AMD Announces Ryzen 7 1800X, 1700X, 1700 And Pricing, Pre-orders Begin Today

[bit_user]

With this article, they did a hybrid of their Intel benchmarks to the benchmarks given by AMD. It is too bad that they didn't mention whether they were using dual or quad channel memory though. 🙁

Well, they got 68% of the difference between AMD's number and the overclock3d.net number I posted above. So, that suggests quad-channel.

Anyway, Ryzen did 3.8% better than PC World's score of 1542.

 

[Jarmund]

Yes! finally i can dismantle the mini nuclear reactor in my garden that is my FX-9590!

A question: What will happen with the FX series?

 

[tamalero]

Its clear that AMD wants to focus on high end server market first. Very similar to the release of K8.
I hope this means the resurgence of AMD. The upgrades by intel in performance between generations are pitiful for the huge prices.

 

[alextheblue]

X1800 performs as i5 6600@4.0Ghz in gaming which means Kaby Lake is going to walk over. Besides attractive pricing there is really nothing here to see.

Are we going to see official benchmarks now so people can get disappointed or feel good about their Kaby Lake purchase. It is no mistake AMD is comparing against Intel 2011 v3 socket in gaming where Intel CPU is not even properly overclocked because they know Ryzen is losing big time against Kaby Lake in gaming which is also faster than even 10 Core 2011-V3 CPU.

But i do have to admit, those prices especially for unlocked X1700 are attractive and will push mainstream gaming to utilize more cores 4-6.

Now, waiting for Intel price adjustment so i can update one of the systems to 8 Core.

Well you're kind of using mixed talking points here. 2011 platform isn't really a gaming platform, and you are talking about gaming... but you're also talking about upgrading to 8 cores. You can't upgrade 1151 to 8 core. So if you're considering 2011 you should know that platform will likely get stomped on price/performance vs Ryzen. 115x is a better fight but an overclockable system is less expensive on AMD, and you get more threads for your money. If you're truly discussing gaming performance only, then I think it would be more interesting to look at Ryzen 5 variants.

As you go down to mid-range platforms, Ryzen looks even better especially paired with a B350 board. At current prices i5 would compete with 4C/8T processors for the most part, and the top model unlocked i5 is up against 6C/12T! For entry-mid level chips the situation is even worse for Intel. For the cost of a multiplier-locked i3 with two cores and four threads, you get a 4C/4T unlocked Ryzen. Not even close - Ryzen will blow the i3s away. The top-dog unlocked i3 once again is up against a higher tier - 4C/8T Ryzen 5... not even counting the reduced cost of a B350 OC-capable board. If Intel doesn't want this to happen, they'll have to drop prices... a lot.

 

[bit_user]

You can't upgrade 1151 to 8 core.

I'd read a rumor that Cannon Lake/Coffee Lake would go up to 8 cores in LGA 1151. But the latest I'm finding now is 6 cores:

https://arstechnica.com/gadgets/2017/02/intel-coffee-lake-14nm-release-date/

 

[kyotokid]

You know, people use their computer for things other than gaming.

But also, the i5 6600 for $220 would compete with the Ryzen lowest 4 core 4 thread chip R3 1200x for $150. Also the cheap $70 motherboard would make for a cheap gaming rig under $550. (Power Supply $40, Case $40, 8GB DDR4 ram $50, 128GB SSD $40, rx470 vid card $160). Even cheaper if you can cannibilize parts from an old pc like hard drive, case,vid card and power supply.

Fact is, even if ryzen single thread performance is 10% lower than intel, with the pricing the way it is, it will be competitive and disruptive to the market.

...exactly. I work in 3D CG and so far, the Rizen MBs that I've seen ready for release are insufficient for my needs. Most only have one PCIe 3.0 x16 slot and a maximum of 4 Dual Channel DIMM slots (some only 2).

The system I am working on building that is based on "legacy" tech (Sandy Bridge) would perform better: Dual 8 Core Xeons, 128 GB DDR3 in Quad Channel and x3 PCIe 3.0 x16 expansion slots for when I can get my hands on a trio of 1070s. The big memory is for render engines that do not support GPU rendering and scenes that exceed 8 GB in file size (I've been known to create some very "big" scenes in terms of polycount, textures, and environmental effects). When it comes to CPU/Physical Memory based rendering, DDR4 has little to no advantage over DDR3.

The other nice feature, I can run this on W7 Pro.

 

[bit_user]

the Rizen MBs that I've seen ready for release are insufficient for my needs. Most only have one PCIe 3.0 x16 slot and a maximum of 4 Dual Channel DIMM slots (some only 2).

You want Naples, the server/workstation version of Ryzen, due out a few months later.

http://wccftech.com/amd-zen-naples-soc-benchmarks/

I like workstation hardware. I have a Sandybridge-E, which lets me use ECC RAM, and has enough PCIe 3.0 lanes that I could add a NVMe SSD in a slot directly connected to the CPU.

I was looking forward to Purley, but it's looking like that might largely miss the single-CPU workstation market.

 

[kyotokid]

the Rizen MBs that I've seen ready for release are insufficient for my needs. Most only have one PCIe 3.0 x16 slot and a maximum of 4 Dual Channel DIMM slots (some only 2).

You want Naples, the server/workstation version of Ryzen, due out a few months later.

http://wccftech.com/amd-zen-naples-soc-benchmarks/

I like workstation hardware. I have a Sandybridge-E, which lets me use ECC RAM, and has enough PCIe 3.0 lanes that I could add a NVMe SSD in a slot directly connected to the CPU.

I was looking forward to Purley, but it's looking like that might largely miss the single-CPU workstation market.

...Now that Diesel based system would make a beast of a workstation 128 CPU cores, 128 GB Octo Channel DDR4.

The only questions, would it also support W7 Pro, Nvidia GPUs, and will I win the next Megabucks Lotto?

 

[GoatGuy]

Performs ON PAR with a top-flight Intel CPU
Costs half.
Has similar TDP.
What's wrong with that?

OR ... if you prefer

Has 8 cores. 16 threads. Each core/thread is competent. Power, performance, perf-per-buck.

What's wrong with that?

 

[jwl3]

Welcome back to the party. It must've been cold there living in Intel's shadow for the past 7 years.

 

[kyotokid]

Performs ON PAR with a top-flight Intel CPU
Costs half.
Has similar TDP.
What's wrong with that?

OR ... if you prefer

Has 8 cores. 16 threads. Each core/thread is competent. Power, performance, perf-per-buck.

What's wrong with that?

 

[kyotokid]

Performs ON PAR with a top-flight Intel CPU
Costs half.
Has similar TDP.
What's wrong with that?

OR ... if you prefer

Has 8 cores. 16 threads. Each core/thread is competent. Power, performance, perf-per-buck.

What's wrong with that?

...as I mentioned the forthcoming Ryzen MBs only offer 2 or 4 dual channel DIMM slots and only 1 PCIe3.0 x16 slot. Great for gaming....not so much for serious CG production.

As I work in 3D CG, I need a little more than that. Since I also use some render engines that do not support GPU rendering, the more CPU cores and memory the better. Hence, why I am designing a dual 8 core E52690 Xeon (2.9 GHZ/turbo boost 3.8 GHz) build with 128 GB Quad Channel DDR3 memory. The boards I am looking at support three PCIe 3.0 x 16 slots.

Yeah this is a "shoestring budget" workstation that I could build for less than the price of two of the fastest 8 core Broadwell Xeon CPUs alone (E52667 v4 @ 3.2 GHz).

The other issue is for GPU rendering the engines I use require Nvidia cards as they are all CUDA based. Not sure of how much support AMD gives to their competitor.

 

[GoatGuy]

Performs ON PAR with a top-flight Intel CPU
Costs half.
Has similar TDP.
What's wrong with that?

OR ... if you prefer

Has 8 cores. 16 threads. Each core/thread is competent. Power, performance, perf-per-buck.

What's wrong with that?

...as I mentioned the forthcoming Ryzen MBs only offer 2 or 4 dual channel DIMM slots and only 1 PCIe3.0 x16 slot. Great for gaming....not so much for serious CG production.

As I work in 3D CG, I need a little more than that. Since I also use some render engines that do not support GPU rendering, the more CPU cores and memory the better. Hence, why I am designing a dual 8 core E52690 Xeon (2.9 GHZ/turbo boost 3.8 GHz) build with 128 GB Quad Channel DDR3 memory. The boards I am looking at support three PCIe 3.0 x 16 slots.

Yeah this is a "shoestring budget" workstation that I could build for less than the price of two of the fastest 8 core Broadwell Xeon CPUs alone (E52667 v4 @ 3.2 GHz).

The other issue is for GPU rendering the engines I use require Nvidia cards as they are all CUDA based. Not sure of how much support AMD gives to their competitor.

NICE!
I've always wondered whether CG was cpu bound, or memory bound. Seems to me that'd it would be WAY more cpu bound, given that ray-tracing (if that's employed) is such a total and irreducible CPU pig.

 

[bloodroses]

Performs ON PAR with a top-flight Intel CPU
Costs half.
Has similar TDP.
What's wrong with that?

OR ... if you prefer

Has 8 cores. 16 threads. Each core/thread is competent. Power, performance, perf-per-buck.

What's wrong with that?

...as I mentioned the forthcoming Ryzen MBs only offer 2 or 4 dual channel DIMM slots and only 1 PCIe3.0 x16 slot. Great for gaming....not so much for serious CG production.

As I work in 3D CG, I need a little more than that. Since I also use some render engines that do not support GPU rendering, the more CPU cores and memory the better. Hence, why I am designing a dual 8 core E52690 Xeon (2.9 GHZ/turbo boost 3.8 GHz) build with 128 GB Quad Channel DDR3 memory. The boards I am looking at support three PCIe 3.0 x 16 slots.

Yeah this is a "shoestring budget" workstation that I could build for less than the price of two of the fastest 8 core Broadwell Xeon CPUs alone (E52667 v4 @ 3.2 GHz).

The other issue is for GPU rendering the engines I use require Nvidia cards as they are all CUDA based. Not sure of how much support AMD gives to their competitor.

NICE!
I've always wondered whether CG was cpu bound, or memory bound. Seems to me that'd it would be WAY more cpu bound, given that ray-tracing (if that's employed) is such a total and irreducible CPU pig.

Both actually. I/O usually is the weakest link though. The idea is that you want the CPU to run at 100% of the time with CG, so a lack of memory of any kind would create paging with the much slower storage; hampering performance.

The main reason why Quad Channel memory would be currently preferred in his case is because it supports a max of 128gb of memory vs the max of 64gb on a Ryzen equivalent. While the quad memory will perform faster for I/O, the fact of not having to page the memory is a much bigger advantage.

This thread explains it pretty well:
http://forums.cgsociety.org/archive/index.php?t-166755.html

 

[ttt_2017]

It is Funny how in all the articles here on Tomshardware about the new AMD CPU you guys never mention that this CPU is ECC enabled as well ... sadly Tomshardware turned into a "gaming" PC site that cares only about gaming when it comes to CPUs .. the community as well.

 

[JamesSneed]

It is Funny how in all the articles here on Tomshardware about the new AMD CPU you guys never mention that this CPU is ECC enabled as well ... sadly Tomshardware turned into a "gaming" PC site that cares only about gaming when it comes to CPUs .. the community as well.

It is and I haven't seen a motherboard yet saying they support ECC(Gigabyte show support but will run ECC in non-ECC mode). Most others say something like "4 x DIMM, Max. 64GB, DDR4 2666/2400/2133 MHz Non-ECC, Un-buffered Memory".

If there are no motherboards to support ECC it isn't much of an oversight by TomsHardware

 

[ttt_2017]

It is Funny how in all the articles here on Tomshardware about the new AMD CPU you guys never mention that this CPU is ECC enabled as well ... sadly Tomshardware turned into a "gaming" PC site that cares only about gaming when it comes to CPUs .. the community as well.

It is and I haven't seen a motherboard yet saying they support ECC(Gigabyte show support but will run ECC in non-ECC mode). Most others say something like "4 x DIMM, Max. 64GB, DDR4 2666/2400/2133 MHz Non-ECC, Un-buffered Memory".

If there are no motherboards to support ECC it isn't much of an oversight by TomsHardware

you are joking right ? the CPU supports ECC and I dont care if you seen a motherboard or not yet. does not matter.

anyways Asrocks motherboards support ECC memory you can check the specs yourself if you visit their site.



And it is a mistake from Tomshardware site not to mention ECC support at all about the new AMD CPU. . because the i7-7700K DOES NOT

 

[littleleo]

The fact AMD can match Intel's performance is great. That they do it for half the cost is AMAZING. Then consider AM4 is the ONLY (consumer) socket AMD will have for the forseeable future - Not this "new-motherboard-every-incremental-processor-release" that Intel has always been horrible with. knowing I'll still have a clean upgrade path 3 years from now is what sells me on AMD.

Where have you heard that AMD is going to keep the same socket for 3+ years worth of CPUs? I think they've confirmed that Zen+ is going to be AM4, but I haven't heard anything beyond that (so currently two generations of CPUs confirmed for the socket, which is Intel's historical habit).

At CES this year Mark Papermaster, AMD’s chief technology officer, confirmed the four-year lifespan.

 

[bit_user]

And it is a mistake from Tomshardware site not to mention ECC support at all about the new AMD CPU. .

Thanks for pointing this out! Years ago, I used a Phenom II in a cheap server build, specifically because it supported ECC. In fact, the board I got even had support for patrol scrub, which is a server-grade feature, yet the board was priced like an upper-end desktop board.

because the i7-7700K DOES NOT

Sigh. Most of Intel's mainstream CPUs, since the Core2, do not support it. The main exception is the i3's, although even some recent i3 models don't.

Basically, Intel is trying to make you buy Xeons, if you want ECC.

 

[kyotokid]

Performs ON PAR with a top-flight Intel CPU
Costs half.
Has similar TDP.
What's wrong with that?

OR ... if you prefer

Has 8 cores. 16 threads. Each core/thread is competent. Power, performance, perf-per-buck.

What's wrong with that?

...as I mentioned the forthcoming Ryzen MBs only offer 2 or 4 dual channel DIMM slots and only 1 PCIe3.0 x16 slot. Great for gaming....not so much for serious CG production.

As I work in 3D CG, I need a little more than that. Since I also use some render engines that do not support GPU rendering, the more CPU cores and memory the better. Hence, why I am designing a dual 8 core E52690 Xeon (2.9 GHZ/turbo boost 3.8 GHz) build with 128 GB Quad Channel DDR3 memory. The boards I am looking at support three PCIe 3.0 x 16 slots.

Yeah this is a "shoestring budget" workstation that I could build for less than the price of two of the fastest 8 core Broadwell Xeon CPUs alone (E52667 v4 @ 3.2 GHz).

The other issue is for GPU rendering the engines I use require Nvidia cards as they are all CUDA based. Not sure of how much support AMD gives to their competitor.

NICE!
I've always wondered whether CG was cpu bound, or memory bound. Seems to me that'd it would be WAY more cpu bound, given that ray-tracing (if that's employed) is such a total and irreducible CPU pig.

..it is both as memory is important to handle the size of scene file (polygons and texture files) The more that the is load is spread out among the total memory available, the better the render performance. One person I know brought a 64 GB Dual 6 core Xeon Mac Pro to it's knees with a scene which had an extreme amount of reflectivity. The process went into swap mode as Both CPU load and memory usage were pegged at 100%.

CPU mode is still better for high quality than pure GPU rendering. Many of the big studios like Pixar still use CPU based rendering as physical memory is also far more cost effective than VRAM when setting up warehouse sized render farms. For about 2,000$ less than a single 24 GB Quadro P6000 you can get 512 GB of DDR4 2400 memory (16 x 32 GB).

 

[ttt_2017]

And it is a mistake from Tomshardware site not to mention ECC support at all about the new AMD CPU. .

Thanks for pointing this out! Years ago, I used a Phenom II in a cheap server build, specifically because it supported ECC. In fact, the board I got even had support for patrol scrub, which is a server-grade feature, yet the board was priced like an upper-end desktop board.

because the i7-7700K DOES NOT

Sigh. Most of Intel's mainstream CPUs, since the Core2, do not support it. The main exception is the i3's, although even some recent i3 models don't.

Basically, Intel is trying to make you buy Xeons, if you want ECC.

You are welcome ...

The AMD 1700 8cores/16 threads running at 3Ghz with ECC support and only 65watts is a very good entry to mid level server CPU ...

for $320 ? dirt cheap server CPU compared to 8 cores Xeons at the same clock speed.

I am supporting AMD and buying from them. AMD needs to stay ... even if Intel lowers their Prices I am still getting AMD just to support them.

 

[Rogue Leader]

...Guess what, if you really need CPU for video editing...you can buy 14/28 Xeon Retail running at 2.5Ghz all cores turbo 3.0Ghz and it will eat AMD Ryzen and i7 like there is no tomorrow and cost of CPU is $350. So that argument is really BS.

WHAT are you smoking dude? Xeon 14 core (28 with HT) running at 2.5GHz "Retail CPU" for $350?? WTF none of that makes any sense whatsoever...

The cheapest Xeon 14 core (which runs at 1.7GHz btw) costs $1399.99.. the 2.6GHz version runs $2133.99...
There is NO 2.5GHz model (for sale).

You are just full of sheeet. Go troll somewhere else.

Go check eBay, there are ton of retail Xeon E5 2683 14/28 (NOT ES) 2.0@2.5@3.0 processors for ~ $350. I bought one for myself. I am not a troll but you are just misinformed.

You're talking about a CPU with the same or lower IPC running 1-2 ghz slower than any of the Ryzen CPUs, so no, not a chance in heck of it eating Ryzen, except maybe in processes that can utilize all 14 cores/28 threads. Realize not all video editing software will. Meanwhile in anything else a Ryzen or a newer lower core count Xeon, or a Sky/kabylake will be faster in every way.

 

[bit_user]

The AMD 1700 8cores/16 threads running at 3Ghz with ECC support and only 65watts is a very good entry to mid level server CPU ...

for $320 ? dirt cheap server CPU compared to 8 cores Xeons at the same clock speed.

It's good, and very good price/performance. However, when I say "cheap server build", I mean I paid < $300 for the Phenom II x4 955 (quad core) CPU + the 890FX board I got, which had the ECC support and 6x SATA 3 ports. And that was for new parts, from an authorized reseller. Not that I needed to, but it could also run both of its x16 PCIe 2.0 slots in full x16 mode.

That CPU had a lot more I/O, for its day, than AM4's mere 24 lanes of PCIe represents, today. I get that HyperTransport needed to be replaced, but it did have its good points.

 

[bit_user]

...Now that Diesel based system would make a beast of a workstation 128 CPU cores, 128 GB Octo Channel DDR4.

Here's a 64-core/256-thread Xeon Phi x200 (Knights Landing) running Cinebench R15 w/ 6-channel DDR4 and 16 GB of HMC (probably setup to act as L3 cache).

https://www.servethehome.com/intel-xeon-phi-x200-knights-landing-boots-windows/

As he says, it's not using any KNL optimizations (such as AVX512), so the score is only slightly better than Ryzen's: 1648. Remember, it's just 64 Silvermont Atoms that've been extended to support 4x hyperthreading + AVX512, running at 1.3 GHz.

Interestingly, you can configure that CPU to act like a quad-CPU cluster. If he did that & setup a render farm, the result would certainly be better. If he then enabled AVX512, I wonder if he could even double the score.

BTW, according to this, the top Cinebench R15 score is 4918 - achieved by a dual E5-2699 v4 Xeon (44 Broadwell cores @ 2.2 GHz):

https://us.rebusfarm.net/en/tempbench?view=benchmark

 

[kyotokid]

...Now that Diesel based system would make a beast of a workstation 128 CPU cores, 128 GB Octo Channel DDR4.

Here's a 64-core/256-thread Xeon Phi x200 (Knights Landing) running Cinebench R15 w/ 6-channel DDR4 and 16 GB of HMC (probably setup to act as L3 cache).

https://www.servethehome.com/intel-xeon-phi-x200-knights-landing-boots-windows/

As he says, it's not using any KNL optimizations (such as AVX512), so the score is only slightly better than Ryzen's: 1648. Remember, it's just 64 Silvermont Atoms that've been extended to support 4x hyperthreading + AVX512, running at 1.3 GHz.

Interestingly, you can configure that CPU to act like a quad-CPU cluster. If he did that & setup a render farm, the result would certainly be better. If he then enabled AVX512, I wonder if he could even double the score.

BTW, according to this, the top Cinebench R15 score is 4918 - achieved by a dual E5-2699 v4 Xeon (44 Broadwell cores @ 2.2 GHz):

https://us.rebusfarm.net/en/tempbench?view=benchmark

...impressive that render test looked more like GPU than CPU based. Of course for us peons who only mess around with CG this is way out of our league as it would also require Windows Server Edition (expensive in and of itself) or Linux, the latter which isn't supported by many CG software developers. Only Modo, Maya, and Blender (the latter which I have tried to get into several times but gave up on because of its cumbersome keyboard driven UI that has it's own very steep learning curve) offer Linux compatible versions.

For myself just moving to dual E5 2690s and 128 GB of Quad Channel DDR3 from an old i7 930 with 12 GB of Tri Channel DDR3 will be a pretty big step up in performance.

as to the 2699 v3, a single Quadro P6000 would cost as much as two of those 22 core Xeons. 88 proceeing threads is nothing to sneeze about especially when you have 128 GB of quad channel memory instead of only 24 GB to back it up. Too bad W7 Pro only supports a maximum of 192 GB instead of 256. Not too keen into turning my workstation into a desktop smartphone with 8,1 and leaving a back door open to MS with W10 is out of the question.