AMD Computex 2019 Livestream

Well, what I meant was, they claim to have improved IPC 15% over Zen+, and the numbers they are claiming put them at a 12.5% IPC advantage over Skylake.

Were they really that close to Skylake before?

The numbers don't make sense to me.


With 3200c14, yes they were ~ to Skylake....Lock the clocks and run the tests yourself and you will see. It's been mentioned in MANY MANY TIMES but th IDF ignores it.
 
Don't get your hopes up. If their GPUs are anything to go by, it'll be closer to maxed out from the factory than not.


Those of us with WC'd VIIs are hitting sustained clocks of 22-2300Mhz on the core and 1200+ on the HBM.

Explain to me again how a 25%+ OC headroom is "maxed from the factory"?

The only GPU that has been maxxed has been the Fury X, and that is because AMD took a 20nm design and hacked it together to work with 28nm when the 20nm fab process fell flat on it's fast for anything other then DRAM and NAND.
 
With 3200c14, yes they were ~ to Skylake....Lock the clocks and run the tests yourself and you will see. It's been mentioned in MANY MANY TIMES but th IDF ignores it.

Well, you can mention it, but to do that you need to run a decent board and spend more on memory- and Skylake CPUs just aren't that picky, and can still clock higher. Not really an IDF thing, just generally not worth the effort for less performance, unless you have an AMD tattoo or something.

But if Ryzen 3 is hitting that performance without being so picky (which we should expect), then cool.
 
Well, what I meant was, they claim to have improved IPC 15% over Zen+, and the numbers they are claiming put them at a 12.5% IPC advantage over Skylake.

Were they really that close to Skylake before?

The numbers don't make sense to me.

As marketing numbers they make sense :D

[in reality, I expect we're seeing within 5% clock-for-clock all variables accounted for]
 
Based on the slides:

3700X looks like an 8C/16T
3800X looks like an upclocked 8C/16T
3900X looks like a 12C/24T

Edit: I'm dumb, didn't read previous posts wherein the same was already said about 12 core and not 16
 
Based on the slides:

3700X looks like an 8C/16T
3800X looks like an upclocked 8C/16T
3900X looks like a 12C/24T

3900x is 12/24 I believe. Maybe they'll do a 3990 as the 16/32. Or the tdp for 16/32 is too high for x370/470 so they will launch it later as a x570 only chip? On my lunch break so still trying to catch up on computex stuff.

Lol just noticed you edited it right as I hit reply so I'm posting this anyways after spending all this time typing on my phone. :p
 
3900x is 12/24 I believe. Maybe they'll do a 3990 as the 16/32. Or the tdp for 16/32 is too high for x370/470 so they will launch it later as a x570 only chip? On my lunch break so still trying to catch up on computex stuff.

Lol just noticed you edited it right as I hit reply so I'm posting this anyways after spending all this time typing on my phone. :p

Yeah, I did a quick read on the previous pages. Also just saw the anandtech article on the full release specs.

Looks like the 3700X will be the OC winner unless these are super binned to begin with. With a 6+6 core, the 3900X may also be able to OC quite a bit if there's some headroom on each chiplet.
 
so if u get a golden sample i guess u can hit 5 ghz on the 8/16, and avg. 2-300 mhz.
 
I'm happy for them too, and I'll definitely be reading what they review.

Too bad they couldn't link to here for forums though.

Creating yet another hardware forum means more fragmentation of users.
Well, that is something they never asked, and they will find soon that forums are a loss leader by far, and pretty much a failing business model today from a standpoint of building anew. To build a forum that actually has any kind of revenue model attached to it, you have to turn it into something that is not taken kindly to around these parts. :) Do you wonder why there has ONLY been one banner advert here for decads?

But yet, I do not deny them posting their URLs here trying to pull traffic away from this community. And interestingly enough, no talk about HardOCP leaving/donating all the testing equipment with the editors so that they could move forward with their own endeavors.
 
Last edited:
Gotta say, 3700X looks pretty awesome for an ITX build. Will wait for review but looking likely I will get that once I get my Sentry 2.0 case later this year.
 
O hell yeah :)
And i also understand that " You will definitely see more Threadrippers from us" which a like though i am not going to get a new one just yet, but maybe in a years time when some monster reach a good price / perf level my 12 core part will get replaced.

24/48 cores sound good to my ears.
 
AMD labeled the 9920x as 'Cascade Lake', correct?

I thought Cascade Lake (14nm++) was not released yet as the 9000x hedt cpus were just 7000x hedt cpus with solder (14nm) - both still SkylakeX.
 
AMD labeled the 9920x as 'Cascade Lake', correct?

I thought Cascade Lake (14nm++) was not released yet as the 9000x hedt cpus were just 7000x hedt cpus with solder (14nm) - both still SkylakeX.

yeah 99xx ended up being called Basin Falls or some shit.. until i saw your post and actually looked it up i always thought 99xx was Cascade Lake.. there are some cascade lake processors out there on the server/small business side though. but honestly it's just another refresh of a refresh of skylake..
 
From the AMD slide deck

View attachment 163547

We will have to wait until July to verify them in real reviews, but it was exciting to watch live!

Also, Intel' s Emergency Edition 9900KS is going to close the gap, but not exceed the Ryzen 9's sizeable lead. I see a price cut from them in the near future.

AMD can just save the 16-core for next year's Zen 7nm+ refresh, where it can counter Comet Lake..
Thanks for sharing this slide. It shows the reason they didn't compare the 3700x to the 9900x is because the 9900k is about 5% faster.
 
I think I will wait for the itx 570 or 550 boards to hit stores and grab one with a ryzen 1600 for $80 from microcenter. Then when the 16 core launches or 12 core drops in price, upgrade. I don't think performance per dollar of ryzen 1600 for 80 bucks can be beat right now.
 
I think I will wait for the itx 570 or 550 boards to hit stores and grab one with a ryzen 1600 for $80 from microcenter. Then when the 16 core launches or 12 core drops in price, upgrade. I don't think performance per dollar of ryzen 1600 for 80 bucks can be beat right now.

as some one who owns and uses both a 1600 and a 2600x, i highly recommend either the 2600 or 2600x over the 1600.. the single threaded performance and the updated boost algorithm on zen+ is so much nicer.
 
as some one who owns and uses both a 1600 and a 2600x, i highly recommend either the 2600 or 2600x over the 1600.. the single threaded performance and the updated boost algorithm on zen+ is so much nicer.


Agreed. I have a 1600 that hits 4.1 AC turbo, and the 2700 @ 4.35Ghz was a welcome upgrade. You will be EXTREMELY LUCKY to find a 1600 that can do 4Ghz, let alone 4.1Ghz stable AC without 1.5V and a ton of voltage.

. The newest AGEA code allows for much better ram, so you should be able yo run pretty much any rated kit at 3000 with nice timings, and most with 3200, but with the 2xx series that is basically guaranteed.

On the other hand, if you buy a top both OC'ING board, the 1600 should be a latest batch sku, which will clock to he ceiling I mentioned. If it doesn't, you still have a decent 6c/12t and have a top quality OC'ing board ready for your Zen2 upgrade. So you can't lose either way!

I actually sold my 2700 off since I knew that the Zen2 launch was going to kill resale, and AMD clearing the channel now is offering sorm killer deals.
 
Well, you can mention it, but to do that you need to run a decent board and spend more on memory- and Skylake CPUs just aren't that picky, and can still clock higher. Not really an IDF thing, just generally not worth the effort for less performance, unless you have an AMD tattoo or something.

But if Ryzen 3 is hitting that performance without being so picky (which we should expect), then cool.

You are shifting the Goal Posts...He asked if AMD was already at the single core IPC of Skylake, and I responded with solid facts. Give them 3200c14 ram and lock them to the same clock as th equivalent SL SKU, and SURPRISE, they have the same performance.

I did not say you needed to have this or that etc, and your facts are a bit twisted. A cheap b350 can run 3200c14 just fine thanks to the IMC. You may not hit the max clock speed, but we aren't comparing processes, we are comparing unarch.

The whole AMD "tattoo" is ironic since you defend Intel so much you may be speaking from experience. AMD has made me a TON of money (both their stock rise under Su's excellent leadership) and with crypto mining going back to the Litecoin days.
AMD GPUs allow for the best all around gaming experience via their excellent FreeSync support for the best gaming monitor that can double as a large 4k tv via the Samsung 8k series SKU(s).

So Yes I love them for that, but I've also run nothing but Intel CPUs since I traded my [email protected] for x58 970@ 4.2Ghz and then z77 with a 5.2Ghz 3770k and then jumped to AMD.

I use what works the BEST for me, and the reality is that my AMD 1600 paid for itself, and my 2700 and continues to earn.
 
You are shifting the Goal Posts...He asked if AMD was already at the single core IPC of Skylake, and I responded with solid facts. Give them 3200c14 ram and lock them to the same clock as th equivalent SL SKU, and SURPRISE, they have the same performance.

Point being, Ryzen 2 can get close per clock, if you then invest additional time and money, and then you're still behind on clockspeed and thus single-core performance, which is what we are talking about.

The whole AMD "tattoo" is ironic since you defend Intel so much you may be speaking from experience.

My experience has been running both (and others) over the past two decades, and witnessing achievements and missteps from both. As much as the cheerleaders get excited, in context I still see AMD provably trying to catch up to an Intel uarch that is over half a decade old.

AMD has made me a TON of money (both their stock rise under Su's excellent leadership) and with crypto mining going back to the Litecoin days.

At least you're honest about your bias and niche use cases eventually. That's fine that it works best for you, but I look at a broader set of workloads.
 
I think I will wait for the itx 570 or 550 boards to hit stores and grab one with a ryzen 1600 for $80 from microcenter. Then when the 16 core launches or 12 core drops in price, upgrade. I don't think performance per dollar of ryzen 1600 for 80 bucks can be beat right now.

Bad news, X570 boards aren't compatible with 1st gen Ryzen chips. Puts a wrench in this plan, as I had been considering something similar.
 
This was pretty funny. :)

t5bd68xfbs031.jpg
 
Point being, Ryzen 2 can get close per clock, if you then invest additional time and money, and then you're still behind on clockspeed and thus single-core performance, which is what we are talking about.



My experience has been running both (and others) over the past two decades, and witnessing achievements and missteps from both. As much as the cheerleaders get excited, in context I still see AMD provably trying to catch up to an Intel uarch that is over half a decade old.



At least you're honest about your bias and niche use cases eventually. That's fine that it works best for you, but I look at a broader set of workloads.


While the topic may be about "single core clockspeed", you are using that to defend your opinion of me being wrong, in response to a specific question he asked about IPC. He did not make any mrntion of single core clock speed.

As far as me being biased, I have owned almost every single CPU from both camps (and even 2 Cyrix units before Intel's bullshit bankrupted them like they then tried to do to AMD, but that is a different story) going back to my first purchased Pentium 100 which replaced a gifted hand me down 486 from my uncle.

I have spent a TON of money on this hobby, abd while I gave my personal reasons, I have built 40+ gaming rigs in the last 7-8 years, and they all used Intel CPUs aside from 3 or 4 that used heavily OC'd FX units due to a budget constraint. A large portion also used Nvidia GPU(s), so you can't ding me there either.


You like to stir the pot a little harder then you want to admit, in both CPU AND GPU threads. The fact is that AMD offers things the others do not, and vise versa.

Back on topic, if Intel were not worried, we would not be seeing specially binned, SOLDERED (now it's not a problem all of a sudden) 9900ks with a 5Ghz AC turbo...

I think with the same cooling you will need to tame that new 9900k, call it a higher end 240mm AIO, you will see the AMD CPU outperforming it in a lot of things with some ties and slight wins to Intel due to complier optimizations.

This with the AMD CPU being clocked 400Mhz slower. Even if AMD ONLY LEFT 100Mhz of OC headroom (again, just like Intel outside of true platinum samples). That means AMD will have surpassed Intel on IPC.

We will not know if this holds true until we get parts in our hands, but AMD already did it in their brief demo back in Jan, all while using much lower power then the normal 9900k. You don't think those been special edition verisons will pull more. Power? I do.

Before your buddy and resident Intel/AMD expert juanrga wakes up, I'm aware that was a Cinebench test and it's oh so biased toward AMD.
 
While the topic may be about "single core clockspeed", you are using that to defend your opinion of me being wrong

That's your perception- I didn't say you were wrong, I supported your claim, but noted the lack of applicability.

You like to stir the pot a little harder then you want to admit

I like to deal in fact. Those with strong feelings on the matters at hand tend to get offended.

I'm aware that was a Cinebench test and it's oh so biased toward AMD.

Fact is, the test is an outlier, and it doesn't really enlighten us to average performance or to potential weaknesses.

We will not know if this holds true until we get parts in our hands

We know what AMD has released for marketing purposes. With respect to IPC, they used a CPU with half of the threads for comparison.

AMD has more than earned the community's skepticism here.
 
Fact is, the test is an outlier, and it doesn't really enlighten us to average performance or to potential weaknesses.

Rendering is obviously not the same as gaming and other typical desktop loads, but it has been a good benchmark to illustrate singles threaded performance. There aren't really any other ones out there. That's why it is popular.

I find it amusing that people are calling it an "AMD Biased Benchmark" considering it was accused of being an "Intel Biased Benchmark" during the Bulldozer era not so long ago :p


We know what AMD has released for marketing purposes. With respect to IPC, they used a CPU with half of the threads for comparison.

AMD has more than earned the community's skepticism here.

Hobestly, the thread count concern is completely irrelevant when discussing IPC. That was the important part.

And if they want to show off the fact that they provide more threads at the same price, I have no qualms with that.

I'm all for dollar to dollar comparisons.
 
Rendering is obviously not the same as gaming and other typical desktop loads, but it has been a good benchmark to illustrate singles threaded performance. There aren't really any other ones out there. That's why it is popular.

I find it amusing that people are calling it an "AMD Biased Benchmark" considering it was accused of being an "Intel Biased Benchmark" during the Bulldozer era not so long ago :p

Bulldozer was... broken. Comparing Bulldozer to basically anything is like comparing Netburst to basically anything. And the challenge with CB15 is that rendering loads don't really show off the broader landscape of single-core performance; what CB15 is really good for is for providing a loose FP throughput result that can be compared across generations.

Hobestly, the thread count concern is completely irrelevant when discussing IPC. That was the important part.

Looking at it more closely, I'm tending to agree; however, using a dissimilar CPU still doesn't inspire confidence here.

And if they want to show off the fact that they provide more threads at the same price, I have no qualms with that.

I'm all for dollar to dollar comparisons.

Going back to this whole release being marketing copy from AMD, it should be noted that any changes in Intel's pricing would affect their comparisons. We know that and AMD knows that, so leaving the 9900K out when comparing eight-core CPUs does seem disingenuous on their part. They've been plenty willing to compare their parts to more expensive Intel parts on the very same marketing slides.
 
If they're pitting the 3800X against the 9900K, what else does Intel have that they could pit the 3700X against? It's not like Intel has hyper-threaded octa-core between the two.
 
If they're pitting the 3800X against the 9900K, what else does Intel have that they could pit the 3700X against? It's not like Intel has hyper-threaded octa-core between the two.

Well, that's based on the current Intel lineup. Intel could have a '9900-non-k' tomorrow, and they could drop prices a bit too.
 
So AMD was supposed to compare their new product to products Intel might have tommorow?

More, comparing to Intel's technology as directly as possible as well as comparing to price. Remember that AMD still has to get the channel supplied, and we can't buy these things yet- they've given Intel plenty of time to respond if Intel hasn't already gotten that ready.

So this is marketing. It's not something that we can reliably predict what will be the best price / performance etc. come retail availability as announced.
 
Back
Top