Speculation: Ryzen 3000 series

[exquisitechar]

https://elchapuzasinformatico.com/2019/06/amd-ryzen-5-3600-x470-review/

 

[Markfw]

https://elchapuzasinformatico.com/2019/06/amd-ryzen-5-3600-x470-review/

Not sure I trust this review. Also, no overclocking means that the bios was not "fully compatible".

I still want to wait for real reviews.

 

[Glo.]

ElChapuzasInformatico has once again pulled the review days before the NDA's lift, and sales start. The same thing they have done when Zen 1 was launching, and their results were pretty much confirmed by later reviews, OR - launch day reviews have shown better performance, than ECI's review.

 

[exquisitechar]

Not sure I trust this review. Also, no overclocking means that the bios was not "fully compatible".

I still want to wait for real reviews.

Memory performance is bad too. Still, this is the floor for the 3600's performance, I guess.

 

[Glo.]

Memory performance is bad too. Still, this is the floor for the 3600's performance, I guess.

Pretty much this.

 

[Karnak]

Good first look, but too bad that they didn't get a X570 board in time. Looks like the actual X470 "Zen2 ready" BIOS are more to guarantee that it works. Looking forward for the reviews with X570 or proper BIOS for X470.

 

[Asterox]

Not sure I trust this review. Also, no overclocking means that the bios was not "fully compatible".

I still want to wait for real reviews.

Well at the moment, "overclocking is not important for this let say first real test".

It is legit or expected scores for Zen 2.Ryzen 5 2600/stock has 2700 in Cinebench R20 vs Ryzen 5 3600 with 3500 points.

 

[DrMrLordX]

Looking forward to this.

3 options for overclocking, eh? Interested in what's the max "safe" voltage, too.

If it's 1usmus, it's probably gonna be good.

 

[guachi]

Those leaked results from the Spanish site look quite good for the lowest priced part. At least, they look good to me: $200, 65W, and a provided cooler that can fit in any case.

Looks like it'll still be my pick for my new miniATX build.

 

[Glo.]

I was divided between R5 3600, or the hypothetical Ryzen 5 3500: 3.2 Base/4.0 Boost, 50W TDP CPU. But I may pull the trigger and go with R5 3600, B450 MoBo and Small Navi, when it will be released.

 

[Kenmitch]

With all the leaks, I'm now considering the 3900x. What I'm hoping is that the 3800x will have better binned chiplets and that the 3900x is less able to overclock, which would make me less inclined to spend more on the 3900x. I think its possible that with the disabled cores, they could actually clock higher though, given that the heat/surface area will be better. It feels so good to have options.

Lisa Su said the 3900x was a no compromise offering. What she meant was no clockspeed regression with higher core count parts. 3950x will most likely exceed 3800x in clocks.

 

[realibrad]

Lisa Su said the 3900x was a no compromise offering. What she meant was no clockspeed regression with higher core count parts. 3950x will most likely exceed 3800x in clocks.

Well, unless they have chiplets that are more than 8 cores, clearly there would be two defective chiplets. It could very well be that the defect is just in either one or two of the disabled cores, but, I doubt the 3900x will clock as high as the 3950x. If it does though, that is what I pull the trigger on. Then again, 2 weeks ago I thought the 3800x was my next cpu. Who knows lol.

 

[ubern00b]

Nice single core boost over 2700x, on what is likely a beta/1st Ryzen 2 BIOS for x470, I expect there will be further improvements/agesa revisions to improve performance further, can anyone clarify what RAM speed/ clock speed the benchmarks were running? I plan on getting a 3600 (most likely non-X) to replace my 1st gen 1600 so the performance improvement should be a lot more clear cut which is why I skipped Ryzen+ i'll be more than happy with 4.4/4.5Ghz OC

 

[Karnak]

The thing is that the 3950X and the 3900X will use different chiplets. The 3950X gets the chiplets with all 8 cores while the 3900X gets the chiplets where only 6 cores are active.

So they don't compete with each other, AMD can use the best chiplets with all 8 cores for the 3950X and the best chiplets with 6 cores active for the 3900X.

 

[Kedas]

If true it looks like the 9900K will still have a very small lead in gaming. (averaged)
But looks amazing otherwise.
Also this is probably without the windows patch to handle threads better and there is clearly some memory write issue that needs fixing or maybe it's just that test that is wrong since a copy is twice as fast as writing something.

 

[realibrad]

The thing is that the 3950X and the 3900X will use different chiplets. The 3950X gets the chiplets with all 8 cores while the 3900X gets the chiplets where only 6 cores are active.

So they don't compete with each other, AMD can use the best chiplets with all 8 cores for the 3950X and the best chiplets with 6 cores active for the 3900X.

That is the thing, which is that if they only make 8 core chiplets, and the 3900x is an 8 core chiplet with 2 cores disabled per chiplet, then the thermal density would be lower, which then could allow for higher overclocking. So, if the cores that are not disabled are still of quality, then that could be a better gaming chip given how many games still do not use anywhere close to 12 cores, nevermind 16.

 

[ubern00b]

If true it looks like the 9900K will still have a very small lead in gaming. (averaged)
But looks amazing otherwise.
Also this is probably without the windows patch to handle threads better and there is clearly some memory write issue that needs fixing or maybe it's just that test that is wrong since a copy is twice as fast as writing something.

This is 3600 mid-tier chip, there will be higher binned and clocked 8c/12c and 16c chips that should put the 9900k firmly in it's place in just about everything, gaming included, if this is what the lower end chips can provide. Also it's a beta/1st release bios for Ryzen 2 on x470, memory performance will improve on release bios or at least further agesa updates, that I'm sure of.

 

[Thunder 57]

If that review is true things are looking really good for AMD. Are those single threaded becnhmarks really running at 4.2GHz on the 2600 and 5GHz on the 9900k?? Impressive if true. I wish there was an x265 benchmark so we could see just how improved AVX is.

Gaming looks respectable considering the clock deficit. AMD could end up winning many of those too with higher end models.

 

[birdie]

Ryzen 5 3600 GeekBench 4 results under Windows 10 and Linux:

https://browser.geekbench.com/v4/cpu/compare/13631495?baseline=13629401



I'm really excited.

 

[jamescox]

A thing of beauty:

https://www.reddit.com/r/Amd/comments/bzcufm

That is a good image. I guess the x16 is the kind of pink circuits on the top left and bottom left. The DRAM controller must be along the bottom / bottom right. The other serdes are the x4 to the chipset, x4 for m.2, and then basically the equivalent of x8 used for SATA and USB.

I had been wondering if there was any chance that it has an extra IFOP present, but I don’t know where it would be routed given those connections. I had wondered about the possibility of an extra IFOP since the over 400 square mm IO die seems like a waste for the lower end ThreadRipper parts unless they have a lot of salvage with defective memory or defective IFIS links. I had thought about the possibility of just having an extra IFOP on the IO die. That would allow connecting 2 small, cheap IO die together to make a ThreadRipper level device limited to 4 cpu chiplets. It doesn’t look like there is any room for an extra IFOP in this image though. The IO die used in Epyc is really the size of a gpu by itself, so I guess ThreadRippers may be going up in price, comparatively speaking.

 

[Head1985]

https://elchapuzasinformatico.com/2019/06/amd-ryzen-5-3600-x470-review/

Probably fake.9900k numbers dont make sense and 6700k is way slower than it should be.
Fc5 benchmark 9900k average 129fps

When i tested it with my 6700k at 4.5ghz i get 128fps avg.That game dont scale with more than 4 cores.Also 6700k is not slower than 2700x in final fantasy.Maybe if we ignore 9900k and 6700k numbers and compare only 2700x vs 3600 we can have some indications of performance.

 

[lobz]

Probably fake.9900k numbers dont make sense and 6700k is way slower than it should be.
Fc5 benchmark 9900k average 129fps

When i tested it with my 6700k at 4.5ghz i get 128fps avg.That game dont scale with more than 4 cores.Also 6700k is not slower than 2700x in final fantasy.Maybe if we ignore 9900k and 6700k numbers and compare only 2700x vs 3600 we can have some indications of performance.

6700K stock and 6700K @ 4.5 ACT are worlds apart, especially if you have faster RAM than what they have used. Anyway, you still have some 2 weeks to keep us all depressed

 

[Markfw]

Probably fake.9900k numbers dont make sense and 6700k is way slower than it should be.
Fc5 benchmark 9900k average 129fps

When i tested it with my 6700k at 4.5ghz i get 128fps avg.That game dont scale with more than 4 cores.Also 6700k is not slower than 2700x in final fantasy.Maybe if we ignore 9900k and 6700k numbers and compare only 2700x vs 3600 we can have some indications of performance.

Even comparing those 2 does not make sense. A 12 core 2.9 ghz base clock ? against a 8 core 3.9 ghz ? HEDT vs Desktop ? And why are we looking at Intel benchmarks on a Ryzen 3000 speculation thread ?

 

[ClockHound]

Even comparing those 2 does not make sense. A 12 core 2.9 ghz base clock ? against a 8 core 3.9 ghz ? HEDT vs Desktop ? And why are we looking at Intel benchmarks on a Ryzen 3000 speculation thread ?

Because... the Inteligentsia are nervous?

 

[realibrad]

Because... the Inteligentsia are nervous?

Clearly this posts wins.