Question 10900k real world comparison to 3900k . Maybe, if I can buy one.

[Keysplayr]

For gaming yes. But it's hard to acquire in DIY so it may not even be an option.



Nope, one more next year, called Rocket Lake.



For desktop it's another two years for something that's not 14 nm.

Oh man. I hardly game anymore and am focused on music DAWs and Video creation. I'm thinking then I'll likely go with Matisse refresh or Zen3. Thanks for the info.

 

[JTaylor2005]

So Der8aur did a video of him with a 4kw chiller and watercooling loop, overclocking the 10900k all the way up to 5.4Ghz with a serious water block and achieved a cinebench of under 7000.

Here is a chap doing the same cinebench test getting over 7000 points with a 3900x on a stock AMD cooler.... With stock settings I believe.

 

[amrnuke]

So Der8aur did a video of him with a 4kw chiller and watercooling loop, overclocking the 10900k all the way up to 5.4Ghz with a serious water block and achieved a cinebench of under 7000.

Here is a chap doing the same cinebench test getting over 7000 points with a 3900x on a stock AMD cooler.... With stock settings I believe.

I'm a big fan of AMD, but CB20 scales pretty tightly to # of threads. The 10900K has a 20% thread count deficit. In most reviews, it has a 20% deficit in CB20 MT. Makes sense that throwing more GHz at the problem wouldn't always solve the deficit in threads.

I'm guessing if you put a 4000W cooler on the 3900X it would still lose to the 10900K with a $150 AIO in gaming and Tensorflow.

So I'm not sure the point? Each CPU has its pluses and minuses.

 

[Rigg]

I was looking at a build a guy is considering in Computer Building. I looked at the board you mention, but he mentioned the Asus Prime Z490 A. It looks significantly better in VRM and, supposedly, memory performance, for only $30 more.

I'm pretty sure those 2 boards have the exact same Vcore VRM. It's a 4 phase with 3 x 50A DrMos powerstages per phase. This is the same Vcore design they use on several x570 boards as well.

 

[Markfw]

I'm a big fan of AMD, but CB20 scales pretty tightly to # of threads. The 10900K has a 20% thread count deficit. In most reviews, it has a 20% deficit in CB20 MT. Makes sense that throwing more GHz at the problem wouldn't always solve the deficit in threads.

I'm guessing if you put a 4000W cooler on the 3900X it would still lose to the 10900K with a $150 AIO in gaming and Tensorflow.

So I'm not sure the point? Each CPU has its pluses and minuses.

So even though the IPC is close, the clock speed is almost 1 ghz higher, but the 3900x wins, you don't think that says something good about the 3900x ?

Well, now that newegg had them, I am on the notify list for 3 or 4 sku's.

 

[gorobei]

buildzoid via gamersnexus came out with his z490 recommendations list. he covers a bunch of price ranges and usage cases. closed caption available.

@Markfw if you havent purchased yet, here are his $200 suggestions:

msi mag z490 tomahawk- 6phase(doubled) 12x 55amp powerstage [1gig intel lan]+ [2gig realtek] tabbed memory traces

msi mpg z490 gaming edge - (adds wifi6) same 6phase but 60amp pwstg [2gig rt] tabbed memory traces

gigabyte z490 vision G -12phase 60amp pwstg, 3x pcie slots, lots of rear i/o ports

per BZ 10900 runs hot (low voltage so higher current = more heat) the msi heat sinks are big enough to handle the heat output.
if you dont think you will keep the 10900k rig maybe base your choice on how attractive the board features will be to a 2nd hand buyer.

 

[aigomorla]

Oh man. I hardly game anymore and am focused on music DAWs and Video creation. I'm thinking then I'll likely go with Matisse refresh or Zen3. Thanks for the info.

keys! LTNS yo..
Kinda ironic saying you don't game anymore when you used to be a Nvidia Focus Group.

For me its the otherway around.
I still am a hardcore gamer, so i am always after the fastest gaming choice.
I also do a lot of encodes, so i really need that AVX.
My family would cry if my Plex server could not transcode, but i have a Quadro P2000 doing that over NVEC, and the cpu also has QuickSync, but the Quadro is probably more overpowered for job.

The only reason i would go AMD is probably for the PCI-E lanes as i have broke that 10TB of SSD capacity, and want to double it next time in pure nVME's, which will require a ton of PCI-E lanes as i also run dual Video Cards and also want that 10GBE nic.
So yeah, most likely ThreadRipper is in my next lineup as main.

 

[amrnuke]

So even though the IPC is close, the clock speed is almost 1 ghz higher, but the 3900x wins, you don't think that says something good about the 3900x ?

Yes, it says a lot about the 3900X, of course. The 3900X is stellar in CB20.

1) 3900X all-core is somewhere around 4.1-4.2 GHz. 10900K all-core in this was 5.3 GHz. That's +26% clock speed for 10900K.
2) Scaled SPECint2006 scores for 9900K (not sure it's been done on 10900K yet) and 3900X --- the 3900X has an apparent "IPC" lead of 6.5% (16.01 SPECint2006 per GHz for 3900X versus 15.03 on 9900K).
3) 3900X has a +20% core/thread count

If things scaled perfectly, we should see AMD leading by about 0.5% in CB20 MT, but instead it's a larger lead. So I think what these CB20 scores show is that AMD in multi-threading tasks has a very nice architectural benefit.

However, my point was more that I think it was stupid to put a 4000W cooler on the 10900K - you can all-core boost the 10900K to 5.1GHz with almost any decent 240mm AIO and achieve CB20 scores within 7-8% of a 3900X, on one of the tests in which AMD chips see the highest consistent benefit.

And because of that, I think the 4000W cooler video is just sensationalist garbage. One could just as easily slap a 4000W cooler on the 3900X and show it still losing to Intel in gaming, and it would be just as sensationalist and just as useless.

 

[Markfw]

Yes, it says a lot about the 3900X, of course. The 3900X is stellar in CB20.

1) 3900X all-core is somewhere around 4.1-4.2 GHz. 10900K all-core in this was 5.3 GHz. That's +26% clock speed for 10900K.
2) Scaled SPECint2006 scores for 9900K (not sure it's been done on 10900K yet) and 3900X --- the 3900X has an apparent "IPC" lead of 6.5% (16.01 SPECint2006 per GHz for 3900X versus 15.03 on 9900K).
3) 3900X has a +20% core/thread count

If things scaled perfectly, we should see AMD leading by about 0.5% in CB20 MT, but instead it's a larger lead. So I think what these CB20 scores show is that AMD in multi-threading tasks has a very nice architectural benefit.

However, my point was more that I think it was stupid to put a 4000W cooler on the 10900K - you can all-core boost the 10900K to 5.1GHz with almost any decent 240mm AIO and achieve CB20 scores within 7-8% of a 3900X, on one of the tests in which AMD chips see the highest consistent benefit.

And because of that, I think the 4000W cooler video is just sensationalist garbage. One could just as easily slap a 4000W cooler on the 3900X and show it still losing to Intel in gaming, and it would be just as sensationalist and just as useless.

ok, I think the 4000 watt cooler is a little over the top. I think the point that they were trying to make, is that

1) It required some good cooling to get decent performance out of the 10900k
2) No matter what the cooling, it will still lose on a lot of benchmarks

I have to agree with those points as well as yours.

 

[Arkaign]

I sort of feel the 10900k is a bridge too far in practical terms for desktop on 14nm.

I just have a big air cooler, and get 5.3Ghz on a 9900KS. And for gaming, basically the only reason to have an Intel system in the age of Zen+/Zen2, 10C is no better than 8C once frequency is accounted for (stock 10th gen K turbo is higher along with slightly higher stock 'official' ram speeds, which is moot for tuner OC crowd anyway).

You just end up with extra heat and expenses, and STILL lose handily to 3900X, to say nothing of 3950X in basically any heavy MT work tasks, both of which will be monumentally more efficient than a maxed out 9900/10700/10900 rig.

The only interesting 10th gen parts to me are the 10400f (need a little price cut imho, but a good, efficient CPU surprisingly) and the 10600k (once OC'd and tuned, is up with 9900/10700/10900 gaming performance). The 10700k is ok, but is basically a 9900k with very slightly better stock turbo behavior. The 10900k is just an expensive, hot waste (IMHO). I guess I could give a nod to the i3 models as a cheap option, but also feel the price of a 3300 makes that a better option, particularly with Intel's HORRIBLE decision to limit non Z boards to 2666 Ram (notwithstanding some renegade bioses out there). 3600 Ram is cheap, so a combo of H series board, i3 maybe $20 cheaper than current MSRP, and non kneecapped Ram support for sub $100 mobos, and the i3 wouldn't be such a 'ehhh' option.

It's just so .. stupid 😑

 

[DrMrLordX]

@Arkaign

It's not so much stupidity as it is desperation.

 

[JoeRambo]

The only interesting 10th gen parts to me are the 10400f (need a little price cut imho, but a good, efficient CPU surprisingly) and the 10600k (once OC'd and tuned, is up with 9900/10700/10900 gaming performance). The 10700k is ok, but is basically a 9900k with very slightly better stock turbo behavior. The 10900k is just an expensive, hot waste (IMHO). I guess I could give a nod to the i3 models as a cheap option, but also feel the price of a 3300 makes that a better option, particularly with Intel's HORRIBLE decision to limit non Z boards to 2666 Ram (notwithstanding some renegade bioses out there). 3600 Ram is cheap, so a combo of H series board, i3 maybe $20 cheaper than current MSRP, and non kneecapped Ram support for sub $100 mobos, and the i3 wouldn't be such a 'ehhh' option.

One thing to watch is FSB overclocking, for example there is this dude on reddit:

https://www.reddit.com/r/intel/comments/h94nq9
That is a bit of game changer on those locked CPUs, for example @4.6Ghz 10400F would be much sweeter deal for the price it is now.

 

[lobz]

One thing to watch is FSB overclocking, for example there is this dude on reddit:

https://www.reddit.com/r/intel/comments/h94nq9
That is a bit of game changer on those locked CPUs, for example @4.6Ghz 10400F would be much sweeter deal for the price it is now.

Game changer? I'd be very surprised if you didn't know how long fsb oc has been around. Do you know how lucky you have to be with multiple components at the same time to not fail miserably with any meaningful amount of fsb oc?

 

[JoeRambo]

Game changer? I'd be very surprised if you didn't know how long fsb oc has been around. Do you know how lucky you have to be with multiple components at the same time to not fail miserably with any meaningful amount of fsb oc?

The thing is, the limit was always 102ish, and here is this guy, running 107, that is full 5% extra performance.




If PEG/DMI is on separate clocks, it is quite likely I/O and graphics are fine when touching base clock?

But that is not what i meant with game changer, the fact that supposedly completely locked is.

 

[lobz]

The thing is, the limit was always 102ish, and here is this guy, running 107, that is full 5% extra performance.

View attachment 23165


If PEG/DMI is on separate clocks, it is quite likely I/O and graphics are fine when touching base clock?

But that is not what i meant with game changer, the fact that supposedly completely locked is.

I see how you meant it now

 

[DrMrLordX]

Most platforms haven't had I/O truly locked out from bclk for awhile. Interesting if true.

 

[Arkaign]

Most platforms haven't had I/O truly locked out from bclk for awhile. Interesting if true.

True. I think it's mainly X58/79 that really let you mess with that stuff in (relatively) recent eras for the most part, never much messed with X99. I think the i3/i5/i7 S1156 also was fairly friendly but it's been so long other than cheap OEM experience with those that idk. I went s1366 at the time and then SB after lol for my own gear.

 

[DrMrLordX]

True. I think it's mainly X58/79 that really let you mess with that stuff in (relatively) recent eras for the most part, never much messed with X99. I think the i3/i5/i7 S1156 also was fairly friendly but it's been so long other than cheap OEM experience with those that idk. I went s1366 at the time and then SB after lol for my own gear.

It would be good for the dedicated overclockers if both AMD and Intel could go back to the "good old days" of being able to adjust bclk all wily-nilly. I miss bclk OC like I had on my s754 Sempron 2800+ and AM2 x2-3600+. Kudos to Intel if it's back on Z490, for what it's worth.

 

[Arkaign]

It would be good for the dedicated overclockers if both AMD and Intel could go back to the "good old days" of being able to adjust bclk all wily-nilly. I miss bclk OC like I had on my s754 Sempron 2800+ and AM2 x2-3600+. Kudos to Intel if it's back on Z490, for what it's worth.

True. Man I tell you as far as my experience, Zen OG was kind of painful but promising, Zen+ was quite a bit of fun to tune and cool and tweak, and then Zen2 is this awesome but fairly boring thing as far as conventional tuning. Deep diving into Ryzen memory tuning with Zen2 is pretty rewarding, but for some reason just not as compelling as getting all core OCs and uncore blazing, at least to me.

I have moderate hopes that the Zen2+ XTs will be a little more tunable with mature stepping.

 

[blckgrffn]

True. Man I tell you as far as my experience, Zen OG was kind of painful but promising, Zen+ was quite a bit of fun to tune and cool and tweak, and then Zen2 is this awesome but fairly boring thing as far as conventional tuning. Deep diving into Ryzen memory tuning with Zen2 is pretty rewarding, but for some reason just not as compelling as getting all core OCs and uncore blazing, at least to me.

I have moderate hopes that the Zen2+ XTs will be a little more tunable with mature stepping.

I agree. After years of getting more mhz so easily with the right setups, the idea of basically no extra frequency (which seems to lift all applications performance) the idea of tweaking for memory performance is not nearly as compelling. Also historically speaking it seemed way less impactful than it is with Zen 2.

Once upon a time I obsessed over 1T vs 2T command rates and all that stuff... but I like higher (frequency) numbers Ideally for free and not with stupid increases in TDP

 

[Pumice]

Is the Ryzen 9 perfectly fine at running most games with high FPS? I plan on playing COD, Battlefield, CS Go, Far Cry , Dota2 , Doom and a bunch of other FPS.

I read a thread here that had one guy having lower FPS on some game when he switched from an Core I7 8700K to a Ryzen

 

[LikeLinus]

Is the Ryzen 9 perfectly fine at running most games with high FPS? I plan on playing COD, Battlefield, CS Go, Far Cry , Dota2 , Doom and a bunch of other FPS.

I read a thread here that had one guy having lower FPS on some game when he switched from an Core I7 8700K to a Ryzen

Intel is going to give you the highest overall FPS in gaming.

 

[jpiniero]

Was again available for two hours this morning (~2 AM EST).

 

[Pumice]

Intel is going to give you the highest overall FPS in gaming.

Right, but what I would like to know is if the drop in FPS for the Ryzen causes a large enough drop in FPS that gaming at 1920 x 1080 or 2560 x 1440 or 3840 x 2160 will be unpleasant , or will I still get smooth gameplay ?

At the moment I am either going with the Radeon rx 5700 xt or the GeForce RTX 2060 super.

 

[Zucker2k]

Right, but what I would like to know is if the drop in FPS for the Ryzen causes a large enough drop in FPS that gaming at 1920 x 1080 or 2560 x 1440 or 3840 x 2160 will be unpleasant , or will I still get smooth gameplay ?

At the moment I am either going with the Radeon rx 5700 xt or the GeForce RTX 2060 super.

You will get smooth gameplay, but you'd still be leaving some performance on the table.