AMD Demos 7nm Ryzen 3000 processors (and kicks a bit of ass)

2600x is 4-5% faster than 8700k clock to clock.

2700x vs 9900k , both @ 4.ghz, it is 1805 vs 1702.

So the outlook doesn't look good at all and it suggest minimal IPC gains and almost performance regression.

 

Considering the CPU used was a 75W TDP part with 8c/16t at an unknown frequency vs Intel's 125W 9900K which moonlights as 95W TDP part. The CPU that lines up with that is the 3600/3600X from the leaks. That's mid-range 3xxx vs 9900K. So yeah not too worried here right now. CPU shot clearly had space for a second chiplet on there indicating the 12 and 16 core parts are a strong likelihood too.

 

Until 3xxx hits HHs table we can't say much, however reaching 5ghz,even at the same ipc, which we know won't be the case, is great. Not to mention that, as many people above said, engineering samples are worse and they lack proper bios.

 

https://twitter.com/IanCutress/status/1083099086880952320

Ian Cutress just non-confirmed the existence of a 12 and 16 core part coming...

 

https://twitter.com/markhachman/status/1083097464624693248

 

Nice... It almost beat mye average 9900k

 

So was this CPU and R7 part or an R5??? If this turns out to be the 3600x version leaked before then this could be a great mainstream CPU.

 

Your 9900K is at 5.3GHz...

I highly doubt the 9900K used was anything other than stock so 4.7GHz unless the board did it's own thing and allowed it more than usual power, which it does look like it did. So the normal out of the box issue with all but one board OC'ing the 9900K from BIOS.

 

In case you weren't aware, Intel's own slides showed their first-gen 10nm being slower than their current 14nm++. The only major improvement is in power efficiency, which is what was shown with 7nm. Seems about right.

 

Probably the R5, gonna be real value on this R5's

 

Sorry i don't troll enough for you to be viewed as useful

Actually, i'm not sorry at all.

Though i find your comment interesting as you have for the most part been replying to the useless, bashing, negative based off of no information posts in this thread, yet i call it out as it is and that's your response?

Very interesting indeed.

 

lmao, at people expecting AMD to layout out their full CPU/GPU hand in January. I have no doubt they will be releasing scrumptious bits of more info in the months to come.... looks like AMD will own 2019.

 

damn i was expecting more .... this answered nothing other than give me the impression that 8core ryzen 3xxx are the r5 ....seems that another chiplet can fit in there ... will see !

 

To be honest it was not Jim's fault that some of the stuff was not announced on CES. Because AMD themselves backtracked and changed their minds several times.
I bet that at first Navi was to be mentioned at CES, but something was not right with the performance or yields on 7nm new uarch, then their now "fired" president had an idea to launch Vega 20 (II) with GTX1080Ti like performance and price around 750 dollars... hmm.
Does that sound familiar? AdoredTV's leaker just had old info to leak. But regarding CPU performance parts- he got that right! Yes, the 16C cpu was not presented, but we all can see from the cpu die placments that there is room for one more 8C die. Also Lisa Su herself confirmed in an interviu right after the keynote, that there will be more cores for Ryzen. And chiplet design was also confirmed by Jim. Altough for baby Ryzen 3000chips, it may not be needed.

 

If they had a better hand we would have seen it. AMD are not Nvidia or intel they are too far behind on market share so if they have anything good or ready they will present it. This only shows that CES was a fee months too eraly for ryzen

 

Considering they have an engineering sample with the same cores matching Intel is actually pretty nice. I'm sure they will eek out some more performance by June be it clocks since apparently they certainly have TDP headroom or by a final silicon spin with some tweaks. However what they showed at CES was there mid range part. You do realize they all but confirmed 2 chiplet Ryzens to Anandtech and since the chip has that funky design where the chiplet is set in the corner well you kind of see it coming. So I assume they will have ryzen part that will score better than 2x what a 9900K can pull in Cinebench.

 

True however having better when a product is 7 months away from launch and making better chips by then are two different things. The fact they showed of an early engineering sample was great for them being able to launch this summer. I can see why they would start off with a single chiplet design for validation simplicity at first.

 

I just called out that both your posts in this and the other thread had no value at all. Didn't contribute and only complained. Not even on the topic, just on the people posting. I didn't see the point in either of them.

 

I thing they are holding their cards ( AMD ) very close to their chest and not showing more than they need to show at this moment ,paper launch , time will tell ... But the empty space on the die i bet 100% it is for another chiplet 8 cores meaning a 16 core 32 threads cannot be dismiss. All in all good vibes from AMD and the Ryzen 3000 cpus

 

Well idk this is an engineering sample so clock speeds will be significantly lower than retail samples and the Intel score is inline with the 9900K running all cores at 4.7GHz. For all we know the Ryzen was running all cores at 3GHz, i am guessing maybe around 4GHz however is realistic, but thats significantly higher than what we are used to seeing with engineering samples from either Intel or AMD.

We wish 125w, all core 4.7GHz is around 150w for the 9900K

22w for motherboard, 3w for SSD, 5w for graphics card idle = 30w then the 150w from the CPU

I agree they are holding them close they don't want to set expectations to high like they have done in the past because it always bites them in the ass and we are still months away from launch no point in giving your competitor many months to prepare.

Remember someone came out and said the new Ryzen had 25-30% IPC and AMD officially commented it wasn't and was around 12%, but it turned out the 30% increase was only for AVX workloads so it was very workload specific. That tells me AMD is not wanting to set expectations to high. They maybe trying to set them low so people are amazed like the original Ryzen they said it was 40% IPC over pilldriver but tuned out to be over 50%.

So we know the IPC gains of the Ryzen but we have no idea what clock speeds they will be hitting. Well i say no idea there are leaks of clock speeds similar to that of Intels. Although Intel are hitting them clock speeds on 14nm while it took AMD to go to 7nm to hit them, but the IPC increase will be what separates it in the end.

*** Guess ***
With the 12% IPC then this would appear the engineering sample is running at 4.2GHz Vs the i9 9900K 4.7GHz. However it may have a lower clock because SMT for Ryzen is more efficent than Intels hyper threading.

Still 4.2GHz for an engineering sample does seem quite high to me.

 

Using less power is only "pretty damn sweet" if they can translate that into another product that uses that power to give me performance. Preferably without just shitting out more cores.
Otherwise, less power is just "ok" in my book.