AMD Raven Ridge 'Zen APU' Thread

Page 62 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

mikk

Platinum Member
May 15, 2012
2,415
38
126

Dayman1225

Senior member
Aug 14, 2017
877
77
96
This is the slowest i7-8550U I have seen. No wonder it's from AMD. I wonder why they didn't tell anything about its rated TDP and is it true that the new Vega GPU is only up to 28% better than 7th Gen APU? And what about VP9 decoding in hardware, is it supported finally?
For CB I noticed they omitted the laptop the 2700u used. I checked an XPS 13 review for 8550U and it was scoring 570-620 in CB, I think this is a 25-30w part. The IGPU is better than what we've seen from anyone but I feel like it isn't a massive jump - possibly focusing on efficiency?
 

PhonakV30

Senior member
Oct 26, 2009
944
15
136
For CB I noticed they omitted the laptop the 2700u used. I checked an XPS 13 review for 8550U and it was scoring 570-620 in CB, I think this is a 25-30w part. The IGPU is better than what we've seen from anyone but I feel like it isn't a massive jump - possibly focusing on efficiency?
570-620 ? Videocards says 498
 
Oct 14, 2003
6,068
206
126
Overwatch at 720P/low/79% resolution getting 66FPS seems really low.

the MX 150 (mobile GT 1030) gets 157FPS with 100% render scale
https://www.notebookcheck.net/NVIDI...pecs-of-the-GT-1030-for-Laptops.223530.0.html
The use of 3DMark Time Spy is suspicious. It's not a good benchmark when compared to games. Firestrike would have been better, or 3DMark11.

From Overwatch and DOTA 2, the results seem comparable to Iris 540/640(15W). That's actually fantastic considering Iris needs an expensive eDRAM to do so.

The CPU is performing really well. Single thread performance is within 10% of Intel CPUs, and multi-thread performance kills the Coffeelake U part. Although the success of beating a 15W part would be dependent on whether 2700U is running at 15W or close to 30W to do so, the fact is it beats it soundly so there's a chance its competitive at 15W too. 2700U Cinbench MT scores are on par with 45W Intel quad core chips.
 
Oct 14, 2003
6,068
206
126
That's actually pretty bad considering its 704 Vega ALUs.
Not at all. People who are less experienced tend to believe things will happen even though real barriers prevent it from doing so. To a certain point you can go around that, but eventually it'll tell you "STOP!
 

mikk

Platinum Member
May 15, 2012
2,415
38
126
For CB I noticed they omitted the laptop the 2700u used. I checked an XPS 13 review for 8550U and it was scoring 570-620 in CB, I think this is a 25-30w part. The IGPU is better than what we've seen from anyone but I feel like it isn't a massive jump - possibly focusing on efficiency?

Most are in the range of 550 points. I haven't see a i7-8550U notebook that scores below 500, so AMD made sure they used a subpar performing i7-8550U coupled with cherry picked benchmarks. Cinebench, Povray, Truecrypt...benchmarks where Zen performs above average.

GPU seems to be surprisingly slow. It's not a huge leap forward that I expected if the up to 28% figure turns out to be true.
 
Oct 14, 2003
6,068
206
126
Most are in the range of 550 points. I haven't see a i7-8550U notebook that scores below 500, so AMD made sure they used a subpar performing i7-8550U coupled with cherry picked benchmarks. Cinebench, Povray, Truecrypt...benchmarks where Zen performs above average.

GPU seems to be surprisingly slow. It's not a huge leap forward that I expected if the up to 28% figure turns out to be true.
And...? That's a 30% advantage in CPU, because Coffeelake 4 cores have to throttle. They've got it on the CPU part.

It is a leap forward for a GPU. Because to get Iris configurations they are hard to find, and often configurations that put it $200-300 above the base one. You are spending $300 for an iGPU, think about that. Tick, for a very good GPU.
 
Feb 6, 2011
1,826
183
136

majord

Senior member
Jul 26, 2015
308
1
86
GPU seems to be surprisingly slow. It's not a huge leap forward that I expected if the up to 28% figure turns out to be true.
That's 128%, not 28% , as irrelevant as timespy is

As for Iris and MX-150 comparisons. Well, If this is a 15wTDP, then still seems very good.. Combined TDP of CPU+MX150 sure isn't 15w.. Not even 20-25w. Will wait for actual testing before really giving it much thought though.. Overwatch is the only real 'anomaly' (i.e seems lower than expected, esp on a marketing slide)
 

Dayman1225

Senior member
Aug 14, 2017
877
77
96
And...? That's a 30% advantage in CPU, because Coffeelake 4 cores have to throttle. They've got it on the CPU part.

It is a leap forward for a GPU. Because to get Iris configurations they are hard to find, and often configurations that put it $200-300 above the base one. You are spending $300 for an iGPU, think about that. Tick, for a very good GPU.
We've got actual reviews of the Kaby Lake R lineup, that shows how it acts - we do not have that for RR, I suspect this is best case (Cold boot, didn't mention laptop either - dual channel and 2400+MHz) though it looks promising so far.

But yes the IGPU is much more powerful than what Intel currently offers
 
Apr 27, 2000
11,827
1,042
126
Maybe it would be a good idea to take step back and chill too: this debate is quickly approaching the edges of nonsense land.

Stop projecting our enthusiast skewed views and our gamer habits in this discussion.
Um what? Look man, AMD has a major problem in the laptop/notebook sector, and that is the perception that they are second-rate in nearly every possible way. All it takes is someone trying to use a cheap laptop to run a game or two when they will notice that the graphics performance is maybe not so great. How much performance will Raven Ridge lose in a single-channel configuration? At the present, I just don't know. But I can tell you that all APUs are bandwidth-starved from the get-go. Well AMD's are anyway. You lose 10-20% performance on someone's PUBG experience (or worse) and it's a "game-changer" for that user.

You don't have to have an "enthusiast perspective" or reside in "nonsense land" to realize that shaving maybe $3 off the total cost of the laptop to deliver a sub-standard experience to the end user is an awful, awful idea for everyone involved. Some idiot beancounters just don't get it.

When they are trying to shave a few dollars off, it matters. A 2 x 4 GB kit is a little more expensive than a 1 x 8GB kit, and, it means that the laptop must have two DIMM slots. Lots of cost savings there.
Actually, I would think that the OEM would just have one board layout for an entire segment of their laptop line. Since they will sell some configurations with 2x8 GB RAM, they will probably recycle the same board and use it in their cheaper products that are only getting 8 GB of RAM. As many others have noted in this thread, past notebooks have generally allowed a RAM upgrade through the addition of a second 8 GB SODIMM. It goes to show that OEMs have recently sold low-end laptops with 2 SODIMM slots. I see no reason why they would stop now.

Also - consumers do not know what single channel vs dual channel is. They know that 8GB is better than 4GB, but not that dual channel is better than single channel.
They understand crap performance when they (or their kid) decide to push the little budget lappie out of its intended workload range by making it play some Steam title. Which is going to happen, often. Dude comes along with his dGPU-equipped laptop (and there are many in the sub-gamer category) and whoops up on the poor little AMD machine. When this happens, it's AMD missing out on the opportunity to really impress someone. If you can take your discount laptop and make it game "okay"-ish then maybe it raises a few eyebrows. If - as expected - it flounders about with poor/unplayable framerates, nobody is impressed.

So, dual channel memory is something you cannot even market to the average consumer and it costs a few dollars. No wonder they cut it back. AMD might not have the force to be able to say that they do not support single channel memory on the higher end Raven Ridge chips, which is what they should do.
The point isn't to market dual-channel memory configurations. The point is to give the end-user the best experience possible within the price range set for the product.

I do wonder if graphics performance doesn't matter a little bit more than it used to - not in terms of games but mobile phones and things have meant that people paying real premium prices kind of expect everything to be premium.
I don't know that it's MORE important, but with people ditching desktops for mobile devices, I do think that it could be more important. As in, there's no way they're going to put down their lappie and say, "Gee, I guess I'd better get on the desktop to play that f2p game I wanted to check out on Steam". They're just gonna fire it up on their laptop/2-in-1 and give it a whirl.
 
Oct 14, 2003
6,068
206
126
That's 128%, not 28% , as irrelevant as timespy is
That is 28%. 128% means, 1.28x.

How am I sure of that? Because they claim 200% for CPU. Based on what you are saying, 200% should mean 3x, and yes that's the correct way of saying it, except its not what they are saying.

POV-Ray 3.7 result on VC shows 1.99x, or practically 200% of the score.

I don't know that it's MORE important, but with people ditching desktops for mobile devices,
People are ditching Desktops AND Laptops for mobile devices. Intel share ratio between Desktop and Laptop isn't drastically different than few years ago.
 

mikk

Platinum Member
May 15, 2012
2,415
38
126
And...? That's a 30% advantage in CPU, because Coffeelake 4 cores have to throttle. They've got it on the CPU part.

The up to 28% figure is about GPU not CPU. And there is no Coffeelake 4C ULV, this is KBL-R.
 

Abwx

Diamond Member
Apr 2, 2011
8,870
213
126
Most are in the range of 550 points. I haven't see a i7-8550U notebook that scores below 500, so AMD made sure they used a subpar performing i7-8550U coupled with cherry picked benchmarks. Cinebench, Povray, Truecrypt...benchmarks where Zen performs above average.
.
Nothing cherry picked but rather the 8550U being set at a fixed 15W with PL2 disabled, otherwise we are in desktop TDP range at 50W..

That being said RR should score 510 pts at 15W in CB R15.


https://www.notebookcheck.net/Dell-XPS-13-i7-8550U-QHD-Laptop-Review.257650.0.html
 

Yakk

Golden Member
May 28, 2016
1,574
0
81
This is looking mighty interesting! Can't wait to see how it really does!
 
Oct 14, 2003
6,068
206
126
The up to 28% figure is about GPU not CPU. And there is no Coffeelake 4C ULV, this is KBL-R.
Coffeelake Kabylake Refresh: That part is my bad

You said 550 for 8550U. 2700U gets ~700. That's 30%. 28% is about something else. Please take time to read.
 

majord

Senior member
Jul 26, 2015
308
1
86
That is 28%. 128% means, 1.28x.

How am I sure of that? Because they claim 200% for CPU. Based on what you are saying, 200% should mean 3x, and yes that's the correct way of saying it, except its not what they are saying.

POV-Ray 3.7 result on VC shows 1.99x, or practically 200% of the score.
Sorry but I think you're both wrong.

Timespy:

9800p: 400
2700U : 915

915/400 = 1.28x (+128%)

There's nothing to suggest the 200% figure is actually pov-ray.. it could be Cinebench, which they haven't actually listed 9800p performance in the lower graphs, but we know it scores around 240 https://www.notebookcheck.net/AMD-Bristol-Ridge-FX-9800P-Notebook-Processor.181435.0.html

CB15

719/240 = 2.99x (+200%)
Makes more sense (since it gives the impression of a larger uplift than pov-ray)
 
Oct 14, 2003
6,068
206
126
Sorry but I think you're both wrong.

Timespy:

9800p: 400
2700U : 915

915/400 = 1.28x (+128%)

There's nothing to suggest the 200% figure is actually pov-ray.. it could be Cinebench, which they haven't actually listed 9800p performance in the lower graphs, but we know it scores around 240 https://www.notebookcheck.net/AMD-Bristol-Ridge-FX-9800P-Notebook-Processor.181435.0.html

CB15

719/240 = 2.99x (+200%)
Makes more sense (since it gives the impression of a larger uplift than pov-ray)
You know what? You are correct. Those figures correspond much better.

That still puts into question how the GPU will really do.
 

Bouowmx

Senior member
Nov 13, 2016
853
20
116
To reduce variance because of power and cooling, use Intel Core i7-7700HQ 3.4 GHz (like a Core i5-8250U) to compare: 734 cb (median).

Description of "Mobile XFR" is confusing:
Raises sustained clock speeds in premium notebooks with great cooling solutions
Shouldn't a processor do that by itself?
 

coercitiv

Diamond Member
Jan 24, 2014
3,242
602
136
You have ceased looking at this objectively, as I told you before, take a step back and rethink this.

Food for thought: I bought an ultrabook with 8GB of RAM in single channel config, bought 1 more stick at retail pricing (not inflated OEM margins) and was good to go. If you had your way, I would have had to buy a unit with 2x4GB RAM, buy a two channel kit of 2x8GB and then find a buyer for the smaller sticks. Or better yet, pay a premium for a 16GB unit that probably came with an even higher SSD size, inflating the final price even more. How's that for beancounters that just don't get it?

Mobile reviewers will have to take notice of the single channel configs and do comparative testing just like desktop reviewers took notice of 2133-2666 RAM performance on Ryzen systems and did comparative testing with 3200 CL14 RAM. People who buy APU based products to play games will need to consult proper reviews just like everyone else, and buy accordingly. Meanwhile the overwhelming majority who does not intend to game on their 15W TDP APUs will simply enjoy their products and maybe take advantage of the easy RAM upgrade path.

That having been said, can't wait to see the outrage when Zen APU based notebooks show up sporting small Polaris dGPUs.
 

Abwx

Diamond Member
Apr 2, 2011
8,870
213
126
To reduce variance because of power and cooling, use Intel Core i7-7700HQ 3.4 GHz (like a Core i5-8250U) to compare: 734 cb (median).
That s a 45W CPU while this RR is set at 35W, that s 25% better perf/watt, and as said the 8550U has a TDP that extend up to 51W when cool, in NBC tests it s still at 20W after 40mn of stress test, so these mobile chips apparent perfs are to be taken with a bucket of salt...
 


ASK THE COMMUNITY

TRENDING THREADS