Support NeoGAF

[Durante]

It's gaf, so this talk is normal. For a large group of people, though, the question of 1{7|8}00 vs whatever intel has in that price range simply does not stand - rysen's core advantage is absolutely decisive /MT represent

Well, I love /MT, (and its big brother make -j ), but if you are in that situation then usually the cost of your development hardware is not such a big deal compared to the cost of you waiting for it. As such, AMD would have to actually be faster and not just cheaper.

FWIW, in that respect I find the leaked pricing (if true) somewhat discouraging. If they were fully competitive on a performance basis I would expect somewhat higher prices. We'll see soon enough.

 

[DieH@rd]

Excited to see some benchmarks with recent releases but I only just found out they're stating no support for Windows 7. Isn't that bad news considering that's still the most common OS?

January 2017 Steam stats:
Windows 10 64 bit 48.49%-0.48%
Windows 7 64 bit 29.74%+0.72%
Windows 8.1 64 bit 8.14%-0.31%
Windows 75.37%+0.14%
Windows XP 32 bit 1.24%+0.10%
Windows 101.18%-0.04%
Windows 8 64 bit 1.07%-0.07%
Windows 8.10.26%0.00%
Windows Vista 32 bit 0.11%-0.02%
Windows 80.10%0.00%
Windows Vista 64 bit 0.06%0.00%

 

[daninthemix]

Many doesn't imply all. Some games are threaded well.

Yeah, but most people have one computer with one CPU - so they have to choose between more IPC or more threads, and basically which one will provide the better performance for the majority of games they play.

 

[Kayant]

January 2017 Steam stats:
Windows 10 64 bit 48.49%-0.48%
Windows 7 64 bit 29.74%+0.72%
Windows 8.1 64 bit 8.14%-0.31%
Windows 75.37%+0.14%
Windows XP 32 bit 1.24%+0.10%
Windows 101.18%-0.04%
Windows 8 64 bit 1.07%-0.07%
Windows 8.10.26%0.00%
Windows Vista 32 bit 0.11%-0.02%
Windows 80.10%0.00%
Windows Vista 64 bit 0.06%0.00%

*Sample size not known
*Most other OS marketshare surveys but 7 as still being the dominant platform.
Picture of the dominant platform isn't too clear.

Excited to see some benchmarks with recent releases but I only just found out they're stating no support for Windows 7. Isn't that bad news considering that's still the most common OS?

Likely would even if it wasn't (depending on what surveys you go with) it's still a significant market to "ignore". Although whilst their are no official drivers they said it was validated on both for what it's worth.

 

[Lagspike_exe]

Well, I love /MT, (and its big brother make -j ), but if you are in that situation then usually the cost of your development hardware is not such a big deal compared to the cost of you waiting for it. As such, AMD would have to actually be faster and not just cheaper.

FWIW, in that respect I find the leaked pricing (if true) somewhat discouraging. If they were fully competitive on a performance basis I would expect somewhat higher prices. We'll see soon enough.

Or they want to gain back a decade of lost market share. Positioning their products on performance/price axis so close to Intel makes absolutely zero sense. AMD needs market share.

 

[Durante]

Or they want to gain back a decade of lost market share. Positioning their products on performance/price axis so close to Intel makes absolutely zero sense. AMD needs market share.

Sure, close to Intel would be a mistake. But the current prices are closer to half the Intel price per core. That is either a very bold market share move or an indicator that per-core performance is not as close as we would like. We'll see soon enough.

 

[Aztechnology]

Sure, close to Intel would be a mistake. But the current prices are closer to half the Intel price per core. That is either a very bold market share move or an indicator that per-core performance is not as close as we would like. We'll see soon enough.

I'm assuming you're correct here. But I'd love to eat crow.

 

[Coulomb_Barrier]

These chips actually aren't that expensive to produce, so if AMD wants to undercut Intel like that, they can do it.

Goes without saying though that they need to recoup R & D costs, which are significant to them for this specific line of processors, unlike Intel, who have basically been refining the same processor since Sandy Bridge.

 

[Deleted member 59090]

Excited to see some benchmarks with recent releases but I only just found out they're stating no support for Windows 7. Isn't that bad news considering that's still the most common OS?

Intel doesn't support W7 on Kaby Lake either AFAIK. It's just Microsoft pushing for people to move over to W10.

 

[Myoclonic Jerk]

My last PC build was and AMD 64 (something) in 2000 or something, then switched to mac

Just few weeks ago I decided to build a new PC and get into PC gaming again. Seems like it's going to be an AMD

 

[kraspkibble]

the prices don't seem too great...

an R7 1700X is £375 and the other image says £283 excluding VAT so add 20% to that and it'll be about £340. is that supposed to be on par with a 7700K? those cost £340.

i don't feel the need to upgrade my 6700K any time soon but if AMD actually do well then i will seriously consider them in the future.

 

[LelouchZero]

the prices don't seem too great...

an R7 1700X is £375 and the other image says £283 excluding VAT so add 20% to that and it'll be about £340. is that supposed to be on par with a 7700K? those cost £340.

i don't feel the need to upgrade my 6700K any time soon but if AMD actually do well then i will seriously consider them in the future.

The 1700X is rumored to be an 8 core processor which is pretty crazy to see if it performs close to the Intel chips.
The 6900K is Intel's Broadwell-E 8 core 16 thread processor and it's MSRP is around $1000.

Sure, close to Intel would be a mistake. But the current prices are closer to half the Intel price per core. That is either a very bold market share move or an indicator that per-core performance is not as close as we would like. We'll see soon enough.

Indeed, I'm really looking forward to seeing the performance and a confirmation of the pricing.

 

[rav]

Sure, close to Intel would be a mistake. But the current prices are closer to half the Intel price per core. That is either a very bold market share move or an indicator that per-core performance is not as close as we would like. We'll see soon enough.

This is precisely why I'm still waiting for more hard facts.

 

[TC McQueen]

If the ####X CPUs are the ones that have the auto-overclocking ability, I'll probably go for the 6c/12t version instead of the lowest 8c/16t. I could do with less hassle in my life when it comes to overclocking.

 

[Lonely1]

Well, I love /MT, (and its big brother make -j ), but if you are in that situation then usually the cost of your development hardware is not such a big deal compared to the cost of you waiting for it. As such, AMD would have to actually be faster and not just cheaper.

Filthy rich Australian Computer Scientists...

 

[Thraktor]

I'm really interested in how well Ryzen works in rendering (which is likely to be at least respectable, given they used Blender as their only comparison point thus far). The R7 1700 could be perfect for a low-cost distributed rendering setup, at barely over €300 ex VAT for a 3GHz+ 8C/16T CPU and being able to use a cheap A320/B350 motherboard. Per-node licensing costs would usually push people up to higher core count multi-socket nodes, but the sheer cost savings here would more than counteract it. Plus with a surprisingly low 65W TDP you could squeeze them into 1U cases and still get decent density.

FWIW, in that respect I find the leaked pricing (if true) somewhat discouraging. If they were fully competitive on a performance basis I would expect somewhat higher prices. We'll see soon enough.

It's possible that they're accounting for an Intel price drop in advance. Intel have plenty of scope in their margins to drop prices (or do things like the rumoured enabling of hyper threading in i5's), but there's likely only so far they would be willing to go to keep AMD out. AMD have effectively zero marketshare in this segment, so they have a lot less to lose by taking a low-margin approach to gain marketshare.

 

[blu]

Well, I love /MT, (and its big brother make -j ), but if you are in that situation then usually the cost of your development hardware is not such a big deal compared to the cost of you waiting for it. As such, AMD would have to actually be faster and not just cheaper.

FWIW, in that respect I find the leaked pricing (if true) somewhat discouraging. If they were fully competitive on a performance basis I would expect somewhat higher prices. We'll see soon enough.

AMD are clearly after the market share. And they've been running off fuel vapours for some time now (i.e. console royalties), so that'd actually be an improvement to their bottom line.

 

[TC McQueen]

I'm really interested in how well Ryzen works in rendering (which is likely to be at least respectable, given they used Blender as their only comparison point thus far). The R7 1700 could be perfect for a low-cost distributed rendering setup, at barely over €300 ex VAT for a 3GHz+ 8C/16T CPU and being able to use a cheap A320/B350 motherboard. Per-node licensing costs would usually push people up to higher core count multi-socket nodes, but the sheer cost savings here would more than counteract it. Plus with a surprisingly low 65W TDP you could squeeze them into 1U cases and still get decent density.

Ryzen's actually going to suffer a bit in the latest version of Blender, because it's got some stuff that benefits more from some of Intel's CPU features:
https://www.pcper.com/news/General-Tech/Blender-Foundation-Releases-278b-Performance

Also, some of the optimizations solve bugs with Intel’s CPU implementation as well as increase the use of SSE 4.1+ and AVX2. Unfortunately for AMD, these were pushed up right before the launch of Ryzen, and Blender with Cycles has been one of their go-to benchmarks for multi-threaded performance. While this won’t hurt AMD any more than typical version-to-version variations, it should give a last-minute boost to their competitors on AMD’s home turf.

 

[blu]

Ryzen's actually going to suffer a bit in the latest version of Blender, because it's got some stuff that benefits more from some of Intel's CPU features:
https://www.pcper.com/news/General-Tech/Blender-Foundation-Releases-278b-Performance

Increased use of SIMD could also benefit ryzen which does AVX2, be that via 128-bit ALUs.

ed: no idea if this has been seen on gaf before, so here it is: https://www.reddit.com/r/Amd/comments/5mklpw/ryzen_vega_bristol_ridge_am4_a_summary_and/

 

[Renekton]

Sure, close to Intel would be a mistake. But the current prices are closer to half the Intel price per core. That is either a very bold market share move or an indicator that per-core performance is not as close as we would like. We'll see soon enough.

The latter was the first thing that came to my mind as well. AMD can't be that generous.

 

[dr_rus]

First AMD Ryzen 7 1700X benchmarks are here

Some more results via the link.

 

[Ozorov]

First AMD Ryzen 7 1700X benchmarks are here



Some more results via the link.

sooooo... is this good?

 

[Aztechnology]

sooooo... is this good?

Lower than the comparative Intel tests. But also keep in mind they're running at much lower frequencies. The Intel processors are insanely OC'd. My 5820k with water cooling and a really good mobo, can barely reach 4.5ghz. It's actually unstable. They have the 7700k running at 5ghz etc.

Edit: Oh RUS put up the other benchmarks as well. Good man.

 

[pa22word]

Lower than the comparative Intel tests. But also keep in mind they're running at much lower frequencies. The Intel processors are insanely OC'd. My 5820k with water cooling and a really good mobo, can barely reach 4.5ghz. They have the 7700k running at 5ghz etc.

Also worth noting that the ryzen chips were running flat stock speed due to turbo being disabled. Also bad memory timings and iffy mobo on the test as well.

 

[Aztechnology]

Also worth noting that the ryzen chips were running flat stock speed due to turbo being disabled. Also bad memory timings and iffy mobo on the test as well.

About to try a test myself just to see how mine compares to their 5820k setup. Grabbed the performance test.

 

[Osiris]

Don't look that good to me, only a small increase over a 4C/8T CPU... that sucks and starts to look a little familiar with respect to AMD overhyped CPU's.

The prices for these had better be bargain basement.

 

[Risk Breaker]

Those are results at 3.4 GHz? Yeah, that looks good. Looking forward to full reviews and jumping ship from 4 cores

 

[opticalmace]

Lower than the comparative Intel tests. But also keep in mind they're running at much lower frequencies. The Intel processors are insanely OC'd. My 5820k with water cooling and a really good mobo, can barely reach 4.5ghz. It's actually unstable. They have the 7700k running at 5ghz etc.

Edit: Oh RUS put up the other benchmarks as well. Good man.

Keep in mind the 7700k overclocks way better than the 5820k. Hitting 5ghz or close (say 4.8+) is pretty normal.

 

[Durante]

It doesn't look too bad to me except for that "Physics" test. But without knowing exactly what that does it's hard to interpret that result.

I'd wait for more real-world tests.

 

[Aztechnology]

Keep in mind the 7700k overclocks way better than the 5820k. Hitting 5ghz or close (say 4.8+) is pretty normal.

No I realize that, they just have the 5820k going at 4.8ghz. That's a real win in the silicone lottery for that CPU to do that from my experience.

 

[TC McQueen]

At first, I was super concerned about the performance, then I noticed that literally every Intel CPU on the list was overclocked. If this isn't the definition of an unfair test, it's damn close.

For stuff like this to be worthwhile, they needed to either OC the Ryzen CPU, or if they couldn't do that, get stock clocked Intel CPUs.

 

[blu]

It doesn't look too bad to me except for that "Physics" test. But without knowing exactly what that does it's hard to interpret that result.

I'd wait for more real-world tests.

Keep in mind that compiler zen-specific optimisaitons are yet to come in full swing.

 

[opticalmace]

No I realize that, they just have the 5820k going at 4.8ghz. That's a real win in the silicone lottery for that CPU to do that from my experience.

Ah gotcha. Am I going crazy or are the first set of benchmark images not in that link? I can't seem to find the page they came from.

 

[LelouchZero]

First AMD Ryzen 7 1700X benchmarks are here



Some more results via the link.

Oh very interesting stuff, thanks for posting! Real-word testing will be fascinating!

 

[napata]

It doesn't look too bad to me except for that "Physics" test. But without knowing exactly what that does it's hard to interpret that result.

I'd wait for more real-world tests.

Prime numbers too. Is it possibly because of the lower AVX performance in Ryzen? What would the effects of that be in real life scenarios like gaming?

 

[Aztechnology]

It doesn't look too bad to me except for that "Physics" test. But without knowing exactly what that does it's hard to interpret that result.

I'd wait for more real-world tests.

You have a 5820k as well right? Is it OC'd? My Physics score is lower than expected.

 

[Durante]

You have a 5820k as well right? Is it OC'd? My Physics score is lower than expected.

It's OC'd a bit, running at 4 GHz. I get 1006 in the Physics test.

 

[teh_pwn]

sooooo... is this good?

For PC gamers with a discrete GPU, uniquely competitive value. We're looking at 90% the performance of an i7-6900K (8 cores, $1000) but with 8 cores at $450. Basically you can get nearly double the performance per dollar. Looking at it from another angle, if you were in the market for a high end quad core i7, now you can spend $50 and get 8 cores of nearly the same performance per core. Considering how software is finally starting to work multithreaded in most apps, that's what I would select if I were building a new system.

This would have a lot of potential in the datacenter server market, as balancing core count and per core performance has a lot of applications. Zen would do well against budget systems looking at 2 socket E5s at high frequency. Even if it got 15% of Intel's market, that would be huge for AMD. But unfortunately server tech usually takes 1-2 years to roll out new designs from desktop to server, at least with Intel.

 

[teh_pwn]

Don't look that good to me, only a small increase over a 4C/8T CPU... that sucks and starts to look a little familiar with respect to AMD overhyped CPU's.

The prices for these had better be bargain basement.

The first benchmark listed is how the CPU will perform with the vast majority of modern applications using the CPU for integer operations and using a discrete GPU for video acceleration.

So for old single threaded x86 stuff it nearly holds it's own on a single core. With modern games, encoding, .Net 4.x+ based apps, Windows 10 OS, etc it's 50+% faster than an i7 quad core.

The benchmarks for floating point stuff is more APU. The weird synthetic benchmarks highlighting niche usage of new instruction sets are extremely application specific.

 

[Aztechnology]

It's OC'd a bit, running at 4 GHz. I get 1006 in the Physics test.

I'm at 4.5 and showing 793... I wonder why. Oh well, not having any issues.

 

[AstroNut325]

Those benchmarks look fantastic so far.

I'm switching to AMD if the price is right!

 

[michaelius]

January 2017 Steam stats:
Windows 10 64 bit 48.49%-0.48%
Windows 7 64 bit 29.74%+0.72%
Windows 8.1 64 bit 8.14%-0.31%
Windows 75.37%+0.14%
Windows XP 32 bit 1.24%+0.10%
Windows 101.18%-0.04%
Windows 8 64 bit 1.07%-0.07%
Windows 8.10.26%0.00%
Windows Vista 32 bit 0.11%-0.02%
Windows 80.10%0.00%
Windows Vista 64 bit 0.06%0.00%

That's gaming only - corporate clients are still sitting on windows 7 in majority of cases (at least those who already migrated from xp )

 

[teh_pwn]

That's gaming only - corporate clients are still sitting on windows 7 in majority of cases (at least those who already migrated from xp )

True, but anyone looking at an 8 core system should be using Windows 10 or the latest Linux LTS distribution of their choice.

There are major differences between versions of Windows. Basically most things in XP are x86 only, single threaded, 20 years behind on optimizations. Everything from CIFS, robocopy, Windows update, explorer, CPU<->GPU communication...everything is going to be way slower in an older version of Windows regardless of how good the CPU is.

 

[Durante]

The first benchmark listed is how the CPU will perform with the vast majority of modern applications using the CPU for integer operations and using a discrete GPU for video acceleration.

The first benchmark is a synthetic integer benchmark. It really doesn't represent any common real-world workload I can think of all that well. Especially since real-world integer-heavy workloads are often very branchy.

To know how a given modern application performs on this architecture, you need a benchmark of that modern application. We'll have a lot of those soon enough.

 

[tuxfool]

Prime numbers too. Is it possibly because of the lower AVX performance in Ryzen? What would the effects of that be in real life scenarios like gaming?

I don't know about physics but prime number tests are usually very AVX2 heavy (almost exclusively) and almost always act as power viruses. They are also the leading way to test whether a CPU has an unstable overclock, which is why Kaby Lake added the ability to control the top rate for AVX instructions.

 

[Paragon]

Those PassMark results are somewhat concerning - though it is only a single point of reference right now.
Single-threaded performance is lower than my 2500K @ 4.5GHz (scores 2405) and the physics test is only 60% higher despite having twice as many cores.
While I realize that the system doesn't have the fastest DDR4 RAM, it scores 800 points lower than my system running DDR3 @ 1600MHz CL8 and latency seems incredibly high at 76ns.
It seems to be a good performer in some areas, but has unexpectedly low performance in others. Curious to see how that works out for gaming.

 

[faint.]

***RUMOR***

AMD Ryzen: Intel allegedly briefed its partners with PR phrases
http://www.pcgameshardware.de/AMD-Zen-Codename-261795/News/Ryzen-Intel-Kaby-Lake-S-1220311/

Looks like Intel is training their associates about Ryzen - funny evasions 101
https://www.reddit.com/r/Amd/commen...ing_their_associates/?st=iywwxwd8&sh=b58f7978

Is this is real I can't believe they would expect people asking about fabrication nodes to assume that means equivalent performance.

 

[dr_rus]

The original post was updated with a couple of new tests and a non-OC Intel CPU results btw:

I'll reserve any judgement till we see other benchmarks. It's usually hard to make any parallels between a synthetic benchmark and real world performance. Ryzen does look pretty good here if you consider the leaked price for the 1700X SKU but that was somewhat expected in heavily threaded workloads.

 

[longdi]

3.4Ghz 1700X beating 4.2Ghz 7700K bodes well for AMD and gamers!

Intel x700K are pretty popular amongst gamers, now they all have 8 core goodness, makes 7700K redundant imo.

Im pissed Intel has been quietly raising their x700K prices, generation after generation. I want AMD to smash that $350 price range.

 

[abracadaver]

Will this push AMD stock to $20?