lower rank modules -MORE ranks usually just means they plop in more ICs) being used... and then it was only natural to run those modules FAST. AMD's set up benefits less from this vs somewhat slower and more ranks. 

That's a side question I wonder about at times. Just how much ram should be present in a gaming system today? I don't see a compelling need to go beyond 16GB for general gaming myself. Some specific games, especially if heavily modded, might need more, but I'll leave those exceptions out for the masses. The problem then is that 8GB modules today are all single rank. You'd have to go back years to find single rank modules at 8GB capacity. Even 16GB modules are going single rank, so that's not even a safe choice for ultimate performance seekers any more. 4GB modules seem to be on the way out, but running 4x4GB could be one way to get 16GB with 2 DPC (not strictly the same as 2 rank modules, but for practical purposes it is). 4x8GB is probably the safest way to do it now, but you then get more ram than might be really necessary.

 

5 hours ago, comander said:

Also... who the heck uses top end cards at 1080p?

That's more a test requirement in order to see the small differences these CPUs offer. IMO you reach "good enough" performance at lower CPU level than these best of best offerings. Still, for the manufacturers, it is a marketing win to be able to claim to have the best performing CPU in some area.

[cj09beira]

On 12/30/2020 at 10:17 PM, CarlBar said:

 

Intels latest 14nm nodes are actually ahead of their planned 10nm performance. I'd have to dig around to be sure if my memory is acurratte but i seem to recall 7nm+ is supposed to have about the same perf per watt advantage over 7nm as 5nm does over 7nm+.I belive it works out at a cumulative 60% perf per watt between 7nm and 5nm.

if i remember right 10nm+ is when they expect to beat 14nm++++

[porina]

13 minutes ago, cj09beira said:

if i remember right 10nm+ is when they expect to beat 14nm++++

The first is not used any more, and the latter never existed.

 

Look at Tiger Lake for a practical example. This is what Intel now call 10nm SuperFin (10SF). Compared to the 10nm version used for Ice Lake, they've overcome the clock wall and it runs much more efficiently now. The first Tiger Lake offerings went up to 4.8 GHz, and the just announced ones nudge 5.0 GHz. Also these are mobile CPUs, if implemented in a desktop form more might be possible. Depending on how you count it, Tiger Lake is also 1 to 2 architecture updates from Skylake with the IPC improvements that come with that.

[RejZoR]

@porina

 

Why would single or dual rank affect things? I always thought single rank was better because they usually clock higher and can run at tighter timings than dual rank RAM. But having same speed and same timings, why would there be a 20% performance difference between single and dual rank RAM ?

[XenosTech]

6 minutes ago, RejZoR said:

@porina

 

Why would single or dual rank affect things? I always thought single rank was better because they usually clock higher and can run at tighter timings than dual rank RAM. But having same speed and same timings, why would there be a 20% performance difference between single and dual rank RAM ?

One dual rank dimm is equivalent to two single rank dimms of the same speed and timings

 

Yes in some cases single rank dimms will OC  way better but board and cpu imc still plays a key factor there and the quality on the chips on either ram set up

[leadeater]

13 minutes ago, RejZoR said:

@porina

 

Why would single or dual rank affect things? I always thought single rank was better because they usually clock higher and can run at tighter timings than dual rank RAM. But having same speed and same timings, why would there be a 20% performance difference between single and dual rank RAM ?

 

Quote

A memory rank is a set of DRAM chips connected to the same chip select, which are therefore accessed simultaneously. In practice all DRAM chips share all of the other command and control signals, and only the chip select pins for each rank are separate (the data pins are shared across ranks).

 

 

Quote

Multi-rank modules allow several open DRAM pages (row) in each rank (typically eight pages per rank). This increases the possibility of getting a hit on an already open row address. The performance gain that can be achieved is highly dependent on the application and the memory controller's ability to take advantage of open pages

 

[kewtz]

Maybe it's #toosoon - the measurement isn't 1080p anymore.

The new standard is 1440p, that should be the new measurement if Intel/AMD are going to display one-upping each other.

 

I'm surprised Intel touted this, honestly. It doesn't seem to be enough to convert a Zen2 and certainly not enough to convert a Zen3 owner.

We can reasonably assume this chip will be hotter, for a minimal 1080p game in a very limited set of titles.

For the enthusiasts we claim we are, this new chip isn't delivering enough, imo.

 

Until Intel sorts out 7nm, meh.

 

Also I just bought a 5900x, I'm kinda sorta talking to myself to justify the purchase.

[porina]

2 hours ago, RejZoR said:

@porina

 

Why would single or dual rank affect things? I always thought single rank was better because they usually clock higher and can run at tighter timings than dual rank RAM. But having same speed and same timings, why would there be a 20% performance difference between single and dual rank RAM ?

The topic has been covered again and again in the past, most recently with Zen 3 launch. I'd suggest looking up the GN video on it as a starting point. It was one of their least worst ones.

[RejZoR]

47 minutes ago, porina said:

The topic has been covered again and again in the past, most recently with Zen 3 launch. I'd suggest looking up the GN video on it as a starting point. It was one of their least worst ones.

It really wasn't. Not to mention how insanely difficult it is to dig out actual info on the RAM sticks you're buying. It's hard to find out if it's SR or DR and finding out which die revision it is, well that's almost mission impossible. And I always got perception everyone was praising SR memory sticks. Seriously, I've been building PC's for 20 years and this is the first time I heard such a thing. Go figure.

[porina]

30 minutes ago, RejZoR said:

It really wasn't. Not to mention how insanely difficult it is to dig out actual info on the RAM sticks you're buying. It's hard to find out if it's SR or DR and finding out which die revision it is, well that's almost mission impossible. And I always got perception everyone was praising SR memory sticks. Seriously, I've been building PC's for 20 years and this is the first time I heard such a thing. Go figure.

It really isn't new. I only first heard of it about 5 years ago, but it was around far longer than that. But because Zen 3 is the new hotness it got rediscovered when different people got different benchmark results. The SR > DR thing is mostly a misunderstanding and irrelevant to anyone not overclocking ram hardcore. In practice DR > SR.

[Kisai]

32 minutes ago, RejZoR said:

It really wasn't. Not to mention how insanely difficult it is to dig out actual info on the RAM sticks you're buying. It's hard to find out if it's SR or DR and finding out which die revision it is, well that's almost mission impossible. And I always got perception everyone was praising SR memory sticks. Seriously, I've been building PC's for 20 years and this is the first time I heard such a thing. Go figure.

 

SR can get higher clocks, but its fair(er) to compare two SR to one DR, as there's only two memory channels, and many motherboard manuals have stuff like this in it:

Basically mixing SR/DR will only let you operate the memory at the slower spec, and you only get the higher spec with one module in each channel because two DR > two SR > four SR > four any configuration.

 

A DR module is essentially better than two SR modules, but it also means if you want a higher speed you can't populate the other slot on the same channel.

 

 

[Unixsystem]

9 hours ago, kewtz said:

Maybe it's #toosoon - the measurement isn't 1080p anymore.

The new standard is 1440p, that should be the new measurement if Intel/AMD are going to display one-upping each other.

These are CPU benchmarks, upping the resolution only increases the chance of encountering a bottleneck somewhere else in the system. For GPUs or full system benchmarks 1440p+ makes sense but if you're doing direct CPU vs CPU comparisons then the lower the resolution the better (within reason). 

[PocketNerd]

So from a statistics point of view this is nearly all within margin-of-error with only one game being just barely statistically signficant. Ergo, Intel is still creating a worse part overall (when you factor in thermals, die size, etc).

RIP Intel