Ryzen 2500u benchmark by The Tech Report (img heavy)

[Bit_Guardian]

59 minutes ago, The Benjamins said:

that make no sense, AMD wants a more competitive product that's why.

 

The performance of the new Ryzen mobile APU's are a HUGE step up from the old ones. they are also faster then all of Intels 7th generation parts at the same wattage.

For some, but not all workloads.

 

@cj09beira The bribing is not happening anymore. Otellini and every executive from that time has left the company, many now working for Google. Further, Intel is under surveillance by the FTC for a couple more years as part of its punishment for that before. Your paranoia is baseless.

[Dabombinable]

4 hours ago, NumLock21 said:

I had a amd apu netbook and it ran like crap even with 6gb of ram, ssd, and a clean os. So no, it has nothing to do with Intel. The reason why everyone expect more from it is because of VEGA.

My A8 4555M has lower single threaded performance at 1.8GHz (it never turbos to 2.4GHz for long) than my Pentium 4 HT 3.2GHz (DDR 333 because Ati chipset). They really shouldn't have taken CMT out of the server space.

[Bit_Guardian]

10 minutes ago, Dabombinable said:

My A8 4555M has lower single threaded performance at 1.8GHz (it never turbos to 2.4GHz for long) than my Pentium 4 HT 3.2GHz (DDR 333 because Ati chipset). They really shouldn't have taken CMT out of the server space.

CMT isn't the issue per-se. However, Bulldozer was crap. There need to be separate prefetch and TLB units for each side of a module if you're going to do CMT.

[cj09beira]

3 hours ago, Bit_Guardian said:

For some, but not all workloads.

 

@cj09beira The bribing is not happening anymore. Otellini and every executive from that time has left the company, many now working for Google. Further, Intel is under surveillance by the FTC for a couple more years as part of its punishment for that before. Your paranoia is baseless.

i really hope you are right, we will know in a few years

[cj09beira]

4 hours ago, NumLock21 said:

I had a amd apu netbook and it ran like crap even with 6gb of ram, ssd, and a clean os. So no, it has nothing to do with Intel. The reason why everyone expect more from it is because of VEGA.

ya but that was with a shitty cpu, now it has ryzen on it with is a lot better

[Dabombinable]

11 minutes ago, Bit_Guardian said:

CMT isn't the issue per-se. However, Bulldozer was crap. There need to be separate prefetch and TLB units for each side of a module if you're going to do CMT.

It is when AMD put in 2x ALU instead of 2x FPU, with FPU being far more important for games (and why AMD needed to release the K6-3/K6 3D in the late 90's-the FPU was slow so they needed to invent special instructions to make it work more efficiently). AMD's CMT designs only lends themselves well to primarily server orientated tasks.

[NumLock21]

4 hours ago, The Benjamins said:

that make no sense, AMD wants a more competitive product that's why.

 

The performance of the new Ryzen mobile APU's are a HUGE step up from the old ones. they are also faster then all of Intels 7th generation parts at the same wattage.

Amd want to be more competitive? They gave up on that a long time ago with them apus and ancient fx. Causing intel to milk their products since there is no competition. It just recently they decided to get back into the game and made Ryzen. 

 

3 hours ago, Bit_Guardian said:

For some, but not all workloads.

 

@cj09beira The bribing is not happening anymore. Otellini and every executive from that time has left the company, many now working for Google. Further, Intel is under surveillance by the FTC for a couple more years as part of its punishment for that before. Your paranoia is baseless.

Otellini is dead.

[Prysin]

Just now, Dabombinable said:

It is when AMD put in 2x ALU instead of 2x FPU, with FPU being far more important for games (and why AMD needed to release the K6-3/K6 3D in the late 90's-the FPU was slow so they needed to invent special instructions to make it work more efficiently). AMd's CMT designs only lends themselves well to primarily server orientated tasks.

no, games are not "hard" riding the FPU. FPU math is slow, much slower then ALU math, simply because the code is slower. You want to use as little FPU math as possible if you want a fast application. You only use floating point when you desire absolute accuracy, as you need lots of decimals.

[Bit_Guardian]

24 minutes ago, Prysin said:

no, games are not "hard" riding the FPU. FPU math is slow, much slower then ALU math, simply because the code is slower. You want to use as little FPU math as possible if you want a fast application. You only use floating point when you desire absolute accuracy, as you need lots of decimals.

Please read the x86 manual. floating point mull/div is much faster than integer.

[Prysin]

Just now, Bit_Guardian said:

Please read the x86 manual. floating point mull/div is much faster than integer.

depends on the code. Yes, in hardware action, it is, as there is fewer steps. Program code wise however, from what i know of coding, which is limited, float takes longer.

[Bit_Guardian]

Just now, Prysin said:

depends on the code. Yes, in hardware action, it is, as there is fewer steps. Program code wise however, from what i know of coding, which is limited, float takes longer.

No. Go ahead and benchmark a few loops. You can use Google's benchmark framework to write it if you want.

[Prysin]

1 minute ago, Bit_Guardian said:

No. Go ahead and benchmark a few loops. You can use Google's benchmark framework to write it if you want.

hi patrick

[Bit_Guardian]

Just now, Prysin said:

hi patrick

???

[Trixanity]

18 minutes ago, Prysin said:

hi patrick

You noticed it too. Didn't want to say anything.

[Prysin]

1 minute ago, Trixanity said:

You noticed it too. Didn't want to say anything.

posts was a giveaway, bio sealed the deal

[Bit_Guardian]

2 minutes ago, Trixanity said:

You noticed it too. Didn't want to say anything.

 

1 minute ago, Prysin said:

posts was a giveaway, bio sealed the deal

First, name's Marcus. Second, who?

[The Benjamins]

1 hour ago, NumLock21 said:

Amd want to be more competitive? They gave up on that a long time ago with them apus and ancient fx. Causing intel to milk their products since there is no competition. It just recently they decided to get back into the game and made Ryzen. 

 

Otellini is dead.

recently? you mean the 4-5 years of development of zen and them having to scrape all other work to get this to market as soon as possible. this was the only way not to give up.

[Dabombinable]

1 hour ago, Prysin said:

no, games are not "hard" riding the FPU. FPU math is slow, much slower then ALU math, simply because the code is slower. You want to use as little FPU math as possible if you want a fast application. You only use floating point when you desire absolute accuracy, as you need lots of decimals.

Why then does having a shit FPU=bad performance in games? Because that's the way been since the days of the 486 and Pentium, where FPU became common and games required more complex calculations. Note that this example while old, still holds true:
https://www.anandtech.com/show/160/10

Think about it, why does the FX 8350 perform like a slow quad core in most games?

[Dabombinable]

1 hour ago, Prysin said:

hi patrick

Personal attacks are not how you have a debate with people. You should know that by now.

[Bit_Guardian]

33 minutes ago, Dabombinable said:

Why then does having a shit FPU=bad performance in games? Because that's the way been since the days of the 486 and Pentium, where FPU became common and games required more complex calculations. Note that this example while old, still holds true:
https://www.anandtech.com/show/160/10

Think about it, why does the FX 8350 perform like a slow quad core in most games?

The bad game performance is primarily having 1 prefetch for every 2 "cores" and having piss poor cache design. 

[jkampman]

Hey folks, author of the review here! Glad to see the piece spark so much discussion, just wanted to clear up a couple things about it. 

On 11/27/2017 at 3:50 PM, WereCat said:

Single channel 8GB is the issue... most likely

The Envy x360 has dual-channel DDR4-2400 out of the box, so not an issue. 

On 11/28/2017 at 9:27 AM, DocSwag said:

I'm willing to bet the thing is heavily memory bandwidth and latency constrained. I thought HBCC could help with that but it looks like it's not enough.

The Vega IGP on Ryzen Mobile actually doesn't use the HBCC. AMD confirmed that it's using a custom-built graphics memory controller for this chip. 

On 11/28/2017 at 5:13 PM, rawrdaysgoby said:

Second of all why are there even i7-7700HQ numbers in there if we take that out and compare the CPU to their U versions it is exactly what we expected

We always try to bookend our results with a higher-end chip so that people can see what the next thing up the stack gets you. In this review, that meant the i7-7700HQ  

 

Just as a general statement, AMD confirmed it's enabled mobile XFR for the Envy x360, so the chip is actually operating at a 25W TDP instead of 15W. That has major implications for performance-per-watt and overall performance comparisons with Intel parts that we're still accounting for.

 

Thanks for reading and please let me know if I can clarify anything else. 

[Trixanity]

20 minutes ago, jkampman said:

Hey folks, author of the review here! Glad to see the piece spark so much discussion, just wanted to clear up a couple things about it. 

The Envy x360 has dual-channel DDR4-2400 out of the box, so not an issue. 

The Vega IGP on Ryzen Mobile actually doesn't use the HBCC. AMD confirmed that it's using a custom-built graphics memory controller for this chip. 

We always try to bookend our results with a higher-end chip so that people can see what the next thing up the stack gets you. In this review, that meant the i7-7700HQ  

 

Just as a general statement, AMD confirmed it's enabled mobile XFR for the Envy x360, so the chip is actually operating at a 25W TDP instead of 15W. That has major implications for performance-per-watt and overall performance comparisons with Intel parts that we're still accounting for.

 

Thanks for reading and please let me know if I can clarify anything else. 

Why would XFR raise TDP?

Isn't XFR supposed to boost beyond the standard speeds if power and thermals allow it? Similar to desktop that is.

[WMGroomAK]

1 minute ago, Trixanity said:

Why would XFR raise TDP?

Isn't XFR supposed to boost beyond the standard speeds if power and thermals allow it? Similar to desktop that is.

My understanding is that the mobile Ryzen chips are standardized as 15W TDP, however if they have mobile XFR enabled and there is adequate cooling, the mXFR allows it to go up to a 25W TDP load.  Not sure on how much this is affecting performance and battery use as I don't think there is a non-mXFR 2500U laptop out there to compare it to yet.

[NumLock21]

28 minutes ago, jkampman said:

Hey folks, author of the review here! Glad to see the piece spark so much discussion, just wanted to clear up a couple things about it. 

The Envy x360 has dual-channel DDR4-2400 out of the box, so not an issue. 

The Vega IGP on Ryzen Mobile actually doesn't use the HBCC. AMD confirmed that it's using a custom-built graphics memory controller for this chip. 

We always try to bookend our results with a higher-end chip so that people can see what the next thing up the stack gets you. In this review, that meant the i7-7700HQ  

 

Just as a general statement, AMD confirmed it's enabled mobile XFR for the Envy x360, so the chip is actually operating at a 25W TDP instead of 15W. That has major implications for performance-per-watt and overall performance comparisons with Intel parts that we're still accounting for.

 

Thanks for reading and please let me know if I can clarify anything else. 

Is it possible for you guys to do a review on the intel version of the x360, since both of them use the same design. Many sites only focuses on the Spectre.

[jkampman]

21 minutes ago, Trixanity said:

Why would XFR raise TDP?

Isn't XFR supposed to boost beyond the standard speeds if power and thermals allow it? Similar to desktop that is.

Mobile XFR is a different feature from desktop XFR, despite the similarity in naming. Basically, desktop XFR is about using extra thermal headroom to enable higher boost frequencies when possible; mobile XFR (to the best of my understanding) is a specific premium feature that will allow for higher overall performance under sustained workloads. As AMD puts it, mobile XFR means that a laptop so equipped "converts the thermal headroom of a premium notebook’s superior cooling solution into higher average processor frequencies." It specifically requires a cooling solution capable of dissipating 25W, which the x360 has. In short, it is all about that higher TDP.

In practice, what that all seems to mean is that you won't see higher single-core boost over spec (3.6 GHz is it for the Ryzen 5 2500U in Cinebench 1T) but all-core clocks might be way higher than the base clock (the Envy x360 hits 2.6 GHz to 2.8 GHz all-core in Cinebench, compared to the official 2 GHz spec).