Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[krumme]

It's not about the perception that AMD needs to be cheap at all.
The quad SKUs are lacking in ST without a good reason.
If they had listed clocks similar to the 8 cores SKUs and such quads at 149$, 169$ and 199$, it would be an easy choice to buy AMD and they would sell twice as well.
As they did it, the quads look weak because of ST and XFR not being properly listed.
Few overclockers would buy the 199$ SKU anyway but such a SKU would change the perception of the entire line.
At this price range, getting close to Intel in ST was achievable and AMD failed to do so.That's all they needed to do, push ST up a bit more.
The 6 cores are great but they messed up the quads.

Dont get carried away from those graphs. AT bm suite is methology from the stoneage and gives the wrong impression.
Who runs a single piece of software where in real world testing you are solely ST limited on a ryzen?

The cpu loads we have today is limited by a mix of st and mt or mt loads.
Yet they have a suite where its solely st limited and not mt limited but a mix. Their mt suite is not a mt suite. Its a mixed suite.

We still have a handfull of taxing newer games that rely strongly on st and memory where zen gets below 90 fps. Is it 4 or 5 games of relevance?
Outside of that no professional nor comsumer will in 99.9% be st limited with the perf a ryzen gives.

The way AT present it had its practical relevance when bd was on the market. Its not reflecting user load today.

 

[french toast]

His point is that the numbers don't tell the whole story. Many have already reported that Ryzen seems to remove the stutter from games in comparable playthroughs. More and more videos seem to be coming out of people showing a deficit in FPS with Ryzen, but a better overall actual gameplay experience because the play is buttery smooth compared to somewhat choppy performance with Intel. This doesn't seem to bear out well enough with the "accepted" numbers game, afaik.

Yes ive heard that, there is just so much rubbish going around ATM, ryzen is good for gaming, not the best but very good for a number of reasons, smoothness being one of them.

 

[krumme]

A 1700 with 1 CCX disabled as a 4+0 is going to be better in 99%+ of games than any 2+2 or even 3+3.
That really ticks me off. I don't want to spend more than I need to for twice the cores over a 1400 or 1500X just to have to disable cores and restart my computer to get the most out of some games and the performance that the Ryzen 4 cores SHOULD have.
You're talking an average 20% difference between 2+2 and 4+0 in games, when 4+0 is already going to be 10-20% behind the 7700k.

No, I didn't expect a 50% better CPU for half the price, but I'd have hoped for a 4c/8t that's 50-65% the cost of a 7700k and 80-90% the performance. Which is what a 1700 with 1 CCX disabled would have been.

Yes AMD needs to "maximize yields" so what are they doing to CPUs with 3 cores defective on one side, but all 4 cores fine on the other?

They needed to release a 4+0 part in the R5 series to truly compete.
The 1600X and 1500X both don't compete. They'll both be slower in many games than a 6 year old i5-2500 or the 2c/4t i3-7350K because of the CCX issue that virtually no game accounts for.

Hell I'd have been happy to pay $200+ for a 4c/8t that's actually 4 cores on one CCX without that CCX issue. I don't want to wait/hope that the issue is fixed in Windows scheduler and game patches for years old games.
Even at 3.5/3.9ghz and its lower IPC, it still probably would have been faster than the i5-7600k in most games thanks to SMT. But no, we get 2+2s that'll be worse than an Intel 2 core in gaming.

I'd bet thousands of dollars that the average performance for the 1500X is going to be worse than the i3-7350K across any decent average selection of games that reviewers commonly use, if there was any legit betting pool going on. That's how bad this is.
It's not just bad, it's sad. So sad and stupid that AMD didn't prioritize salvaging 4+0 cores and just sold the 2+2 to OEMs or something.


... or they're being sold as cheaper R3s later this year?

Much drama for 5% difference on 120fps.

Do you guys lack historical memory?
Any 6c will beat any 4c in bf1 mp for min in the new maps we have no matter the name on it.
Thats where is heading and its happening today.
Do you think bf in 1.5 year will be a lighter load?

 

[krumme]

I believe you but may i see numbers ?

Here is you numbers. Ryzen kills core $vs$
4c4t is damn bad. 8c rocks.
http://www.pcgameshardware.de/Battlefield-1-2016-Spiel-54981/Specials/Battlefield-1-They-Shall-not-Pass-Benchmarks-1223170/

And that is with 2800 ram for the 3.9zen in 4x configuration. That gives slow sub timings. You can easily go 2933 if you buy the right ram and use 2 sticks.
If you need help to select go to the builders thread.
Obviously is important to get fast ram with zen but thats also due to latency issues.
Get some trident that works at 3200 at solid cl.

 

[lolfail9001]

Here is you numbers. Ryzen kills core $vs$

That's not what i am asking, though, read the context.

Besides, BF1 is hardly representative of anything.

 

[beginner99]

What I mean is what if each 2 core CCX had access to 4MB+4MB with the second being an exact copy of the other 4MB on its sister CCX across the way so that the cross CCX latency is eliminated by keeping the cache mirrored

There are 2 hard things in computer science. Naming things and cache invalidation. Keeping these 4 mb caches in sync might be more difficult and take more bandwidth than just syncing when it's needed.

 

[krumme]

That's not what i am asking, though, read the context.

Besides, BF1 is hardly representative of anything.

Its representive of how future gaming is. Have been the last 6 years.
People that bought core2 quad could play bf3 kind of acceptable. People that had dual core could not.
Here is the link in case you lost it.
http://www.pcgameshardware.de/Battlefield-1-2016-Spiel-54981/Specials/Battlefield-1-They-Shall-not-Pass-Benchmarks-1223170/

 

[beginner99]

The cpu loads we have today is limited by a mix of st and mt or mt loads.
Yet they have a suite where its solely st limited and not mt limited but a mix. Their mt suite is not a mt suite. Its a mixed suite.

Maybe the CPU load you have as an enthusiast is a lot of MT, but for most users, its not. A fast dual-core is mostly enough for office and web browsing. Heck my work laptop is a hyper-threaded ivy bridge dual core and I did some pretty heavy computing stuff on it and you know what? The limiting thing is our stupid IT department that only offers 32-bit windows and hence only about 3 GB usable RAM. I also run a software package that cost several 10k a year on it and you know what? that POS software is 100% single-threaded even were it easily could be multi-threaded.

point is the difference between average user and enthusiast is getting bigger and bigger. AMD needs volume. They should target the average user.

 

[krumme]

Maybe the CPU load you have as an enthusiast is a lot of MT, but for most users, its not. A fast dual-core is mostly enough for office and web browsing. Heck my work laptop is a hyper-threaded ivy bridge dual core and I did some pretty heavy computing stuff on it and you know what? The limiting thing is our stupid IT department that only offers 32-bit windows and hence only about 3 GB usable RAM. I also run a software package that cost several 10k a year on it and you know what? that POS software is 100% single-threaded even were it easily could be multi-threaded.

point is the difference between average user and enthusiast is getting bigger and bigger. AMD needs volume. They should target the average user.

All core and ryzen is fast enough for normal office use.
For bd line it made some sense to bm st perf because it was so crappy.
Normal consumers is limited by a mixed load of st and mt and not st perf.
Obviously intel skl is faster in those mixed loads but its marginal and comes either at high cost or a serious hit on number of cores. You get 15% st perf for half the cores. Thats a pretty steep price imo in todays mixed loads for average consumers.
Obviously there is enthusiast cornercases for either arch but i think those segments is actually far bigger for zen.

 

[CatMerc]

I have an uncanny ability to hate products i purchase. I hated all of my phones, i hated myself for buying C2D, and i still hate myself for purchasing a cooler out of habit that i could even mount on my motherboard properly.

But yeah, i do not have Ryzen right now, i have just laid off $400 (damn 20% VAT) for it should ITX motherboards and the stepping with better data fabric clocks come out though.

Out of curiosity, what made you hate your C2D?

 

[Whitestar127]

Many have already reported that Ryzen seems to remove the stutter from games in comparable playthroughs. More and more videos seem to be coming out of people showing a deficit in FPS with Ryzen, but a better overall actual gameplay experience because the play is buttery smooth compared to somewhat choppy performance with Intel.

But did they specifically mention "choppy Intel performance" though? In the (admittedly few) reviews I've seen they mention the smoothness of the Ryzen, but I didn't get the impression that the Intel performance was less smooth or choppy compared to it.

My own experience on my Sandy Bridge is mostly buttery smooth in most games. But I do get stutters in say GTA5 at regular intervals. And I also get a sort of micro stutter in Fallout 4, especially when near objects or just looking at the ground. Not sure if those can be attributed to the CPU or not. Also, in Just Cause 3 I get something that can only be described as micro stutter at 75Hz vsync on mouse/head movement, while it's not there at 60Hz vsync, and strafe action is buttery smooth at both. Go figure. I guess it could be anything from game code to CPU to video drivers.

Ok, that was neither here nor there really, but my point is: Is there really a general consensus that Ryzen is smoother than Intel, or is it just that they want to point out that it's just as good as Intel, and therefore mention how smooth it is in the reviews? If you can point me to some videos where they straight out says that the Intel performance is choppy in comparison, then I'll happily accept that as a clear indicator that it is in fact smoother.

 

[coercitiv]

Ok, that was neither here nor there really, but my point is: Is there really a general consensus that Ryzen is smoother than Intel, or is it just that they want to point out that it's just as good as Intel, and therefore mention how smooth it is in the reviews? If you can point me to some videos where they straight out says that the Intel performance is choppy in comparison, then I'll happily accept that as a clear indicator that it is in fact smoother.

Link here. The reviewer also explains the implications, since this does not translate into an automatic win for Ryzen.

 

[PotatoWithEarsOnSide]

A few review videos have shown Intel quads to stutter at points where Ryzen is uber smooth, and they have commented on this at the time.

 

[imported_jjj]

Dont get carried away from those graphs. AT bm suite is methology from the stoneage and gives the wrong impression.
Who runs a single piece of software where in real world testing you are solely ST limited on a ryzen?

The cpu loads we have today is limited by a mix of st and mt or mt loads.
Yet they have a suite where its solely st limited and not mt limited but a mix. Their mt suite is not a mt suite. Its a mixed suite.

We still have a handfull of taxing newer games that rely strongly on st and memory where zen gets below 90 fps. Is it 4 or 5 games of relevance?
Outside of that no professional nor comsumer will in 99.9% be st limited with the perf a ryzen gives.

The way AT present it had its practical relevance when bd was on the market. Its not reflecting user load today.

It's about how people make the purchasing decision and AMD crippled the ST perf for their quads for no reason.
Have you seen anyone at all excited about the quads? I haven't and that's precisely because the ST clocks are too low and bellow expectations.
They had the opportunity to get close in ST to Intel in the sub 200$ segment and they just missed it.
They win in MT, they have more L3$, longer life for the socket and unclocked but they create a huge headwind with the ST clocks.
It's the difference between not needing to even think about what to purchase (and chose AMD by default) and pondering about it for days.
And remember it's self inflicted, AMD made a major miscalculation when "calibrating" these SKUs and it will be very costly from a financial point of view.

Intel has
7500 3.4/3.8GHz
7400 3.0GHz/3.5GHz

AMD has
1500X 3.5/3.7GHz with 200MHz XFR at 189$
1400 3.2/3.4GHz with 50MHz at 169$

What if AMD had
3.6/4GHz plus 100MHz XFR (or even without as long as they list 4GHz ) at 199$
3.4/3.8 at 179$
3.2/3.4 at 149$

Can you really say that my proposed SKUs would not lead to much better sales?
It not only makes the line a lot more appealing but they would also reach lower price points so quite a bit more customers now not in the second half.
The top 2 SKUs with slightly higher ASPs than their offering, would likely lead to a somewhat similar overall ASP for the R5 quads but much higher volumes.
Just by listing lower turbo clocks than Intel they lose a lot of customers as a large chunk of people will chose on clocks, just like they chose GPUs based on amount of VRAM.
What AMD did is dumb and they need to adjust the quads right now before launch.

 

[tamz_msc]

i5 7400/7500 are locked - you're forgetting that. Plus they don't have SMT.

 

[imported_jjj]

i5 7400/7500 are locked - you're forgetting that. Plus they don't have SMT.

If you read my post again you'll see that i did not forget.
"They win in MT, they have more L3$, longer life for the socket and unclocked"

 

[PotatoWithEarsOnSide]

You fail to put any value on those though.

 

[tamz_msc]

If you read my post again you'll see that i did not forget.
"They win in MT, they have more L3$, longer life for the socket and unclocked"

You can OC to your proposed speeds and achieve parity in ST. Of course, the IPC deficit with respect to Kaby Lake will remain, and how the cores on different CCXs behave will need more testing, but the end result won't be too bad for AMD.

 

[beginner99]

All core and ryzen is fast enough for normal office use.

Yes my point was more so is the new pentium for $64.

 

[beginner99]

Can you really say that my proposed SKUs would not lead to much better sales?

No but do you know AMD can supply enough chips that actually reach 4 ghz on a sane voltage and within sane TDP? bet that is the main issue and also why the 1800x is so expensive. It's the best bin and there aren't enough such chips to offer such a quad-core.

 

[Whitestar127]

Link here. The reviewer also explains the implications, since this does not translate into an automatic win for Ryzen.

A few review videos have shown Intel quads to stutter at points where Ryzen is uber smooth, and they have commented on this at the time.

Thanks! Sounds promising. Smoothness > FPS any day of the week, so this looks good for the Ryzen.

Btw @PotatoWithEarsOnSide, do you remember which videos, apart from the one coercitiv linked to?

 

[imported_jjj]

@ tamz_msc
OC is a niche and they need to be compelling for the broader market.

@ beginner99
Likely yes and at the end of the day it's better to have shortages than to present a weak lineup.
Demand for the 1800X will ease as the wave of early adopters with plenty of $ to spare is saturated.Short term the 1600X would be a headwind but manufacturing improves slightly and they would be fine.
AMD even left room in the naming scheme for a 1900X.

@ PotatoWithEarsOnSide
The value depends on the customer's needs and perception.
Haven't pointed this out before but the weaker than ideal ST clocks will create a negative image in gaming again. The battery of games used by reviewers vary a great deal but plenty use ST dependent games. 2+2 might cripple games that do scale so it's rather messy.



This is about what they could have done vs what they did. Some will argue that the quads don't suck but are those SKUs the best they could have done - from a financial perspective so best for AMD.
Not going bellow 169$ is about protecting the R3 as they seem to want to price those from 129$ and up. 129$ is almost 2x that popular Pentium and not great considering that they won't have dual cores Zen this cycle. IIMO they should extend R3 to 99$ and many folks that would spend less will make an effort to get the 99$ quad.
Likely AMD is worried about Bristol Ridge too ,maybe they have inventory and that's why the R3 is delayed. That's not ideal and BR should be marketed for its GPU.
They need to be a bit bolder and not let fear rule their actions.They got the octa and hexa right so far but not the quads.

 

[french toast]

It's about how people make the purchasing decision and AMD crippled the ST perf for their quads for no reason.
Have you seen anyone at all excited about the quads? I haven't and that's precisely because the ST clocks are too low and bellow expectations.
They had the opportunity to get close in ST to Intel in the sub 200$ segment and they just missed it.
They win in MT, they have more L3$, longer life for the socket and unclocked but they create a huge headwind with the ST clocks.
It's the difference between not needing to even think about what to purchase (and chose AMD by default) and pondering about it for days.
And remember it's self inflicted, AMD made a major miscalculation when "calibrating" these SKUs and it will be very costly from a financial point of view.

Intel has
7500 3.4/3.8GHz
7400 3.0GHz/3.5GHz

AMD has
1500X 3.5/3.7GHz with 200MHz XFR at 189$
1400 3.2/3.4GHz with 50MHz at 169$

What if AMD had
3.6/4GHz plus 100MHz XFR (or even without as long as they list 4GHz ) at 199$
3.4/3.8 at 179$
3.2/3.4 at 149$

Can you really say that my proposed SKUs would not lead to much better sales?
It not only makes the line a lot more appealing but they would also reach lower price points so quite a bit more customers now not in the second half.
The top 2 SKUs with slightly higher ASPs than their offering, would likely lead to a somewhat similar overall ASP for the R5 quads but much higher volumes.
Just by listing lower turbo clocks than Intel they lose a lot of customers as a large chunk of people will chose on clocks, just like they chose GPUs based on amount of VRAM.
What AMD did is dumb and they need to adjust the quads right now before launch.

Are my eyes playing tricks on me?
What my eyes are telling me is intel has a 50mhz clock advantage (vs 1400) and a 100mhz deficit (vs 1500x), what the hell is the problem here?

You admit yourself the advantages of ryzen, wayy more L3 cache, DOUBLE the threads, lower power? Cheaper? Free decent cooler, upgrade ability.
For what penalty compared to i5? 10% lower ST? MAYBE 10-15% lower game performance in some games?

Ryzen 5 is extremely good value for every scenario, even if it gets beat in a couple of areas vs i5, overall its clearly better imo.

 

[imported_jjj]

Are my eyes playing tricks on me?
What my eyes are telling me is intel has a 50mhz clock advantage (vs 1400) and a 100mhz deficit (vs 1500x), what the hell is the problem here?

You admit yourself the advantages of ryzen, wayy more L3 cache, DOUBLE the threads, lower power? Cheaper? Free decent cooler, upgrade ability.
For what penalty compared to i5? 10% lower ST? MAYBE 10-15% lower game performance in some games?

Ryzen 5 is extremely good value for every scenario, even if it gets beat in a couple of areas vs i5, overall its clearly better imo.

The difference is about a billion dollars in revenue this year that AMD misses out on for having lower ST clocks than they could.
And that doubles if you include the short and mid term marketing impact.

 

[french toast]

The difference is about a billion dollars in revenue this year that AMD misses out on for having lower ST clocks than they could.
And that doubles if you include the short and mid term marketing impact.

Right, can i borrow your crystal ball to see for myself?
We have no clue.