Leaked AMD Ryzen Benchmarks?

[GroundZero]

When every cell phone is a recording device all it would take is one AMD fan to expose Intel if they resort to monopolistic practices.

I don't see it happening this time around. All the companies that are relevant know by now that Ryzen is going to be huge. They want their piece of the pie. When the benchmarks finally come out no one in their right mind would buy Intel for gaming. It will be get XFR or go home.

 

[NKD]

Intel doesn't sells to to IT departments, outside of hosting providers (large) and Data cetners, and volume buyers. You are talking about a tens of thousands of chips per year , if you aren't one of those people, they tell ya to go to OEM's and 3rd party and get from them. They will help you as in when you go to those people, they will tell that OEM, or 3rd party, they can give discounts to the OEM to pass along to the buyer. But the OEM still gets their cut.

I am very doubtful of people that say Intel rep came to me, cause if you aren't the guy that signed the contract or in the same room, you aren't going to know what the actual price that Intel is quoting the company. As I stated, most xeons aren't sold at Ark prices anyways, they sell for around 60% less to volume buyers. Factor that in to Ark prices, this guy is talking crap.

I believe you and I understand. At no point I said he is legit. I just said it wouldn't be surprising to me if Intel does this. That's all. And yes I am not Sure this happened.

 

[OrangeKhrush]

I don't necessarily believe that article but, I do not think it is beyond the realms of intel to resort to underhanded tactics if there is enough erosion of their market share.

 

[Trimlock]

I don't necessarily believe that article but, I do not think it is beyond the realms of intel to resort to underhanded tactics if there is enough erosion of their market share.

If AMD can keep up with stock and maybe take some OEM contracts away from them, I see Intel putting their monetary might into action and make it difficult or impossible for AMD to compete.

 

[Shintai]

http://www.microcenter.com/category/4294966995,4294964566/Intel-Processors

The prices.....they are falling!!!

No they are not. Microcenter got these deals for the last 10 years or so. Check slickdeals or the hot deals forum.

 

[juanrga]

No sure if posted before. Latency and BW benches for Ryzen

http://ranker.sisoftware.net/show_r...e6d5ecdfecdbeec8ba87b791f491ac9cbac9f4c4&l=en

http://ranker.sisoftware.net/show_r...e4dceadceddae3c5b78aba9cf99ca191b7c4f9c9&l=en

 

[lolfail9001]

No sure if posted before. Latency and BW benches for Ryzen

You are posting about Zen being good.

Wow, it must be really good, then.

EDIT: Wait, never mind, you decided to skip cache benches and went with memory ones. Granted, even memory ones are great save for latency going full retard. But we knew that already.

 

[juanrga]

Not exactly a leaked bench, but a rework of David Kanter estimation for SPECint

http://www.linleygroup.com/mpr/article.php?id=11666

He used the 40% target given by AMD plus optimizations to get the next figure

AMD claims that Ryzen has finally a 64% higher IPC than Excavator on SPECint, therefore, we have for 3.7GHz Zen

93.6 * 1.64 = 153.5

EDIT: Note that Kanter mixed IPC claims with rate scores. Below I did a new estimation using non_rate scores.

 

[lolfail9001]

Not exactly a leaked bench, but a rework of David Kanter estimation for SPECint

Kanter's figure uses SPECint_rate without libquantum, however, so AMD's figures by themselves do not apply here. At all.

 

[juanrga]

Kanter's figure uses SPECint_rate without libquantum, however, so AMD's figures by themselves do not apply here. At all.

Yes he mixed IPC target with rate. Edited my post. THX

 

[CSI_PC]

Yeah yeah, WCCFTECH.... but... It looks like what many of us expected. Excellent value for multi-threaded applications, and a big leap, but behind in single threaded applications. These were at stock clocks (didn't mention if it was locked at base or locked at turbo frequencies). Assuming they were locked at base (4.2 GHz and 3.4 GHz on the 1700X) that would make the Zen core about the same clock for clock as Kaby Lake. Then there is the issue of average/max overclocks. If Kaby Lake holds a GHz advantage it should still be a measurably superior chip in single thread performance. The real question is how well will Ryzen overclock.

From what it looks like Ryzen is beat but still performs reasonably well in single threaded applications, however Ryzen buries a similarly priced chip from Intel. To me that makes it an overall better chip (again from what we have seen so far). Ultimately it will come down to your workloads which is the better chip for you.

http://wccftech.com/amd-ryzen-7-1800x-1700x-1700-benchmarks-leaked/

Had to be Khalid
Has he already forgot AMD themselves showed the 6900K with the same single thread performance as the 1800X
Both had 162 at the AMD event, so the leak Khalid is now referencing is with 6900K with Max 3.0 disabled in the BIOS, bit naughty tbh as it is integral to the CPU and why AMD themselves ensured it was on.

Also to say I think juanrga is wrong about all the bells and whistles (XFR/boosting to at least 4.1GHz/etc) and the reason for the 1800X score, critically because 162 is what was scored at the AMD event using the Wraith cooler and lets be honest it is only marginally better than the stock Intel cooler.
Cheers

 

[CSI_PC]

Didn't take Intel long did it

I am really wary about posts like this, what is to stop anyone saying such a thing.
Someone could post I have been approached by AMD and they are offering us large incentives to move away from Intel.
In reality both will be offering incentives for the enterprise sector.
I agree Intel will offer stronger incentives (they have more cash/more leeway with their channels and strategy/critically starting from higher profit margin position/etc), but because of their actions all those years ago they now have to be careful how far they can go with such incentives otherwise it is seen as being anti-competitive and there will be fines at the very minimum.
They may shrug and accept those fines in US if one accepts the logic they will break legal anti-competive laws to stop Ryzen (I tend to think they will not break the anti-competive behaviour laws) but they would be very wary of Europe where there is a habit now of being a pain in the backside of Tech market leaders in essence controlling the market.

So I am dubious of any posts appearing on the internet on this subject, and in reality all large tech companies offer incentives for Enterprises and more so for large contracts, just that Intel can push this more than AMD as Intel started at a much higher profit margin let alone the other factors but they will still have to obey anti-competive behaviour rules and Europe these days can be a right pain for the dominant market leader on this subject.

Cheers

 

[CSI_PC]

Not exactly a leaked bench, but a rework of David Kanter estimation for SPECint

http://www.linleygroup.com/mpr/article.php?id=11666

He used the 40% target given by AMD plus optimizations to get the next figure

AMD claims that Ryzen has finally a 64% higher IPC than Excavator on SPECint, therefore, we have for 3.7GHz Zen

93.6 * 1.64 = 153.5

EDIT: Note that Kanter mixed IPC claims with rate scores.

Based upon model assumptions and estimates way before he could have known any of the actual data performance for Ryzen, good grief that article goes back to August 2016.
It is handy article in many ways, but performance figures as he says relies upon certain estimates, and those assumptions seem to have been too low (especially as AMD in the event clearly stated they achieved higher IPC than originally goal)
Cheers

 

[lolfail9001]

It is handy article in many ways, but performance figures as he says relies upon certain estimates, and those assumptions seem to have been too low.

Nah, the only real faulty assumption was using multithreaded bench as estimation without accounting for CMT penalty or SMT benefit. Had he used ST SPEC, he would probably land right on the mark of where Ryzen is.

In a way, AMD's claim of 40% over Excavator was lowballing it.

 

[CSI_PC]

Nah, the only real faulty assumption was using multithreaded bench as estimation without accounting for CMT penalty or SMT benefit. Had he used ST SPEC, he would probably land right on the mark of where Ryzen is.

In a way, AMD's claim of 40% over Excavator was lowballing it.

LOL, that is what I just added while you responded
And it means the performance estimate in the Kanter chart is not correct to be used now, it was always an estimate and he says that.
Just coming from the OP context.

Edit:
Also he is not using HEDT Intel but 4C/8T compared to Ryzen, in reality needs to be HEDT product line as this is what we have from Ryzen so far.
The Haswell figure should in reality be lower and Broadwell should had been included as Ryzen as released is in essence HEDT equivalent, as I say it is useful in some ways but Kanter did not have enough information to put it into right context for performance let alone that the figures are wrong.
The article is less performance relevent than the performance figures from the AMD event IMO.
Cheers

 

[GroundZero]

I am really wary about posts like this, what is to stop anyone saying such a thing.
Someone could post I have been approached by AMD and they are offering us large incentives to move away from Intel.
In reality both will be offering incentives for the enterprise sector.
I agree Intel will offer stronger incentives (they have more cash/more leeway with their channels and strategy/critically starting from higher profit margin position/etc), but because of their actions all those years ago they now have to be careful how far they can go with such incentives otherwise it is seen as being anti-competitive and there will be fines at the very minimum.
They may shrug and accept those fines in US if one accepts the logic they will break legal anti-competive laws to stop Ryzen (I tend to think they will not break the anti-competive behaviour laws) but they would be very wary of Europe where there is a habit now of being a pain in the backside of Tech market leaders in essence controlling the market.

So I am dubious of any posts appearing on the internet on this subject, and in reality all large tech companies offer incentives for Enterprises and more so for large contracts, just that Intel can push this more than AMD as Intel started at a much higher profit margin let alone the other factors but they will still have to obey anti-competive behaviour rules and Europe these days can be a right pain for the dominant market leader on this subject.

Cheers

I see it as a good thing either way. If it's true then I'm glad it came to light. If not then I am glad attention is being brought to the possibility that it could happen..might give Intel cold feet about pursuing such tactics.

 

[juanrga]

Had to be Khalid
Has he already forgot AMD themselves showed the 6900K with the same single thread performance as the 1800X
Both had 162 at the AMD event, so the leak Khalid is now referencing is with 6900K with Max 3.0 disabled in the BIOS, bit naughty tbh as it is integral to the CPU and why AMD themselves ensured it was on.

Also to say I think juanrga is wrong about all the bells and whistles (XFR/boosting to at least 4.1GHz/etc) and the reason for the 1800X score, critically because 162 is what was scored at the AMD event using the Wraith cooler and lets be honest it is only marginally better than the stock Intel cooler.
Cheers

Precisely the stock cooler is the reason why leaks are reporting that XFR provides only ~100MHz above stock turbo. Custom watter cooler would provide much more than 100MHz extra.

We know turbo was activated in those scores. The doubt is if XFR was activated as well. We know that Ryzen at 3.7GHz does 146. The chart given by WCCFTECH reports 163 for the 1800X

( 163 / 146 ) * 3.7 GHz = 4.13 GHz.

It seems XFR was activated and real turbo was 4.1GHz for the 1800X.

In any case those are minor details. The main point is that the poster whom I replied said that those scores are for Ryzen chips locked at base frequencies, which is not true.

 

[CSI_PC]

Precisely the stock cooler is the reason why leaks are reporting that XFR provides only ~100MHz above stock turbo. Custom watter cooler would provide much more than 100MHz extra.

We know turbo was activated in those scores. The doubt is if XFR was activated as well. We know that Ryzen at 3.7GHz does 146. The chart given by WCCFTECH reports 163 for the 1800X

( 163 / 146 ) * 3.7 GHz = 4.13 GHz.

It seems XFR was activated and real turbo was 4.1GHz for the 1800X.

In any case those are minor details. The main point is that the poster whom I replied said that those scores are for Ryzen chips locked at base frequencies, which is not true.

If I accept your logic then you also need to accept that in the AMD event the 6900K was also setup to use its own features to boost performance.
The chart at WCCFT is pointless for reason I mentioned before.
We know Ryzen does 162 in normal operation meaning, we know 6900K also does 162 with its options on in the BIOS for Max Tech 3.0 boost (I linked the config info of the event on this forum twice and once in its own topic).
You are deliberately skewing this and making wild assumptions about base clock operation.

The ONLY accurate Cinebench results WE KNOW are from the AMD event, we even have the HW and setup for those as well, and we know 6900K was setup in a way that it should be for optimal performance.
AMD never released base clock results so your making assumptions by using other 'independent leaked' results that are not comparable or even reliable to make a point against AMD.....

Not saying any more on this, but as least I know when and where to be critical of AMD and this is not one of them.
Cheers

 

[razor1]

I see it as a good thing either way. If it's true then I'm glad it came to light. If not then I am glad attention is being brought to the possibility that it could happen..might give Intel cold feet about pursuing such tactics.

Definitly not true, why do you think it took so long for Intel's "rebates" to come to light 10 years ago, cause low level people like IT managers aren't involved in those deals.

Those types of discussions are made by senior VP level and up guys. Not by low level managers, Intel wouldn't waste their time with low level managers.

 

[CSI_PC]

Definitly not true, why do you think it took so long for Intel's "rebates" to come to light 10 years ago, cause low level people like IT managers aren't involved in those deals.

Those types of discussions are made by senior VP level and up guys. Not by low level managers, Intel wouldn't waste their time with low level managers.

Yeah any in those meetings have been in the business for quite awhile and they know how incentives work and would not do a post with 'ranting' in it because Intel is adjusting their profit margin on their contract, many businesses will turn it around to put pressure on Intel/etc by using RFP/RFQs.
And these are under NDA for all involved in the process, with such experience those employees tend to not discuss it in a way like that poster even if it was a more casual conversation - IMO anyway.
Although RFP/RFQs seem to work oppsite in government and makes prices go up rather than down like in the private sector

And just to say separately, considering the profit margin of Intel's product and Ryzen's profit margin, Intel and their sales/distribution channels have a lot of flexibility with lowering prices, if Intel decides to do this and that still is not anti-competitive but I really do not think they will slash prices.
Cheers

 

[juanrga]

I reworked fully Kanter's estimations for RyZen. The score of 153.5 in my above post #1056 is incorrect because Kanter's estimation mixed IPC with SPECint_rate_base2006 scores and I used his former estimation as baseline.

Instead using his former estimation as baseline I am starting from scratch. I am now using SPECint_base2006 scores to not mix IPC with SMT and CMT factors.

Starting point is the score for the A10-7850K without libquantum, resulting in an adjusted score of 24.91. Increasing that number by 15% for an Excavator-based design should yield 28.64, and a 64% boost yields 46.97 for RyZen.

The adjusted score for Haswell (again without libquantum) is 44.47.

The above score for RyZen is using Kanter's assumption that Excavator IPC is 15% higher than Steamroller. AMD itself said 5% at ISSCC

http://www.anandtech.com/show/8995/amd-at-isscc-2015-carrizo-and-excavator-details

and reviews showed same IPC or even a slight regression compared to Steamroller

https://semiaccurate.com/2016/05/24/godavari-versus-carrizo-65-watt-showdown/

Alternative scores for Excavator and Ryzen using the 5% boost are 26.16 and 42.89 respectively.

 

[Ocellaris]

I reworked fully Kanter's estimations for RyZen. The score of 153.5 in my above post #1056 is incorrect because Kanter's estimation mixed IPC with SPECint_rate_base2006 scores and I used his former estimation as baseline.

Instead using his former estimation as baseline I am starting from scratch. I am now using SPECint_base2006 scores to not mix IPC with SMT and CMT factors.

Starting point is the score for the A10-7850K without libquantum, resulting in an adjusted score of 24.91. Increasing that number by 15% for an Excavator-based design should yield 28.64, and a 64% boost yields 46.97 for RyZen.

The adjusted score for Haswell (again without libquantum) is 44.47.

The above score for RyZen is using Kanter's assumption that Excavator IPC is 15% higher than Steamroller. AMD itself said 5% at ISSCC

http://www.anandtech.com/show/8995/amd-at-isscc-2015-carrizo-and-excavator-details

and reviews showed same IPC or even a slight regression compared to Steamroller

https://semiaccurate.com/2016/05/24/godavari-versus-carrizo-65-watt-showdown/

Alternative scores for Excavator and Ryzen using the 5% boost are 26.16 and 42.89 respectively.

7/10 napkin math.

 

[lolfail9001]

7/10 napkin math.

OTOH it confirms what we knew half a year ago when GB4 and GB3 for Naples leaked.

Whatever, it is literally 4 days to go until i can compile a list of people whom Intel paid off.

 

[Ocellaris]

OTOH it confirms what we knew half a year ago when GB4 and GB3 for Naples leaked.

Whatever, it is literally 4 days to go until i can compile a list of people whom Intel paid off.

I am just waiting for a full review to "leak" any second now

Like this?

Only include GTA5 and Cinebench scores.

 

[razor1]

Yeah any in those meetings have been in the business for quite awhile and they know how incentives work and would not do a post with 'ranting' in it because Intel is adjusting their profit margin on their contract, many businesses will turn it around to put pressure on Intel/etc by using RFP/RFQs.
And these are under NDA for all involved in the process, with such experience those employees tend to not discuss it in a way like that poster even if it was a more casual conversation - IMO anyway.
Although RFP/RFQs seem to work oppsite in government and makes prices go up rather than down like in the private sector

And just to say separately, considering the profit margin of Intel's product and Ryzen's profit margin, Intel and their sales/distribution channels have a lot of flexibility with lowering prices, if Intel decides to do this and that still is not anti-competitive but I really do not think they will slash prices.
Cheers

Yeah companies will be asking for price negotiations and doing double bid requests.

 

[razor1]

I am just waiting for a full review to "leak" any second now

Like this?

Only include GTA5 and Cinebench scores.

Those min frames are disconcerting.

 

[os2wiz]

Nah, the only real faulty assumption was using multithreaded bench as estimation without accounting for CMT penalty or SMT benefit. Had he used ST SPEC, he would probably land right on the mark of where Ryzen is.

In a way, AMD's claim of 40% over Excavator was lowballing it.

Yes 64% above EXcavator is a huge difference than 40%. I say Broadwell IPC.

 

[os2wiz]

Those min frames are disconcerting.

Listen he ran the 7700K overclocked at 5.0 GHZ. He ran 1700 without overclocking . That is NOT a proper test. Had he overclocked to 4.05 GHZ it would have been a lot higher FPS.

 

[Master_shake_]

7700k 4.2ghz stock 95watt tdp edit:91watts

ryzen 1700 3.4 GHz 65watt tdp.

and you are worried about minimums....

 

[Pieter3dnow]

7700k 4.2ghz stock 95watt tdp edit:91watts

ryzen 1700 3.4 GHz 65watt tdp.

and you are worried about minimums....

Make that 3.0GHZ 3.7GHZ Turbo (single core even)

 

[Ocellaris]

7700k 4.2ghz stock 95watt tdp edit:91watts

ryzen 1700 3.4 GHz 65watt tdp.

and you are worried about minimums....

People are mostly concerned with CPU performance, not a small power difference. Actual power draw can vastly differ from the rated TDP as well.

 

[razor1]

Listen he ran the 7700K overclocked at 5.0 GHZ. He ran 1700 without overclocking . That is NOT a proper test. Had he overclocked to 4.05 GHZ it would have been a lot higher FPS.

The stock 6700k has higher min frames :/

 

[razor1]

7700k 4.2ghz stock 95watt tdp edit:91watts

ryzen 1700 3.4 GHz 65watt tdp.

and you are worried about minimums....

TDP isn't measured the same way on Intel and AMD and when you look at the figures Intel gives which are rounded up, and AMD's which are rounded down you need to keep that in mind, so until we see actual TDP figures from lets say Tom's hardware, I would take that with a grain of salt.

Typical gaming load the 7700k doesn't get any where near that 95 or even 91 watts

 

[Master_shake_]

People are mostly concerned with CPU performance, not a small power difference. Actual power draw can vastly differ from the rated TDP as well.

was more a shot at clock speed variance. also the 7700k will turbo to 4.5.

Make that 3.0GHZ 3.7GHZ Turbo (single core even)

my bad.

 

[Master_shake_]

TDP isn't measured the same way on Intel and AMD, so until we see actual TDP figures from lets say Tom's hardware, I would take that with a grain of salt.

Typical gaming load the 7700k doesn't get any where near that 95 or even 91 watts

so if the 91 watts is a little to liberal on intel does that mean that the 65 watts its too conservative on AMD?

 

[razor1]

so if the 91 watts is a little to liberal on intel does that mean that the 65 watts its too conservative on AMD?

Don't know yet, gotta wait and see, but Intel the way they do their TDP, they test numerous applications and average them and round up to the socket, AMD does the opposite they round down in the past, so end results, they might actually end up around the same, if AMD is still calculating their TDP the way they did with BD.

 

[Ocellaris]

so if the 91 watts is a little to liberal on intel does that mean that the 65 watts its too conservative on AMD?

Nope they are both BS My 7700K can use just under 130W when running AVX instructions Then it barely goes over 30W playing games, it's all over the place.

 

[juanrga]

If I accept your logic then you also need to accept that in the AMD event the 6900K was also setup to use its own features to boost performance.
The chart at WCCFT is pointless for reason I mentioned before.
We know Ryzen does 162 in normal operation meaning, we know 6900K also does 162 with its options on in the BIOS for Max Tech 3.0 boost (I linked the config info of the event on this forum twice and once in its own topic).
You are deliberately skewing this and making wild assumptions about base clock operation.

The ONLY accurate Cinebench results WE KNOW are from the AMD event, we even have the HW and setup for those as well, and we know 6900K was setup in a way that it should be for optimal performance.
AMD never released base clock results so your making assumptions by using other 'independent leaked' results that are not comparable or even reliable to make a point against AMD.....

Not saying any more on this, but as least I know when and where to be critical of AMD and this is not one of them.
Cheers

The chart at WCCFTECH can be useful if one knows what is being measured.

I am NOT making "wild assumptions about base clock operation". And I am NOT making a point against AMD.

I will try to explain this a last time. TordanGow reported the chart and did the assumption that chips were locked at base clocks; he then concluded that Ryzen has about the same IPC than Kabylake: "Assuming they were locked at base (4.2 GHz and 3.4 GHz on the 1700X) that would make the Zen core about the same clock for clock as Kaby Lake." In my reply to him I noted that turbo was activated.

The doubt is if XFR is activated or not in that chart. If it is not activated the score of 163 is for 4GHz RyZen. If XFR is activated the frequency for that score could be 4.1GHz or even 4.2GHz. Using a independent score for a 3.7GHz Ryzen we obtain that a score of 163 is compatible with a 4.1GHz clock. This is compatible with XFR being activated in the WCCFTECH chart.

But again it is a minor point if XFR is activated or not. This is statistically insignificant 2.5% variation. The main point is that stock turbo was activated, contrary to his assumption.

 

[Magic Hate Ball]

I am just waiting for a full review to "leak" any second now

Like this?

Only include GTA5 and Cinebench scores.

MFW someone posts OC benchmark numbers for one piece of hardware vs only stock on another:

Also, curious if he ran into the memory configuration issues for running at speed. If he got the memory to run @ 3000mhz for the Ryzen or not given all we've heard in rumors.

 

[os2wiz]

The chart at WCCFTECH can be useful if one knows what is being measured.

I am NOT making "wild assumptions about base clock operation". And I am NOT making a point against AMD.

I will try to explain this a last time. TordanGow reported the chart and did the assumption that chips were locked at base clocks; he then concluded that Ryzen has about the same IPC than Kabylake: "Assuming they were locked at base (4.2 GHz and 3.4 GHz on the 1700X) that would make the Zen core about the same clock for clock as Kaby Lake." In my reply to him I noted that turbo was activated.

The doubt is if XFR is activated or not in that chart. If it is not activated the score of 163 is for 4GHz RyZen. If XFR is activated the frequency for that score could be 4.1GHz or even 4.2GHz. Using a independent score for a 3.7GHz Ryzen we obtain that a score of 163 is compatible with a 4.1GHz clock. This is compatible with XFR being activated in the WCCFTECH chart.

But again it is a minor point if XFR is activated or not. This is statistically insignificant 2.5% variation. The main point is that stock turbo was activated, contrary to his assumption.

So with turbo activated that would only be a10% advzntage. If you deduct 10% off Kaby Lake IPC you get A drop higher than Broadwell for Ryzen IPC.