The Poll: Allow E.S. or disallow ?

[BenchZowner]

Just a flashback.

It's May 2011 now, and we have yet to see a retail 980X/990X surpass the E.S. @ SuperPi, 3D Mark03/05/Vantage.

[knopflerbruce]

There are retails that are in the same league as the golden A0's (check the 980x and 990x rankings if you don't believe me).

 

Rev. 4 will change the ES game a bit as well, in case you didn't know.

[BenchZowner]

980X - SuperPi 32M ranking:

Places 1 ( 6851MHz ) to 9 = ES

Place 10 ( 6480MHz ) = Retail

 

980X - SuperPi 1M Ranking:

Places 1 ( 6930MHz ) to 6 = ES

Place 7 ( 6589MHz ) = Retail

 

990X - CPU-z Ranking:

Place 1 ( 7092MHz ) = ES

Place 3 ( 6836MHz ) = Retail

 

990X - PiFast Ranking:

Place 1 ( 6946MHz ) = ES

Place 5 ( 6707MHz ) = Retail

 

990X - wPrime 32m Ranking:

Place 1 ( 6654MHz ) = ES

Place 7 ( 5871MHz ) = Retail

 

990X - wPrime 1024m Ranking:

Place 1 ( 6585MHz ) = ES

Place 7 ( 5742MHz ) = Retail

 

Close call ?

[Massman]

*yawn*

 

Sandy Bridge.

[K404]

Who cares about SB?

 

The ES Vs retail debate will keep going with every architecture.... EVERY time ES is better than retail.... until a decision is made.

 

Binned ES is the pinnacle of suckleware.... it should be left for the Pro league at best.

[Massman]

The debate is not 'ES' it's 'binning'.

[richba5tard]

I thought ES was only allowed in the pro oc league in rev4, or am I mistaken?

[knopflerbruce]

Here's the complete list for 980x (top 10) and 990x (top 5):

 

980x:

 

wp32m: 1,2,3,4,5,6,8,9,10: retail, 7: A0 ES

wp1024m: 1,2,3,5,6,7,8,9,10: retail, 4: A0 ES

spi32m: 10: retail, 1,3,4,5,6,7,8,9:A0 ES 2: N/A (thumbnail)

spi1m: 6,8: retail, 1,4,5,7,9: A0 ES, 2,3,10 N/A (thumbnail)

pifast: 3,4,6,8,9,10: retail, 1,7: A0 ES, 2,5: N/A (thumbnail)

cpuz: 1,2,3,5,6,7,9,10: retail, 4,8: A0 ES

 

ES vs retail: 35-19

 

990x:

 

wp32m: 2,3,4,5,: B1 ES, 1: N/A (thumbnail)

wp1024m: 1: retail, 2,3,4,5: B1 ES

spi32m: 3,4: retail, 1,2,5: B1 ES

spi1m: 2: retail, 1,3,4,5: B1 ES

pifast: 5: retail, 2,3,4,5: B1 ES

cpuz: 3: retail, 1,2,4,5: B1 ES

 

Looks like even steven to me, but wprime seems better on retail chips, and superpi better on A0 ES. Very few people have benched 990x retails, not even 20 pifast results in the database - ES included, so it's just the 980x story all over again - when enough chips have been tested, retails will be alot more visible in the rankings.

[BenchZowner]

I think it's more than safe to say that:

 

In this ~2 year period, people have tested MUCH MORE 980X Retails than A0 ES.

Do we really need to see it in numbers like... retails tested ( count the hundreds of RMAs... ? ) 10.000 pieces... ES... 1.000 pieces ?

[El Gappo]

Completely abolishing es's would put a big ass dent in hardware sharing too.

What's going to happen with the FX es's? Can't usually tell none retail amd chips from cpu-z.

[knopflerbruce]

Completely abolishing es's would put a big ass dent in hardware sharing too.

What's going to happen with the FX es's? Can't usually tell none retail amd chips from cpu-z.

 

You can, it says "AMD Engineering sample":)

[El Gappo]

Had a few non retail chips, either before they went retail or chips that never went retail and none have been listed as es in cpu-z.

 

740 would be a good example, or some of the earlier ( and higher )975 results. They always have a different stepping to retail chips i.e CACDC vs AACDC or are they just for reviews and oem, I dunno.

[Massman]

AMD usually has media samples with a different stepping. If the first letter is 'A', it's a media sample ~ AACDC. It shows as retail in CPU-Z. Intel is marking their media samples with (ES) through microcode.

[BenchZowner]

But are they distinguishable clocking-wise ?

If they clock really similarly to the retails ( as in same clocks, not a single advantage over retail ) it shouldn't be a problem at all.

[El Gappo]

You tell me lol, hard to see from the data base as you can't see. 740's clearly outpace 720's by a long way. "Media sample" 975's all hit 7gig which was pretty impressive.

 

Not worth thinking about as you cant tell anyways. Maybe BD will be different.

[Massman]

There were rumours that the 'A' samples were better than the 'C' samples. Like Gappo said ... hard to tell.

[knopflerbruce]

But are they distinguishable clocking-wise ?

If they clock really similarly to the retails ( as in same clocks, not a single advantage over retail ) it shouldn't be a problem at all.

 

No difference from my experience. Best chips from each type listed below:

 

FX53: C

FX55 Clawhammer: C

FX55 San Diego: C

FX57: C

FX60: C

FX62: C

FX64. A (only A's available)

 

Opteron:

 

144: C

146: C

148: C

150: C

152: C

154: C

156. A

165. C

170: L

175: C

180: L

185: C

190: L

 

...and so on. Plus, this "A=ES" theory hasn't been confirmed AFAIK. Early FX60's (retail ones) were ACB2E 0536 chips.

[knopflerbruce]

You tell me lol, hard to see from the data base as you can't see. 740's clearly outpace 720's by a long way. "Media sample" 975's all hit 7gig which was pretty impressive.

 

Not worth thinking about as you cant tell anyways. Maybe BD will be different.

 

My 740 is an A sample, and it's a very good one:) But i've got shitty Phenom A chips as well, just to make that clear:p

[Massman]

Related to this ES debate, I would also pose the question: Why is Intel sending ES samples to media?

 

The famous Dr.Who said a few days before the launch of Sandy Bridge that we shouldn't be looking at the early leaks because they weren't showing the performance of retail silicon. The more I wonder why Intel is not providing the media with retail silicon. And if the silicon of the ES media samples is the same like retail silicon, why would they claim the performance in early leaks are not reflecting retail performance.

 

Extremely Shady business

[BenchZowner]

Intel's not sending ES to all media.

And it also depends on the timeline.

[El Gappo]

No difference from my experience. Best chips from each type listed below:

 

FX53: C

FX55 Clawhammer: C

FX55 San Diego: C

FX57: C

FX60: C

FX62: C

FX64. A (only A's available)

 

Opteron:

 

144: C

146: C

148: C

150: C

152: C

154: C

156. A

165. C

170: L

175: C

180: L

185: C

190: L

 

...and so on. Plus, this "A=ES" theory hasn't been confirmed AFAIK. Early FX60's (retail ones) were ACB2E 0536 chips.

Yeah, but I bet it would look a lot different if you had unlimited access to "A" chips.

Time will play a factor too, I bet the es chips you've got were made before the chip went retail. Revised retail steppings following later on.

Related to this ES debate, I would also pose the question: Why is Intel sending ES samples to media?

 

The famous Dr.Who said a few days before the launch of Sandy Bridge that we shouldn't be looking at the early leaks because they weren't showing the performance of retail silicon. The more I wonder why Intel is not providing the media with retail silicon. And if the silicon of the ES media samples is the same like retail silicon, why would they claim the performance in early leaks are not reflecting retail performance.

 

Extremely Shady business

 

At least they mark them. Someone should write an editorial and oust the buggers

[Massman]

At least they mark them. Someone should write an editorial and oust the buggers

 

Yeah. Imagina what would happen if those ES samples were being sold on Ebay!!

[Christian Ney]

There is going to be more bidders on ES chips than on my Ney's Edition hardware

[El Gappo]

Like dis?

 

Took me all of 5 mins to see what review it came from, looks like a nice chip

[knopflerbruce]

Yeah, but I bet it would look a lot different if you had unlimited access to "A" chips.

Time will play a factor too, I bet the es chips you've got were made before the chip went retail. Revised retail steppings following later on.

 

Well, I wonder why the lucky ones who had access never posted scores that would indicate those chips were better than the C ones?

 

If chips improve with time, then whatever difference between A and C chips would be irrelevant as an argument to not allow A chips as later batches would perform better anyway.