AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

[CAT-THE-FIFTH]

I should have thought about a Budget Intel ASRock, VRM heat sync still look's inferior, like they are trying to save money, the SOC doesn't have one at all.
They technically are 3 phase, yes, but current is still split across 6 Hi and Low Fet's and Chockes, because there are twice as many sets the Ampage is divided up between more components = less heat per component and that makes a huge difference.

74c is the higest i have ever seen them and that was overclocked during extensive stress testing, they are rated for 125 to 150c.

You can get the B450M Mortar for £80ish now and that has a solid VRM and good cooling so not sure why OcUK is selling them for so much. The ASRock boards also appear to at least implement properly finned heatsinks on their B450 boards.

Also WTF is the Ryzen 5 2600 costing £170 here?? Its £140 to £150 at most other places I looked at!!

 

[humbug]

You can get the B450M Mortar for £80ish now and that has a solid VRM and good cooling so not sure why OcUK is selling them for so much. The ASRock boards also appear to at least implement properly finned heatsinks on their B450 boards.

Also WTF is the Ryzen 5 2600 costing £170 here?? Its £140 to £150 at most other places I looked at!!

I know, just a little shopping around you can get a pretty cheap 2600 now.

 

[chrcoluk]

Come on guys. Not that hard lol...

LUL but it seems it is, as for a 9400f to 2600X value comparison you somehow added the wrong cpu

Personally because I think XFR is so damn good and you do get a better stock cooler as well, I think the cost of a 2600x is worth it over a 2600 imo. There is no guarantee a 2600 will reach 2600x clocks, silicon lottery and all.

 

[Journey]

It's hard to chose between the 2600 and 2600X, especially when you can now get the 2600 at £135 and the 2600X at £165, that's a big chunk of cash; £30 on a low-mid budget build you can go from a GTX 1660 to a 1660 Ti, or get a better board, and maybe faster RAM.

I've yet to have a bad 2600 in a build, just lucky as I had quite a few rubbish 2700's so I know that silicon lottery is not always on your side.

 

[chrcoluk]

I do agree % wise its a big chunk, its a fair debate, but the 2600X was used as it was used the actual chip to compare.

I have had so much bad luck with silicon lotteries tho I am paranoid because of it, e.g. my auros 1080ti cannot reach auros 1080ti extreme speeds and I even recently found out it cannot be stable on its OC mode that gigabyte designed for the card. So I could probably even RMA it if I wanted as its that bad of a loser chip.

My EVGA 970 FTW card was below kingpin clocks as another example (although also unstable on shipped clocks).

Sods law if I brought a 2600 instead of a 2600x, I bet it wouldnt achieve a 4.15ghz all core overclock which the 2600x will do as standard out of the box config. Never mind the 4.25ghz single/dual core speed. I couldnt even get 4.15ghz all core stable on my 2600x manually, I think XFR does more then just tune vcore, as an equal config set manually freezes up. Someone on reddit also rightly pointed out the better cooler on the 2600x could be the difference in a decision to buy after market cooler or not, and if it is then there goes a big chunk of the cost saving.

So in my opinion if you did a test on a 2600 you would have to "not" o/c it as thats luck based.

 

[tamzzy]

LUL but it seems it is, as for a 9400f to 2600X value comparison you somehow added the wrong cpu

2600 oc and 2600x oc is about the same.
Also at work so can't spec. Hence the quote from previous post
But I take your point.

 

[EsaT]

Expecting the first silicon from a new fab process to have an increase of 600Mhz over an older more mature fab

Mature phone/tablet CPU node...
Because of themselves being unable to develop working 14nm node, GloFo bought/licensed Samsung's mobile SOC node.
That's why voltage/clocks scaling is so bad for current Ryzens and getting over 4GHz needs lots of volts.

TSMC again has separate 7nm nodes for phone/tablet CPUs and High Performance parts.
And you can bet that Zen2s aren't made on mobile SOC production line...

I've yet to have a bad 2600 in a build, just lucky as I had quite a few rubbish 2700's so I know that silicon lottery is not always on your side.

Lot easier to find dies with six high clocking cores than eight high clocking cores.
So good quality full chips are likely heavily prioritized for 2700X and Threadripper, while 2700 gets mostly left overs.

 

[Double07]

Oops Sorry

 

[beany_bot]

I was thinking 4.7GHz or so, not that it matters given the expected improvements in other areas. They could be running at 2GHz for all I care so long as they deliver on performance, which the demo at the very least implies they should.

Exactly

 

[chrcoluk]

Yes this is why I try to always use the term "per core performance". As IPC may be different meaning 5ghz wont necessarily be the same as 5ghz on coffeelake.

I am hoping they start winning on IPC, and then something like a 4.5-4.6ghz chip might match the 4.8 on my 8600k.

Intel feel like they have hit their ceiling, whilst AMD are on that upward projectile.

 

[DragonQ]

Well technically Intel hit the IPC ceiling years ago, in fact it has regressed because of vulnerability patches.

 

[chrcoluk]

indeed

 

[Murphy]

Mature phone/tablet CPU node...
Because of themselves being unable to develop working 14nm node, GloFo bought/licensed Samsung's mobile SOC node.
That's why voltage/clocks scaling is so bad for current Ryzens and getting over 4GHz needs lots of volts.

TSMC again has separate 7nm nodes for phone/tablet CPUs and High Performance parts.
And you can bet that Zen2s aren't made on mobile SOC production line...

Yea but no, GloFlo bought the 14nm plant, process, and a number of 14nm IPs from IBM back in 2014. GloFlo then brought Samsung on-board and further developed the original IBM IP into a full FinFet node that started with low-power early, then a low-power performance, and then the two of them went of on their own development paths with GloFlo developing their own 12nm leading performance node based on the previous work with Samsung.

To say it's a phone/tablet CPU is highly misleading as firstly it didn't start out as that and secondly when GloFlo developed their own 12nm, that Zen+ is fabricated on, it moved away from being a low power node.

The original Ryzen used a node that focused on low power draw but Ryzen 2 didn't.

 

[humbug]

Yea but no, GloFlo bought the 14nm plant, process, and a number of 14nm IPs from IBM back in 2014. GloFlo then brought Samsung on-board and further developed the original IBM IP into a full FinFet node that started with low-power early, then a low-power performance, and then the two of them went of on their own development paths with GloFlo developing their own 12nm leading performance node based on the previous work with Samsung.

To say it's a phone/tablet CPU is highly misleading as firstly it didn't start out as that and secondly when GloFlo developed their own 12nm, that Zen+ is fabricated on, it moved away from being a low power node.

The original Ryzen used a node that focused on low power draw but Ryzen 2 didn't.

12nm is 14nm Tweaked, the difference is minimal, the Fin and Poly pitch are identical, they have slightly smaller cells on the same substraight which gives them about 1.2X density vs the Samsung 14nm it originated.

Other than that there really is no difference, the smaller cell sizes are slightly more efficient giving you about 10% higher clocks at the same volts but they are still based on the same 3.0Ghz Samsung 14nm node, so they are now more like 3.3Ghz.

 

[Murphy]

12nm is 14nm Tweaked, the difference is minimal, the Fin and Poly pitch are identical, they have slightly smaller cells on the same substraight which gives them about 1.2X density vs the Samsung 14nm it originated.

Other than that there really is no difference, the smaller cell sizes are slightly more efficient giving you about 10% higher clocks at the same volts but they are still based on the same 3.0Ghz Samsung 14nm node, so they are now more like 3.3Ghz.

Considering little is known about any of the Samsung nodes outside of the first gen 14nm (14LPP,LPC, LPU) IDK what you're basing that assessment on, besides there's a lot more that goes into making a node more power efficient or clock higher than simply the fin, poly pitch and standard cells sizes.

 

[humbug]

Considering little is known about any of the Samsung nodes outside of the first gen 14nm (14LPP,LPC, LPU) IDK what you're basing that assessment on, besides there's a lot more that goes into making a node more power efficient or clock higher than simply the fin, poly pitch and standard cells sizes.

Well hang on a minute, i've just given you specific aspects of SS 14nm vs 12nm, this gathered from reading around the net, do you not trust what i'm say? which is fair enough, is it the source that you want? it would take me some time to find them all again.

And i'm responding to your assertions which are far more blanket than mine, what do you base your assertions on?

 

[Murphy]

Well hang on a minute, i've just given you specific aspects of SS 14nm vs 12nm, this gathered from reading around the net, do you not trust what i'm say? which is fair enough, is it the source that you want? it would take me some time to find them all again.

And i'm responding to your assertions which are far more blanket than mine, what do you base your assertions on?

No i don't trust what you're saying, based on your contributions in another thread you tend to play fast and loose when it comes to information, facts, and evidences.

That's backed up by you seemingly being unaware that there are four generations of Samsung's 14nm node and not specifying what generation you're referring to when you say 12nm is 14nm tweaked.

And what assertion that i made are you referring to? Just throwing out a "blanket statement" that someone has made assertions is pointless unless you're willing to point out exactly what assertion you're taking issue with.

 

[humbug]

No i don't trust what you're saying, based on your contributions in another thread you tend to play fast and loose when it comes to information, facts, and evidences.

That's backed up by you seemingly being unaware that there are four generations of Samsung's 14nm node and not specifying what generation you're referring to when you say 12nm is 14nm tweaked.

And what assertion that i made are you referring to? Just throwing out a "blanket statement" that someone has made assertions is pointless unless you're willing to point out exactly what assertion you're taking issue with.

I'll just take your assertions as read then

 

[Murphy]

I'll just take your assertions as read then

Feel free to point out what assertions you have issue with, while you're at it how about you tell people what generation of 14nm is 12nm based on?

Or failing that as you're probably going to do your typical equivocation to avoid committing yourself to anything resembling an answer you could just read this that explains 12nm is more than a simple "tweak" of 14nm.

 

[humbug]

Feel free to point out what assertions you have issue with, while you're at it how about you tell people what generation of 14nm is 12nm based on?

You've just told me you don't trust anything i say, this has been the problem right from the start, you want me to go running round and round in circles proving my self to you for your amusement, you're trolling, again you've just told me you don't believe anything i say and yet you're still trying to push me into engaging with you.

Go amuse yourself else where, don't do it on my time.