AMD Ryzen Gen 2 Set For Q2 2018

[epsilon84]

As someone who switched from a 6600K to a 1700, don't let the pretty benchmarks fool you. These are done on clean systems with nothing running in the background.

In a real use case where you have things running in the background (discord, YouTube playing music, maybe Windows decides it want to do some maintenance, etc), frametime variance can be hit quite hard on the 6600K.

Meanwhile Ryzen just chugs along since it has a lot of spare resources to deal with background tasks. There is a real and perceptible difference to using the two in a real world scenario.

I hear ya, but that is due to lack of threads more than anything. I doubt you will see such a variance with a 4C/8T CPU like a 6700K for example. 4C/4T CPUs are starting to struggle with the latest AAA games, but personally I simply upgraded from a 2500K to a 3770 (overclocked to 4.4GHz) and I'm good to go for another couple of years at least with 'Ryzen level' gaming performance.

 

[CatMerc]

Thats true, but if you are buying it today 8600k has 6 cores so it will be able too handle game and some background programs running

Never claimed otherwise. The talk was specifically about why one would want to upgrade from a 4670K, and I gave a reason why.

I hear ya, but that is due to lack of threads more than anything. I doubt you will see such a variance with a 4C/8T CPU like a 6700K for example. 4C/4T CPUs are starting to struggle with the latest AAA games, but personally I simply upgraded from a 2500K to a 3770 (overclocked to 4.4GHz) and I'm good to go for another couple of years at least with 'Ryzen level' gaming performance.

Ditto

 

[csbin]

AMD Ryzen 5 2600 and ASUS ROG Crosshair VII Hero spotted

https://videocardz.com/74829/amd-ryzen-5-2600-and-asus-rog-crosshair-vii-hero-spotted

 

[Insert_Nickname]

AMD Ryzen 5 2600 and ASUS ROG Crosshair VII Hero spotted

Interesting. So we can expect plus 200 (or maybe 300MHz on a good day) within the same TDP. That'd take the high-end X's to 4.3GHz XFR. Maybe a bit higher with Precision Boost. Sounds pretty good so far.

Pretty much what I was expecting.

 

[neblogai]

Interesting. So we can expect plus 200 (or maybe 300MHz on a good day) within the same TDP.

Summit Ridge engineering sample had 3.3/3.7 in the name, retail model has specs of 3.2/3.6, but really works at 3.4GHz (100% load). So with this engineering sample having 3.4/3.8 in the name- same principle would net real life 3.5GHz 100% load clocks at 65W.

 

[Atari2600]

AMD Ryzen 5 2600 and ASUS ROG Crosshair VII Hero spotted

https://videocardz.com/74829/amd-ryzen-5-2600-and-asus-rog-crosshair-vii-hero-spotted

I have to admit, 200 MHz is a little... underwhelming. ~5.5% freq uplift.

That said, I suppose if they can run reliably at much higher memory frequencies with good timings then the average user* might see a 10% performance improvement.

*not those on here that have tuned their memory settings to the nth degree.

 

[eek2121]

Interesting. So we can expect plus 200 (or maybe 300MHz on a good day) within the same TDP. That'd take the high-end X's to 4.3GHz XFR. Maybe a bit higher with Precision Boost. Sounds pretty good so far.

Pretty much what I was expecting.

That's an engineering sample, so take it with a grain of salt. However, it lines up with exactly what I've been saying all along, 10% boost in clocks for all models. I bet we are going to see higher boost clocks for the final product, since we now get single core boost.

One interesting tidbit that someone at AMD was talking about was that Precision boost 2.0 would boost as high as TDP would allow. This means that with a single core, you should actually see higher boost clocks for single core...possibly as high as 4.7 GHz if the process allows for it.

EDIT: Oh and I can get my Threadripper to boot at 4.45 GHz on 8 cores at the current process. It's not stable due to the fact that the board can't handle power delivery that well (the system shuts off when I put load on the machine, the lights on the board and everything stay lit, but the GPU shuts down, etc.). Stay away from the MSI boards. The reason I bring this up is that a slightly refined process would take this to 4.9 GHz.

 

[Insert_Nickname]

That's an engineering sample, so take it with a grain of salt. However, it lines up with exactly what I've been saying all along, 10% boost in clocks for all models. I bet we are going to see higher boost clocks for the final product, since we now get single core boost.

Since it is an ES, I already do take it with a large grain of salt. But since it lines up so well with expectations, and what AMD has stated, I think it is not that far off the mark.

One interesting tidbit that someone at AMD was talking about was that Precision boost 2.0 would boost as high as TDP would allow. This means that with a single core, you should actually see higher boost clocks for single core...possibly as high as 4.7 GHz if the process allows for it.

EDIT: Oh and I can get my Threadripper to boot at 4.45 GHz on 8 cores at the current process. It's not stable due to the fact that the board can't handle power delivery that well (the system shuts off when I put load on the machine, the lights on the board and everything stay lit, but the GPU shuts down, etc.). Stay away from the MSI boards. The reason I bring this up is that a slightly refined process would take this to 4.9 GHz.

Ryzen performance at 4.2/4.3GHz should be a somewhat known quantity, because of TR. So any real gains would be whatever low hanging fruit AMD has picked. If precision boost can get one or two cores up to 4.7GHz, it would be quite impressive, and should silence any doubts about single thread performance.

I almost can't wait to see these things used for Threadripper Mark 2. That'll be something.

 

[eek2121]

Some interesting takeaways here, Sandra SUCKS at detecting hardware. I set my threadripper to 6 cores/12 threads/4GHz fixed and it detected the speed as 3.4 GHz for the test. Due to this, it's nearly impossible to tell the average boost speed, or how effectively boost works. At 4.0 GHz (albeit with quad channel RAM) my 4.0 GHz 6 core 12 thread threadripper beats this chip by around 11%. Part of that may be due to the fact I have quad channel RAM as well. When I tried using regular CPU boost (which defaults to 3.7 GHz-4.1 GHz and only drops to 3.4 GHz when thermals get out of whack), I scored lower than this chip. I have some work to do, but for funsies, I'm going set my clock speeds to 3.8 and run one more time.

 

[eek2121]

So, fun fact, the listing in question is equivalent to a Threadripper 1950x with 6 cores disabled @ 3.7 GHz and 32 GB Quad Channel 2400 MHz CL14 DDR4 RAM. I have 1 more test to run before I quit. I'm going to set my clock speed back to auto, and see if sandra pulls down boost clocks to minimums. If it does, that could mean we have a considerable IPC increase on our hands.

EDIT: Oh and if anyone with a Ryzen 5 1600 or 1600x wants to try and reproduce these results, feel free to post.

EDIT: My first test with boost clocks was incorrect. My boost clocks were actually disabled, I should have tested this before starting.

Final Edit: While the chip performs nearly identically to my chip with quad channel RAM, it does NOT perform identically to a 3.7-3.8 GHz 1600/1600x. It is actually quite a bit faster from the results I've looked at so far. This means that they have reduced the latency of the L2 cache, and may have other improvements as well. The quad channel RAM might be a very big factor in the sandra tests, so someone with a 1600 would really have to run sandra to see what type of result they get. Also, sandra appears to be reporting other people's clock speeds correctly...unsure why mine is not being reported correctly.

 

[PeterScott]

Interesting. So we can expect plus 200 (or maybe 300MHz on a good day) within the same TDP. That'd take the high-end X's to 4.3GHz XFR. Maybe a bit higher with Precision Boost. Sounds pretty good so far.

Pretty much what I was expecting.

I wouldn't be surprised if it is less than that. It is probably like the Intel Kaby Lake improvement. Higher clock rates at midrange, but the top speed really doesn't move much at all.

 

[Jan Olšan]

That's an engineering sample, so take it with a grain of salt. However, it lines up with exactly what I've been saying all along, 10% boost in clocks for all models. I bet we are going to see higher boost clocks for the final product, since we now get single core boost.

Look at the code, this thing is a qualification sample (QS), not an ES. Which means these clocks are likely final specs for one particular SKU. Unless it is fake. QS chips should match final SKUs, unlike ES.

 

[french toast]

Could be that these clocks are there for segmentation purposes...2800x may well get more MHz but consumed more power than 95w.
Maybe 200-300mhz xfr2.

 

[Mockingbird]

Look at the code, this thing is a qualification sample (QS), not an ES. Which means these clocks are likely final specs for one particular SKU. Unless it is fake. QS chips should match final SKUs, unlike ES.

QS samples of Ryzen 3 2200G and Ryzen 5 2400G only ran at 3.0/3.3 GHz and 3.3/3.5 GHz respectively.

ZD3000C5M4MFB_33/30_Y

ZD3300C5M4MFB_35/33_Y

 

[exquisitechar]

Well, that's disappointing. I hope the OC ceiling and the top models (2600x, 2800x, etc.) will improve the clock speeds a bit more than that.

 

[neblogai]

QS samples of Ryzen 3 2200G and Ryzen 5 2400G only ran at 3.0/3.3 GHz and 3.3/3.5 GHz respectively.

ZD3000C5M4MFB_33/30_Y

ZD3300C5M4MFB_35/33_Y

Also, Ryzen 5 1600: ZD3301BBM6IF4_37/33_Y

 

[Topweasel]

Could be that these clocks are there for segmentation purposes...2800x may well get more MHz but consumed more power than 95w.
Maybe 200-300mhz xfr2.

Unlikely they have a CPU go over 95w in this market. Unlike FX, they have OEM customers to consider. They won't like changing Specs on a platform that is supposed to be stable for several years.

 

[Topweasel]

Well, that's disappointing. I hope the OC ceiling and the top models (2600x, 2800x, etc.) will improve the clock speeds a bit more than that.

That's the thing. I don't know what people were expecting but no matter what the new process had to provide in max clocks we aren't going to see at stock. Specially on something that is supposed to be a 65w unit. Maybe there is more headway to 95w. I would expect that OC will have much more headroom this gen.

 

[neblogai]

Unlikely they have a CPU go over 95w in this market. Unlike FX, they have OEM customers to consider. They won't like changing Specs on a platform that is supposed to be stable for several years.

They could do like Intel did with the i7 8700: increase all core boost to what is unsustainable at 95W TDP at 100% load. So that games- which usually load CPU at 50% or less- fit into the 95W, and perform very well. But when CPU is loaded 100% on all cores- clocks drop to fit into those same 95W.

 

[snarfbot]

That's the thing. I don't know what people were expecting but no matter what the new process had to provide in max clocks we aren't going to see at stock. Specially on something that is supposed to be a 65w unit. Maybe there is more headway to 95w. I would expect that OC will have much more headroom this gen.

that would be pretty awesome though, as even 4ghz wasn't guaranteed for ryzen 1

 

[Topweasel]

They could do like Intel did with the i7 8700: increase all core boost to what is unsustainable at 95W TDP at 100% load. So that games- which usually load CPU at 50% or less- fit into the 95W, and perform very well. But when CPU is loaded 100% on all cores- clocks drop to fit into those same 95W.

They already do that with XFR. Which I assume we will get a better version of this. But instead of pulling an Intel and having a configuration that only allows for 2 minutes of boost clock, let it rely completely on cooling, if CPU is temp is lower than X, boost, if X or higher, no boost. They hopefully aren't going to change max non-XFR to exceed specced power usage.

 

[Mockingbird]

They could do like Intel did with the i7 8700: increase all core boost to what is unsustainable at 95W TDP at 100% load. So that games- which usually load CPU at 50% or less- fit into the 95W, and perform very well. But when CPU is loaded 100% on all cores- clocks drop to fit into those same 95W.

It was introduced in Raven Ridge as Precision Boost 2.

https://community.amd.com/community/gaming/blog/2017/11/27/asdasd

 

[neblogai]

It was introduced in Raven Ridge as Precision Boost 2.

https://community.amd.com/community/gaming/blog/2017/11/27/asdasd

It was- but it still says Precision Boost 2 "continues to boost—on any number of cores— until reaching the maximum clockspeed printed on the box". I'm all for it- but AMD is likely to set the limit (clockspeeds printed on the box) at what was tested to be achievable at said TDP at 100% load. Instead- if Zen+ is capable to clock higher at higher power use than TDP- they could follow Intel's example and set maximum boost to go even higher when load is less than 100%, and still fit into TDP.

 

[Jan Olšan]

QS samples of Ryzen 3 2200G and Ryzen 5 2400G only ran at 3.0/3.3 GHz and 3.3/3.5 GHz respectively.

ZD3000C5M4MFB_33/30_Y

ZD3300C5M4MFB_35/33_Y

Hmm, they do start with Z, but they don't sound like QSes for the APU SKUs to me.
For one, they would have to contain that BBM string (65W, AM4) like this one. C5M could perhaps mean 35W chips for AM4 (2200S, 2400S), if I was to guess blindly?

 

[Mockingbird]

Hmm, they do start with Z, but they don't sound like QSes for the APU SKUs to me.
For one, they would have to contain that BBM string (65W, AM4) like this one. C5M could perhaps mean 35W chips for AM4 (2200S, 2400S), if I was to guess blindly?

C5 is Raven Ridge

M is AM4 socket