AMD Ryzen 2600 Benchmark Spotted

[mackerel]

https://videocardz.com/75194/amd-ryzen-2000-series-exposed-pricing-performance-leaked

Alleged leaked slides with the 2nd gen details.

No 2800X!
2700X, base 3.7, turbo 4.35, 105W TDP, 369 USD
2700, base 3.2, turbo 4.1, 65W TDP, 299 USD
2600X, base 3.6, turbo 4.25, 95W TDP, 249 USD
2600, base 3.3, turbo 3.9, 65W TDP, 199 USD

Not sure, looks like they might be including coolers with the X parts now?

Difference between X370 and X470 chipsets seems to be XFR2 Enhanced and Precision Boost Overdrive, for more clock on more cores. May help out of box performance but not sure there's any benefit to manual overclocking.

I'm gonna play wait and see, will probably get either 2700(X) to play with, but not sure I'm going to spring on a new mobo. Just swap my CPUs around and maybe sell on the 1600 if anyone still wants it at that point.

 

[EarthDog]

I see TDP has gone up to reach those clocks (on two)... interesting. I would have thought the die shrink and any improvements would have prevented that...

 

[mackerel]

Unless I'm missing something, only the 2700X is up from 1700X by 10W, presumably they really wanted to get that bit more clock out of the top end as they know it'll be compared against the 8700k. It is also somewhat suggestive of where the wall is, I'm still guessing 4.5 may be attainable with a bit of luck, some voltage and cooling to match.

 

[EarthDog]

Oops, yes one.

If they needed to raise tdp over 10% to get 135mhz, the wall is still painfully obvious. I hope 4.4 is in the cards as these couldnt really get by their xfr clocks with ease...

 

[Johan45]

I'm curious to see if the 2800X has a higher core count but still available on the AM4 socket. It was rumoured back in Nov-Dec I think. It seems like something AMD would do

 

[Woomack]

Rumours were before ryzen premiere and later it was little disappointment to see that AM4 got "only" 8 cores.
I wonder if that turbo is close to max OC on all cores or we will see stable 4.5GHz+ on ambient temps.

 

[MaddMutt]

Doesn't it come with an operating manual?

^ The Operating Manual is on a CD\ONLINE that you can access AFTER you get the system WORKING You are **** until you get the system working or have another system close by that you can use.
Don't you just love progress, no more Paper Manuals to read. They have been moved to the CLOUD to save Space and Money

I have one of the first 1800X cpu's that where released. I get a MAX of 4.6GHz and 3066 memory from Dry Ice. If the new 2800X gives 300 - 400 extra MHz, I could hit 5.0GHz with Dry Ice

More info on the new X470 motherboards Please!!!
I have the ASUS CHVI and would like to know if:
a) Current or Future ones will allow better OC'ing
b) Do they have more PCIe lanes than the X370
c) Is this a motherboard upgrade similar to the AM2 - AM2+ (IIRC the AM2+ introduced a separate power phase\plane\something to the memory)
d) Is it a needed upgrade (More PCIe Lanes) for Benching Multiple GPU's
e) More\Better VRM's section or more secondary power inputs.
f) The BEST part I love about the CHVI is that it can use\take AM3 OR AM4 CPU COOLERS

 

[Johan45]

It was just in December 2017 and so outrageous no one paid attention to it. 12c/24t and 5.1 GHz https://www.tweaktown.com/news/60100/amds-new-ryzen-7-2800x-teased-12c-24t-up-5-1ghz/index.html

 

[Dolk]

Lol, increase core counts for Zen. That Core architecture is set in stone for several generations. We will only have 8c/16t for a few years before thread count goes up. I'm not surprised they couldn't squeeze more speed out without increasing power on the chip. The Zen core has a deep pipeline with a lot of smart tech in it to monitor and anticipate operations.

@ed, this die shrink was a test for their 12nm+ and not really meant to achieve great feats. You can really think of this as a 14nm because the FinFets are still very large compared to what 12nm specifications call out. If they get the 7nm process working we will see a bit of a boost in speed, but that may go away if AMD decides to stick to their current plan.

 

[EarthDog]

Correct dolk.

There is a lot of confusion and false expectations on the web.. confusion over zen2 vs zen+, what we see now. Many expected large(r) increase in clocks and overclock headroom.

 

[MaddMutt]

It was just in December 2017 and so outrageous no one paid attention to it. 12c/24t and 5.1 GHz https://www.tweaktown.com/news/60100/amds-new-ryzen-7-2800x-teased-12c-24t-up-5-1ghz/index.html

^ I (As a BASIC Overclocker\Bencher) cannot justify the cost of a Threadripper 1920x 12c/24t or a 1950x 16c/32t processor. Even though they are CHEAPER than their Intel counterparts. The Elite Benchers are running Intel 79xx CPU's (http://cpu.userbenchmark.com/Compar...m/gadgets/2017/09/intel-core-i9-7960x-review/)

^ The rumor mill was spinning in high gear in Dec from our look at the Threadripper to a AM4+ CPU??? The last that I have heard, is that ZEN-2 may contain a 12c/24t core. This is hoping that the 7nm is successful and that GF does not have to release a 10nm or an improved 12nm+ for AMD to work with.

 

[EarthDog]

It was just in December 2017 and so outrageous no one paid attention to it. 12c/24t and 5.1 GHz https://www.tweaktown.com/news/60100/amds-new-ryzen-7-2800x-teased-12c-24t-up-5-1ghz/index.html

Wow, never ran across that... I'd love to see what the TDP is on those parts... we can see at least one of the three chips details weren't accurate either... assuming what we see here is true of course.

EDIT: Or I cant read... this is 2600, lol

 

[Dolk]

Lol this is what I'm talking about. You can't physically increase the core count of Zen. I mean AMD could but they are not doing it. Zen refers to the microcode architecture, but the chip itself is called something else. I'm not talking about the CPU, but the actual chip that includes all the Zen cores. That chip is copy pasted to create Epyc and Threadripper. This means that whole topological architecture is set in stone for several projects over several generations. Any tweak here gets a ton of butt hurt in other segments. There will be 0-core increase with Zen cores.

Maybe Zen 2, but I've already hinted at what is coming with that CPU

 

[mackerel]

I believe the name for the piece of silicon with two CCXs on it is called Zeppelin.

For speculating about greater than 8 cores in an AM4 part without requiring a new die, I have wondered, could they physically fit two of those under the IHS? Looking at photos of delids, I think there is room but not necessarily for the surrounding components also. I don't know how essential they are, or if they could otherwise be implemented in an alternate way. If they could do such a thing, all IO will likely go through one die to maintain socket compatibility, and to reach the 2nd would have to go through the 1st. If you thought inter-CCX latency was a concern, this is worse.

For really wild speculating, I find it hard to believe they would do that. It would essentially be Threadripper-lite. The only provocation might be if Intel were to release an 8-core part at consumer level, but Intel have their own different set of problems there.

 

[Johan45]

Gary Patton on the GF 7nm process

Well, let's do one more AMD related item that will make your eyebrows frown. Globalfoundries Chief Technical Officer, Gary Patton, recently talked about the upcoming 7nm manufacturing generation. The advantages compared to 14nm have grown significantly. The die sizes could be more than halved, clock rates in a range of 5.0 GHz are realistic.

Full article at guru3d http://www.guru3d.com/news-story/am...undries-expects-5-ghz-in-the-7nm-process.html

 

[Alaric]

Gary Patton on the GF 7nm process


Full article at guru3d http://www.guru3d.com/news-story/am...undries-expects-5-ghz-in-the-7nm-process.html

Wow. I still maintain that if AMD can close the speed gap and maintain the multi thread advantage (that last seems likely given SMT's apparent advantages in Zen's implementation), their IPC disadvantage comes close to being a non event. Outside influences aside (fanboyism and Intel's long established role as the guys on top), AMD could actually graduate from "alternative" to actual competitor on an even playing field. I like AMD, and would like to see them finally shake the "for the money" disclaimer they've been carrying around.

 

[Dolk]

I believe the name for the piece of silicon with two CCXs on it is called Zeppelin.

For speculating about greater than 8 cores in an AM4 part without requiring a new die, I have wondered, could they physically fit two of those under the IHS? Looking at photos of delids, I think there is room but not necessarily for the surrounding components also. I don't know how essential they are, or if they could otherwise be implemented in an alternate way. If they could do such a thing, all IO will likely go through one die to maintain socket compatibility, and to reach the 2nd would have to go through the 1st. If you thought inter-CCX latency was a concern, this is worse.

For really wild speculating, I find it hard to believe they would do that. It would essentially be Threadripper-lite. The only provocation might be if Intel were to release an 8-core part at consumer level, but Intel have their own different set of problems there.

They physically have the space to place them but not to route all the signals. Remember, each CPU has some 1331 pins, you are nearly doubling that when adding a second Zeppelin (you are right about that name). That is why Intel and AMD makes such large CPUs for their Workstation and Server CPUs. They need that super wide PCB to route the thousands and thousands of high-speed signals.

 

[mackerel]

My thinking was, to maintain socket compatibility the interface from the CPU package has to remain the same, thus connectivity to the 2nd die would be indirect.

Trying to think of another example, didn't Intel do this in the ancient times of Athlon X2? They wanted in on multi-core, and literally stuck two dies on on package. Not looked at article in depth, might be this? http://www.tomshardware.com/reviews/65-nm-pentium-d-900s-coming-party,1202.html (you might want to check adblocking is in place before clicking...)

 

[Dolk]

Yes the very first multi-core from Intel was just two dies placed on the same PCB. I'm not sure why they did this rather than go straight to making their own multi core single die CPU, but they did it. It was to counter AMD as they already had two cores on one die.

However, that Pentium Duo was absolutely awful and was completely crushed by Athlon X2 at the time.

 

[Alaric]

Yes the very first multi-core from Intel was just two dies placed on the same PCB. I'm not sure why they did this rather than go straight to making their own multi core single die CPU, but they did it. It was to counter AMD as they already had two cores on one die.

However, that Pentium Duo was absolutely awful and was completely crushed by Athlon X2 at the time.

Yeah, if not for Intel's dirtiness, the CPU landscape today might look quite different. I don't mean drastic differences in hardware, but AMD's presence in the market would have been a lot different if Intel hadn't so wrongfully screwed AMD in the court of public opinion. I had a couple Dells with Pentium D chips. Worst POS rigs I've ever seen. They overheated out of the box, and God forbid you get one of the infamous Dell Dust Bunnies in the cooler intake. I was a noob and it took me two hours to actually look in the green chute trying to figure out why it BSOD'd two minutes after booting.
But, I should have been rich and good looking, too. Lemme tell ya' how well that worked out.... LOL