Thursday, July 13th 2017

AMD CEO Talks Ryzen Threadripper and Ryzen 3 Series in Latest Company Video

In a video presentation posted on the company's official YouTube channel, AMD CEO Lisa Su talked at length about the two new lines of Ryzen desktop processors the company plans to launch later this month. This includes the Ryzen Threadripper HEDT socket TR4 processor at the higher-end of the lineup, and the new Ryzen 3 series socket AM4 processors at the lower-end. AMD is announcing market-availability of two SKUs for each of the two brands. To begin with, AMD will launch two quad-core SKUs in the Ryzen 3 series, beginning with the Ryzen 3 1200 and the Ryzen 3 1300X. Both of these are quad-core parts which lack SMT, leaving them with just four threads. AMD is expected to price them on par with Intel's dual-core "Kaby Lake" Core i3 SKUs.

The Ryzen 3 1200 is clocked at 3.10 GHz, with 3.40 GHz boost, the 1300X is clocked higher, at 3.50 GHz, with 3.70 GHz boost, and XFR (extended frequency range) enabling higher clocks depending on the efficacy of your cooling. Both parts will be available worldwide on July 27. The Ryzen Threadripper HEDT processor lineup is designed to take Intel's Core X series head-on, and will launch with two SKUs, initially. This includes the 12-core Ryzen Threadripper 1920X, and the 16-core Ryzen Threadripper 1950X. Both parts further feature SMT and XFR. The 12-core/24-thread 1920X features clock speeds of 3.50 GHz, with 4.00 GHz boost; while the 16-core/32-thread 1950X ticks at 3.40 GHz, with 4.00 GHz boost. AMD also ran live demos of the Threadripper chips, in which the 12-core 1920X was shown to beat 10-core Intel Core i9-7900X at Cinebench R15 multi-threaded benchmark. The 16-core 1950X was shown to be close to 50% faster than the i9-7900X. The company also confirmed pricing.
The Ryzen Threadripper 1920X is priced at USD $799, while the Threadripper 1950X goes for a stunning $999. Both chips feature 32 MB of L3 cache, a 64-lane PCI-Express root complex, which enables full x16 bandwidth for up to 3 graphics cards; and a quad-channel DDR4 memory interface. Of course, both SKUs are completely unlocked. Both Threadripper parts will be available in the market by "early August" alongside a wave of compatible socket TR4 motherboards based on the AMD X399 chipset. At its SIGGRAPH 2017 event held on July 27, the company will formally launch the Ryzen 3 series, the Ryzen Threadripper series, and the Radeon RX Vega family of high-end graphics cards.
The video presentation follows:

Add your own comment

118 Comments on AMD CEO Talks Ryzen Threadripper and Ryzen 3 Series in Latest Company Video

#26
efikkan
btarunrAMD effectively murdered the entire Intel Core X lineup below the i9-7900X. Even if the i7-7800X or i7-7850X somehow scrape through in CPU performance using Intel's latest spate of PR bullshit; they still can't get away with crippled PCIe (28-lane).
Just the other day TPU were bashing Intel for their PR bullshit, yet you embrace AMD's PR bullshit. It's sad to see the days of unbiased reporting has ended. It's fine to be though on bullshit, but you have to strive for fairness and be unbiased.
TheLostSwedeI'm sorry, but what are you smoking? You think it's too much to pay $999 for a 16-core CPU when Intel wants $1,699 for their yet to launch counterpart. :kookoo:
You know very well Intel has higher IPC and higher clocks. You should compare actual performance levels, not "specifications". Ryzen surely does well in select benchmarks, and of course AMD focuses on those (as everyone does). But what really matters is actual performance in real workloads.
Posted on Reply
#28
R0H1T
efikkanJust the other day TPU were bashing Intel for their PR bullshit, yet you embrace AMD's PR bullshit. It's sad to see the days of unbiased reporting has ended. It's fine to be though on bullshit, but you have to strive for fairness and be unbiased.


You know very well Intel has higher IPC and higher clocks. You should compare actual performance levels, not "specifications". Ryzen surely does well in select benchmarks, and of course AMD focuses on those (as everyone does). But what really matters is actual performance in real workloads.
You should also know that SKL-X dropped the ball on IPC, it isn't consistently faster than BDW-E across the board.
TR should be +5 to -15% clock for clock against it depending on the application being run atm, that's without more Zen specific optimizations that Linux or Windows may bring to the table.
Posted on Reply
#29
Countryside
efikkanJust the other day TPU were bashing Intel for their PR bullshit, yet you embrace AMD's PR bullshit. It's sad to see the days of unbiased reporting has ended. It's fine to be though on bullshit, but you have to strive for fairness and be unbiased.

You know very well Intel has higher IPC and higher clocks. You should compare actual performance levels, not "specifications". Ryzen surely does well in select benchmarks, and of course AMD focuses on those (as everyone does). But what really matters is actual performance in real workloads.
You are trying to throw shite against the wall and hoping that it will stick.
Posted on Reply
#30
Lionheart
efikkanJust the other day TPU were bashing Intel for their PR bullshit, yet you embrace AMD's PR bullshit. It's sad to see the days of unbiased reporting has ended. It's fine to be though on bullshit, but you have to strive for fairness and be unbiased.


You know very well Intel has higher IPC and higher clocks. You should compare actual performance levels, not "specifications". Ryzen surely does well in select benchmarks, and of course AMD focuses on those (as everyone does). But what really matters is actual performance in real workloads.
Posted on Reply
#31
Estaric
efikkanJust the other day TPU were bashing Intel for their PR bullshit, yet you embrace AMD's PR bullshit. It's sad to see the days of unbiased reporting has ended. It's fine to be though on bullshit, but you have to strive for fairness and be unbiased.


You know very well Intel has higher IPC and higher clocks. You should compare actual performance levels, not "specifications". Ryzen surely does well in select benchmarks, and of course AMD focuses on those (as everyone does). But what really matters is actual performance in real workloads.
No offense but I'm not seeing the AMD PR bullshit, as with most announced products they showed scores and how it stakes up against the competition. Of course a company when announcing there products arnt gonna be like "oh yeah BTW our cpu doesn't stand a chance against Intel in single core performance" it wouldn't be smart for a company to come straight out with the flaws of there products
Posted on Reply
#32
Particle
Sempron Guywake me up when you can slap those two on an AM4 socket then you can justify your $850 pricing
It's two of the same dies on a different PCB. Your point is nonsense.
ShurikNAbsolutely not. TR brings more pcie lanes and quad cannel.
It just connects all of the things that two of what is already on an 1800X already has. The 1800X has some of its PCIe lanes disabled.
TheLostSwedeI'm sorry, but how is this "two of those" stuck together? As far as I'm aware, the 1800X has a total of 24 PCIe lanes, which would make this a 44+4 lanes, which it's not, it's 60+4 for starters.

Secondly, is this retail price? The MSRP for the 1800X is $499, not $420 and this is MSRP, so retail might very well be lower, or higher, depending on the retailers selling the chips.

Seriously people, get a grip...
It is literally two of the same dies stuck together. The 1800X you're referencing has the same PCIe root hub inside of it. It's just not fully exposed. If anything, you should be complaining about how the 1800X is artificially limited.

MSRP doesn't matter, only what you can actually buy it for does. At launch, retailers are going to charge TR's full MSRP and it will fall over time as the 1800X's price has.
iOTotally flawless logic. So a 1800X should be $120 because the $480 8 core EPYC consists of 4 dies?
The 1800X is a fully enabled and fully functional die while the same is not true of the dies on an 8-core EPYC. Your point is nonsense.
FxThis isn't like a fast food joint where you save money the larger size of soda you get.

This is a premium CPU and demands a premium price and yet it still isn't anywhere near Intel's pricing.
I agree that it's a good value compared to Intel's offerings, but that isn't the point I was making. I'm also not suggesting a savings where you get more cores per dollar with the high end, just not gouging where you get less.
Trxdwell by your logic its pricing is fine as 1800x retail price announced by AMD was 499$..so if i am not wrong it does justify your logic of two cpu glued together and the retail price also..
The 1800X launched in the past obviously. It doesn't matter what it sold for back then, just what it sells for now. Double the current market price would be fine.
Posted on Reply
#33
TheLostSwede
News Editor
efikkanJust the other day TPU were bashing Intel for their PR bullshit, yet you embrace AMD's PR bullshit. It's sad to see the days of unbiased reporting has ended. It's fine to be though on bullshit, but you have to strive for fairness and be unbiased.
You're aware that it's two different writers that wrote the stories, right? Makes you bashing the writer of this story a bit unfair and unbiased as well...
efikkanYou know very well Intel has higher IPC and higher clocks. You should compare actual performance levels, not "specifications". Ryzen surely does well in select benchmarks, and of course AMD focuses on those (as everyone does). But what really matters is actual performance in real workloads.
Uhm, Intel doesn't have higher clocks when the core count goes up. Assuming (although maybe a bit pre-maturely) that the i9-7980XE will be based on the Xeon Gold 6150, we have a base clock of 2.7 vs 3.4GHz. I doubt Intel can scare up the base clock 700MHz, although in all fairness there's a two core advantage in this case to Intel, but also a $1,999 price tag. The boost clock is 3.7GHz vs 4GHz, so again, advantage AMD. Let's assume Intel gets their boost clock up to 4GHz as well and you might be right that Intel performs slightly better, but will it really be worth the extra money?

Also, what are "real workloads" to you? Admittedly we've only seen a single benchmark so far for Threadripper and it's not a core to core comparison (as Intel has as yet deliver its higher core count parts), but if you were to spend $999 on a CPU, Threadripper looks like a much more attractive option to me compared to the i9-7900X, but each to their own.
Posted on Reply
#34
efikkan
R0H1TYou should also know that SKL-X dropped the ball on IPC, it isn't consistently faster than BDW-E across the board.
TR should be +5 to -15% clock for clock against it depending on the application being run atm, that's without more Zen specific optimizations that Linux or Windows may bring to the table.
You know very well Skylake-X has the highest IPC of any x86 design at the moment.
There is nowhere a 16-core Threadripper will be on par with a 16-core Skylake-X, 1950X will be competing with 10 and 12 core Skylake-X in overall performance.

And once again, you default to the missing "optimizations" for AMD. The facts are simple; there are no special feature sets in Ryzen to optimize for. Stop this BS now.
Posted on Reply
#35
refillable
How is this one not "PR Bullshit" and Intel's one "clearly" was?
Posted on Reply
#36
dir_d
efikkanYou know very well Skylake-X has the highest IPC of any x86 design at the moment.
There is nowhere a 16-core Threadripper will be on par with a 16-core Skylake-X, 1950X will be competing with 10 and 12 core Skylake-X in overall performance.

And once again, you default to the missing "optimizations" for AMD. The facts are simple; there are no special feature sets in Ryzen to optimize for. Stop this BS now.
You forget 1 Major factor and that is price.
Posted on Reply
#37
cdawall
where the hell are my stars
Well now it is a race to whom can deliver a matx board that is good first. I am really leaning towards this threadripper setup it will be a nice swap out for my x99 build I think. Hopefully Asus can deliver something with quality like the x99m-ws
Posted on Reply
#38
Unregistered
efikkanYou know very well Skylake-X has the highest IPC of any x86 design at the moment.
There is nowhere a 16-core Threadripper will be on par with a 16-core Skylake-X, 1950X will be competing with 10 and 12 core Skylake-X in overall performance.

And once again, you default to the missing "optimizations" for AMD. The facts are simple; there are no special feature sets in Ryzen to optimize for. Stop this BS now.
Considering threadripper's clockspeeds and the fact it has 64 working pcie lanes, AMD must have done at least some optimizing.
Posted on Edit | Reply
#39
efikkan
dir_dYou forget 1 Major factor and that is price.
No I did not. i9-7900X and 1950X cost the same, so they are within the same range.
Hugh MungusConsidering threadripper's clockspeeds and the fact it has 64 working pcie lanes, AMD must have done at least some optimizing.
PCIe lanes have nothing to do with optimizations.
Posted on Reply
#40
Gasaraki
btarunrAMD effectively murdered the entire Intel Core X lineup below the i9-7900X. Even if the i7-7800X or i7-7850X somehow scrape through in CPU performance using Intel's latest spate of PR bullshit; they still can't get away with crippled PCIe (28-lane).
Oh come on. Crippled the entire Core X lineup? Who running i5s and i3 care about more than 28 lanes of PCIe? If you are not running SLI 28 lanes is plenty. Ask the Ryzen users they only have 24.
Posted on Reply
#41
R0H1T
efikkanYou know very well Skylake-X has the highest IPC of any x86 design at the moment.
There is nowhere a 16-core Threadripper will be on par with a 16-core Skylake-X, 1950X will be competing with 10 and 12 core Skylake-X in overall performance.

And once again, you default to the missing "optimizations" for AMD. The facts are simple; there are no special feature sets in Ryzen to optimize for. Stop this BS now.
It;s not, SKL-X is not better than BDW-E across the board clock for clock, stop misrepresenting facts, also IPC depends on the application.

www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-5.html
www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-6.html
www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-7.html
www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-8.html

What BS are you talking about did you not see the ROTR or AoS patches, what they did for Zen? That's just for two games, you're telling me win10 is running the best it can on Zen, when the chip itself was unveiled this year? How about SKL-X & that AVX 512, do programs simply run AVX (512) code without having SKL-X specific path as if it was Broadwell?
Posted on Reply
#42
f22a4bandit
efikkanNo I did not. i9-7900X and 1950X cost the same, so they are within the same range.
That's funny, the 16 core/32 thread part is predicted to sell for $999 compared to the 10 core/20 thread part that currently sells for nearly $1,200. How is the 10 core a better deal than the 16 core beyond the brand name?
Posted on Reply
#43
TheLostSwede
News Editor
ParticleIt is literally two of the same dies stuck together. The 1800X you're referencing has the same PCIe root hub inside of it. It's just not fully exposed. If anything, you should be complaining about how the 1800X is artificially limited.

MSRP doesn't matter, only what you can actually buy it for does. At launch, retailers are going to charge TR's full MSRP and it will fall over time as the 1800X's price has.
Ok, please go ahead and make things up, because you clearly knows best. Got any proof that it's "literally two of the same dies stuck together"? How do you know AMD limited the PCIe root hub? Do you have side by side die shots to prove that? If anything, it's "literally" the same as the Epyc 7351P but at higher clocks and with half the PCIe lanes. Until we had die shots of all three side by side, no-one can say that they're the same part.

If you have a look at some of my Ryzen posts, I did point out that it should've had more PCIe lanes, as it's the one thing that disappointed me about it, but hey, I guess you didn't notice that as you were too busy bitching.

What do you mean MSRP doesn't matter? You just complained that AMD charged too much and then say it doesn't matter. Your logic is very confusing. So are you saying Intel's MSRP doesn't matter as well then, as their prices will maybe also drop over time? The only thing anyone can go by at launch is MSRP's until we see what the actual retail prices are. I think you need some serious help dude :kookoo:
Posted on Reply
#44
Countryside
Time to open my beer and the popcorn will be ready in a few this is getting really interesting here.
Posted on Reply
#45
5DVX0130
This thread at the moment. :shadedshu:



Just ignore them, and try to actually comment on the news.

This year has been amazing so far, well at least in the CPU field. It truly reminds of the golden age of computing.
You know when changing the CPU actually made a difference, and you didn’t only do it because you wanted a new chipset.

Competition and innovation FTW :rockout:
Posted on Reply
#46
efikkan
f22a4banditThat's funny, the 16 core/32 thread part is predicted to sell for $999 compared to the 10 core/20 thread part that currently sells for nearly $1,200. How is the 10 core a better deal than the 16 core beyond the brand name?
Overpricing in shops is not the maker's fault. At times the Fury X was sold for a 50% premium.
Posted on Reply
#47
EarthDog
f22a4banditThat's funny, the 16 core/32 thread part is predicted to sell for $999 compared to the 10 core/20 thread part that currently sells for nearly $1,200. How is the 10 core a better deal than the 16 core beyond the brand name?
7900x msrp (tray) is 1000. TR is (rumored) to be the same. It will see the same new cpu i flation as 7900x.

TR is a good value there if you can use more than 10c/20t. Otherwise, you are piling on cores for no reason. Id go for the 'generally' faster ipc chip too... especially since it can overclock to 4ghz+ with relative ease. I can run mine at 4.5ghz (custom loop 3x120mm rad)
Posted on Reply
#48
Particle
TheLostSwedeOk, please go ahead and make things up, because you clearly knows best. Got any proof that it's "literally two of the same dies stuck together"? How do you know AMD limited the PCIe root hub? Do you have side by side die shots to prove that? If anything, it's "literally" the same as the Epyc 7351P but at higher clocks and with half the PCIe lanes. Until we had die shots of all three side by side, no-one can say that they're the same part.

If you have a look at some of my Ryzen posts, I did point out that it should've had more PCIe lanes, as it's the one thing that disappointed me about it, but hey, I guess you didn't notice that as you were too busy bitching.

What do you mean MSRP doesn't matter? You just complained that AMD charged too much and then say it doesn't matter. Your logic is very confusing. So are you saying Intel's MSRP doesn't matter as well then, as their prices will maybe also drop over time? The only thing anyone can go by at launch is MSRP's until we see what the actual retail prices are. I think you need some serious help dude :kookoo:
I would encourage you to re-read my post as you have somehow managed to misunderstand the point I made about MSRP. I'd explain it to you again, but correcting your reading comprehension failure is not my job.

As for the dies being the same, what evidence do you have that they are not? AMD has talked extensively about how part of the beauty of their new product lineup is their extensive use of MCMs with Infinity Fabric since they can produce the smaller dies with much higher yields and then create products in their various product segments from that common stock at lower costs. I would say that your assertion that the dies would not be the same would less follow from common logic, and as such the burden of proof is yours.
Posted on Reply
#49
eidairaman1
The Exiled Airman
Attacking the writer was a big mistake... this thread is about AMD not intel, bias needs to be left at the door.
Posted on Reply
#50
TheLostSwede
News Editor
ParticleAs for the dies being the same, what evidence do you have that they are not? AMD has talked extensively about how part of the beauty of their new product lineup is their extensive use of MCMs with Infinity Fabric since they can produce the smaller dies with much higher yields and then create products in their various product segments from that common stock at lower costs. I would say that your assertion that the dies would not be the same would less follow from common logic, and as such the burden of proof is yours.
For cores, yes, not the PCIe root complex which takes up a HUGE die area. It simply doesn't make sense that they use the same cores when you have 28, 64 and 128 PCIe lanes. That said, I have as much proof as you do, so let's wait for the die shots...

However, looking at the Ryzen die shot, it doesn't look like there's any unused PCIe lanes in there.
Posted on Reply
Add your own comment
Dec 18th, 2024 10:49 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts