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

[easyrider]

Link to proof that your 9900K is 24/7 stable in the first place ? You have been spamming this thread so much i really can't be arsed to go back through your posts to find it.

Fail

 

[eddiew]

Tell you what though, 1.2vcore for 5ghz?

 

[smilingcrow]

It's almost as if they are so confident they are drip feeding this in an underwhelming way.
The power consumption bodes very well for Server where it is very significant.
The release date is a shock which means they don't know the clock speeds yet so maybe that's also why they are playing it cool.
Cool as in only talking about 8C chips as without more than that many will feel betrayed!

 

[DragonQ]

what vcore was being used in the test on the 9900k ?

Seems high for a stock 4.7 ghz 9900k

I’ll check that bit of the video again and compare power usage with my 5ghz 9900k

They said stock so that should mean stock clocks and voltages. Seems fair to me, the Ryzen 3 chip isn't exactly going to be overclocked or undervolted is it, it's not even final clocks yet. It's like Vega vs Pascal. At stock Vega uses a lot more power. It can be aggressively undervolted to mitigate this, but then Pascal can also be overclocked/undervolted so the only possible fair tests are stock or both reasonably tweaked.

Anyway, let's get down to what all this means. Whether it's IPC or clocks that are putting Ryzen 3 on par with the i9-9900K is pretty irrelevant. It means that literally the only advantage the latter has is possibly more overclocking headroom. Let's look at two scenarios:

Best case scenario for Intel
The demo we saw is a maxed out stock Ryzen 3 that has no overclocking headroom at retail, which'd mean Intel has a ~6% lead when both chips are maxed out. That's really not very much and comes at the expense of a load more power usage, since a stock i9-9900K is already using nearly double the power of the demoed Ryzen 3 and the last 300 MHz takes a comparatively large amount of juice.

The AMD chip would also surely be considerably cheaper. Again, best case scenario so let's say the demoed chip is a Ryzen 3700X at 5.0 GHz. It can't realistically be over $400 because that would result in a lower-clocked sibling (R7 3700 at say 4.6 GHz) at around the $350 mark, which would not be a significant improvement on the even cheaper R7 2700X (except for PCIe 4.0). So I think $400 is pretty much the maximum this chip could be, and that's still a lot cheaper than Intel's competition.

Worst case scenario for Intel
The demo we saw is a relatively low clocked Ryzen 3, essentially clocked exactly where it needed to be to match the i9, let's say 4.5 GHz. This would give AMD a healthy IPC advantage (in multithreaded Cinebench anyway), a huge power advantage, and potential for higher clocks too. It'd also mean an overclocked version would beat an overclocked i9-9900K, likely while using less power and costing less. Oh, and there might be 12 and 16 core parts too.


Honestly, I think the truth is somewhere in the middle but due to the low power usage of the chip I'd be amazed if it's actually clocked that high. The only real question mark in my mind is how much headroom there is left in the clocks. If it's 65 W at 4.5 GHz, surely it can get close to 5 GHz.

 

[lltfdaniel]

I am gonna go into my corner now.

 

[TrixX]

They said stock so that should mean stock clocks and voltages. Seems fair to me, the Ryzen 3 chip isn't exactly going to be overclocked or undervolted is it, it's not even final clocks yet. It's like Vega vs Pascal. At stock Vega uses a lot more power. It can be aggressively undervolted to mitigate this, but then Pascal can also be overclocked/undervolted so the only possible fair tests are stock or both reasonably tweaked.

Anyway, let's get down to what all this means. Whether it's IPC or clocks that are putting Ryzen 3 on par with the i9-9900K is pretty irrelevant. It means that literally the only advantage the latter has is possibly more overclocking headroom. Let's look at two scenarios:

Best case scenario for Intel
The demo we saw is a maxed out stock Ryzen 3 that has no overclocking headroom at retail, which'd mean Intel has a ~6% lead when both chips are maxed out. That's really not very much and comes at the expense of a load more power usage, since a stock i9-9900K is already using nearly double the power of the demoed Ryzen 3 and the last 300 MHz takes a comparatively large amount of juice.

The AMD chip would also surely be considerably cheaper. Again, best case scenario so let's say the demoed chip is a Ryzen 3700X at 5.0 GHz. It can't realistically be over $400 because that would result in a lower-clocked sibling (R7 3700 at say 4.6 GHz) at around the $350 mark, which would not be a significant improvement on the even cheaper R7 2700X (except for PCIe 4.0). So I think $400 is pretty much the maximum this chip could be, and that's still a lot cheaper than Intel's competition.

Worst case scenario for Intel
The demo we saw is a relatively low clocked Ryzen 3, essentially clocked exactly where it needed to be to match the i9, let's say 4.5 GHz. This would give AMD a healthy IPC advantage (in multithreaded Cinebench anyway), a huge power advantage, and potential for higher clocks too. It'd also mean an overclocked version would beat an overclocked i9-9900K, likely while using less power and costing less. Oh, and there might be 12 and 16 core parts too.


Honestly, I think the truth is somewhere in the middle but due to the low power usage of the chip I'd be amazed if it's actually clocked that high. The only real question mark in my mind is how much headroom there is left in the clocks. If it's 65 W at 4.5 GHz, surely it can get close to 5 GHz.

QFT

At least there's someone else in this thread with his head screwed on

 

[r22snapper]

Beat me to it journey. I think the insinuation that AMD would have boosted the vcore of the 9900 is pretty funny tbh

 

[Jokester]

The release date is a shock which means they don't know the clock speeds yet so maybe that's also why they are playing it cool.

This for me is the only disappointment (the release date).

 

[d_brennen]

Matching a 9900K, drawing less power and presumably with a mid stack product (if the rumours are correct) is pretty handily besting Intel. The lack of info is for effect, to get the likes of us talking about it. It looks quite a lot like Intel got handed their ass back to them

 

[Stephanie Peterson]

Best case scenario for Intel
....
Worst case scenario for Intel
....
Honestly, I think the truth is somewhere in the middle but due to the low power usage of the chip I'd be amazed if it's actually clocked that high. The only real question mark in my mind is how much headroom there is left in the clocks. If it's 65 W at 4.5 GHz, surely it can get close to 5 GHz.

My honest view is it is the worst case scenario for Intel
The zen chip was an early ES ryzen 5 with 8c/16t running at either its stock base+boost or more likely a little bit less as this is how these things usually go.
The intel chip was a fully stock 9900k, out of box no fancy stuff just right out of box.

The likely very cheep mid range ryzen 5 was practically toe to toe with the stupidly expensive 9900k, once released it may be slightly quicker.
OC wise, well thats all up in the air - however while the 9900 can do a decent turn of Mhzs (at great cost to power and heat) the zen chip might not reach tat far as the older ones struggled with much more than a 150mhz sometimes.. Then again this is a brand new performance orentatied process and this may be BS
What ever way i do suspect that the dirt cheep (i hope) ryzen 5 (x or not) will be either just as fast or slightly faster when maxed out than the 9900k... and use a lot ...LOT less power when doing it.

Thats my view, i think intel is in for a severe ... severe beating in a few months..

 

[Doobedoo]

QFT

At least there's someone else in this thread with his head screwed on

That’s essentially what I’ve been trying to say except DragonQ put it much better .

 

[Vince]

I think bringing this back it was pretty solid showing from AMD today, we got a decent look at competitive products across the lineup, Radeon 7 looks decent enough and compelling for some and a sneak peak at Ryzen 3000 looks perhaps better than we should have been expecting. Claims that the silicon came back "better than expected" seem to be cemented in truth imo, in reality who expects generation on generation bumps like this in tech today?

I think they showed just enough which I think is actually pretty clever, they showed that they can match or beat Intels best mainstream and heavily cherry picked part core for core, thread for thread and most importantly for significantly less power draw, something that they simply haven't been able to do for the last 13 years. I think more than anything this is a millestone for them and validation on investments and critical thinking which happened many years ago.

 

[Potatowithearsontheside]

Bloody hell that's fast..............................it even manged to give you the score before the CB run finished

Not to mention that his score was 2070, but meh lol.
Zoom is a fine thing, especially when the result also says "Your Score" lol
Admittedly, he's had his quoted score in the past.

 

[Scougar]

This for me is the only disappointment (the release date).

Same. I must admit I was hoping it was going to be sooner.

 

[ivu]

Shame about the release date. I was planning to move from laptop to desktop but was wishing for something around march-early april at most ;/
What to do...what to do...

 

[Potatowithearsontheside]

So is that stock? i dont know what the stock 9900k CB multi score is? if that stock then fine.

If thats got MCE and all sorts of other fancy little bits of bios crap turned on then it rubbish.

With the 9900K it's a little unfair to use the word stock; out of the box plug and play is probably more accurate - all top-tier mobos for the 9900K have default PL2 set to 210w, and tau (maximum amount of time that turbo boost can be active, effectively) set to max (unlimited), in other words they default overclock (like XFR2 does for Ryzen in fairness) and you are more likely to run into thermal throttling (or some other protective measure) than you are to be power starved.
A 9900K at a strict 95w TDP (which is only possible on an Asus Maximus Hero XI, so PL2 of 1.25x PL1), typically hits 4.7GHz ACT for 24-28 seconds before dropping to a steady 4.2GHz under full load, which it can then maintain 24/7 if it has to.
The CB 15 MT is too short a run really for even a strict 95w TDP to impact the 9900K score; it'd still be averaging at least 4.5GHz for the entire run. I mean, my own figures do suggest that this may actually have been how the 9900K was configured, since it was on an Asus board, but the power draw figures seem to rule it out.

 

[Journey]

The CB 15 MT is too short a run really for even a strict 95w TDP to impact the 9900K score; it'd still be averaging at least 4.5GHz for the entire run. I mean, my own figures do suggest that this may actually have been how the 9900K was configured, since it was on an Asus board, but the power draw figures seem to rule it out.

The 9900K would have been running in stock 95w TDP mode, only allowing PL2 boost as I said earlier, the total system draw for the 9900K system is circa 180w, and the PL2 allowance is up to around 119w. This would have allowed an all core boost of 4.7GHz on the 9900K for the entire duration of the test.

 

[humbug]

Yeah power was ok but it all depends on clock speed....

No confirmation that Ryzen 2 beats intel in IPC at all.

Who cares? if they haven't beaten Intel on IPC the thing is running at at least 5Ghz, its an 8 core 16 thread chip that beat the 8 core 16 thread 9900K.

what is more likely, higher IPC or higher clock speed, and what would it matter? ITS FASTER ITS BETTER.

 

[Potatowithearsontheside]

I get that, but I doubt it maintained the 4.7GHz for the entire run; Gamers Nexus specifically tested this after the power draw saga after its release, and they found that it dropped to 4.2GHz after 24 seconds.
The Stilt, a highly respected tech guy on Anandtech, had previously referred to a 28s ACT limit, hence why I quoted 24-28s myself.
How long was the CB run?

 

[Journey]

I get that, but I doubt it maintained the 4.7GHz for the entire run; Gamers Nexus specifically tested this after the power draw saga after its release, and they found that it dropped to 4.2GHz after 24 seconds.
The Stilt, a highly respected tech guy on Anandtech, had previously referred to a 28s ACT limit, hence why I quoted 24-28s myself.
How long was the CB run?

CB run was exactly, 22 secs.